Studies in Big Data 67

Prasant Kumar Pattnaik

Raghvendra Kumar
Souvik Pal Editors

Internet
of Things
and Analytics
for Agriculture,
Volume 2
Studies in Big Data

Volume 67

Series Editor
Janusz Kacprzyk, Polish Academy of Sciences, Warsaw, Poland
The series “Studies in Big Data” (SBD) publishes new developments and advances
in the various areas of Big Data- quickly and with a high quality. The intent is to
cover the theory, research, development, and applications of Big Data, as embedded
in the fields of engineering, computer science, physics, economics and life sciences.
The books of the series refer to the analysis and understanding of large, complex,
and/or distributed data sets generated from recent digital sources coming from
sensors or other physical instruments as well as simulations, crowd sourcing, social
networks or other internet transactions, such as emails or video click streams and
other. The series contains monographs, lecture notes and edited volumes in Big
Data spanning the areas of computational intelligence including neural networks,
evolutionary computation, soft computing, fuzzy systems, as well as artificial
intelligence, data mining, modern statistics and Operations research, as well as
self-organizing systems. Of particular value to both the contributors and the
readership are the short publication timeframe and the world-wide distribution,
which enable both wide and rapid dissemination of research output.
** Indexing: The books of this series are submitted to ISI Web of Science, DBLP,
Ulrichs, MathSciNet, Current Mathematical Publications, Mathematical Reviews,
Zentralblatt Math: MetaPress and Springerlink.

More information about this series at http://www.springer.com/series/11970

Prasant Kumar Pattnaik
•

Raghvendra Kumar Souvik Pal

•

Editors

Internet of Things
and Analytics for Agriculture,
Volume 2

123
Editors
Prasant Kumar Pattnaik Raghvendra Kumar
School of Computer Engineering Department of Computer Science
KIIT Deemed to be University and Engineering
Bhubaneswar, Odisha, India LNCT College
Jabalpur, Madhya Pradesh, India
Souvik Pal
Department of Computer Science
and Engineering
Brainware University
Kolkata, West Bengal, India

ISSN 2197-6503 ISSN 2197-6511 (electronic)

Studies in Big Data
ISBN 978-981-15-0662-8 ISBN 978-981-15-0663-5 (eBook)
https://doi.org/10.1007/978-981-15-0663-5
© Springer Nature Singapore Pte Ltd. 2020
This work is subject to copyright. All rights are reserved by the Publisher, whether the whole or part
of the material is concerned, specifically the rights of translation, reprinting, reuse of illustrations,
recitation, broadcasting, reproduction on microfilms or in any other physical way, and transmission
or information storage and retrieval, electronic adaptation, computer software, or by similar or dissimilar
methodology now known or hereafter developed.
The use of general descriptive names, registered names, trademarks, service marks, etc. in this
publication does not imply, even in the absence of a specific statement, that such names are exempt from
the relevant protective laws and regulations and therefore free for general use.
The publisher, the authors and the editors are safe to assume that the advice and information in this
book are believed to be true and accurate at the date of publication. Neither the publisher nor the
authors or the editors give a warranty, expressed or implied, with respect to the material contained
herein or for any errors or omissions that may have been made. The publisher remains neutral with regard
to jurisdictional claims in published maps and institutional affiliations.

This Springer imprint is published by the registered company Springer Nature Singapore Pte Ltd.
The registered company address is: 152 Beach Road, #21-01/04 Gateway East, Singapore 189721,
Singapore
Preface

The main aim of this book is to bring together leading academic scientists,
researchers, and research scholars to exchange and share their experiences and
research results on all aspects of wireless IoT and analytics for agriculture. It also
provides a premier interdisciplinary platform for researchers, practitioners, and
educators to present and discuss the most recent innovations, trends, and concerns
as well as practical challenges encountered and solutions adopted in the fields of
IoT and analytics for agriculture. This book is organized into 13 chapters.
Chapter “IoT-Based Soil Health Monitoring and Recommendation System”
discusses a soil health monitoring system in which a farmer will be able to monitor
soil moisture, soil temperature, and soil pH in his android smartphone. The farmer
will also get the recommendations of lime and sulfur on the basis of pH of the soil.
The area of study is Jaipur, Rajasthan (26.9124° N, 75.7873° E). The proposed
system is tested with the results taken from authorized laboratories. The proposed
system is validated with T-test having no significant difference between calibrated
values and laboratory recorded values. The proposed system is implemented on
android smartphone, so it is useful to farmers, agriculture scientists, agriculture
professionals, and IoT experts.
Chapter “Implementing IoT and Wireless Sensor Networks for Precision
Agriculture” affords the deployment of an Internet of things (IoT)-based precision
agriculture technology using wireless sensor networks (WSNs) in crop manage-
ment. An online crop management system (OCMS) was developed to complement
the IoT platform layer. The application aimed at providing the farm environmental
conditions including the temperature, water level, pH value, and dissolved oxygen
measurements from the paddy field with food crops and horticulture crops being
grown. Further studies could be done by analyzing the inter- and intra-field vari-
abilities such as soil herbicide weed control, side dressing, late blight control, and
haulm killing.
Chapter “Bayesian Network for Development of Expert System in Pest
Management” discusses the application of Bayesian network which may greatly
assist in development of innovative methods of selecting appropriate management
option based on uncertain or tentative information of identified agroecological

v
vi Preface

factors besides crop condition that farmers provide. Application of the methods thus
developed in information and communication technology (ICT)-based expert sys-
tems (ESs) of agriculture would automate and speed up the process of providing
real-time decision support on insect pest management to the farmers.
Chapter “An IoT and Wireless Sensor Network-Based Technology for a
Low-Cost Precision Apiculture” presents the foundational space of apiculture,
challenges, and emerging techniques of the apiarists in beekeeping including pre-
cision apiculture and a low-cost IoT and remote sensing technology deployed for
data gathering and monitoring the vigor and productivity of beehive colonies. The
system has redundancy components and utilized the three-level hierarchical model
including wireless node, local information server, and the cloud data server. The
system also consists of sensors deployed across distinct locations to capture the
beehives’ temperature, relative humidity, and loads of the hives and honeycombs,
and the information is stored to the local and cloud server. This system will
complement the urgent need to countervail the honeybee colonies' collapse by
helping the apiarists, apiculture sub-sector, and agriculture sector to embark
emerging precision apiculture technologies.
Chapter “Adaptive Encryption Technique for Collaborative Cloud Environments”
presents a realistic solution to the problem where numeric data are transformed into
alphanumeric types, and hence, there is minimal probability of storing encrypted
data in the existing numeric field. The proposed algorithm allows translucent
verification of intensity encryption that does not modify the data field category or
the fixed time span.
Chapter “Role of Internet of Things and IT in Health Care” explores in depth the
importance of the Internet of things devices in health care and also the role of
information technology in managing the massive volume of high-security medical
information of patients. This chapter proposes few IoT applications in rural health
care and ways to boost primary health desires of the developing nations.
Chapter “Adopting Big Data Analysis in the Agricultural Sector: Financial and
Societal Impacts” summarizes the state-of-the-art analytical methods that have been
recently used in the agriculture industry, along with their financial and societal
impacts.
Chapter “AI-Based Yield Prediction and Smart Irrigation” acknowledges the
past breakthroughs and emerging artificial intelligence-based techniques in preci-
sion farming specifically for yield prediction and smart irrigation. Artificial
intelligence-based system provides sufficient information about crop yields at an
early stage, and its associated smart irrigation management system is effective in
judicious use of essential resources such as water and energy for agriculture.
Chapter “IoT-Based Smart Tree Management Solution for Green Cities” dis-
cusses key challenges faced by transformation of smart cities and integration of
green technologies, initiatives, and existing systems that are in place to solve these
challenges. With overpopulation, fast urbanization, and environmental changes,
powerful administration of trees and plants is required like never before. We present
Preface vii

an Internet of things (IoT)-based smart tree management architecture for cities

which is able to track various characteristics of a given tree such as the air quality,
sunlight level, sound pollution level, and other important characteristics for efficient
city planning, afforestation plans, and higher-quality produce.
Chapter “Data Analysis and Prediction Using Big Data Analytics in Agriculture”
is composed of four sections. The first section provides an overview of analytics on
big data in agriculture. This is followed by the second section which highlights the
different big data techniques in agriculture. The third section describes two case
studies that apply techniques for big data analysis in the agricultural domain. The
fourth section analyzes the various research areas related to agriculture that can be
explored by the academicians and researchers. The summary of the chapter is
provided at the end.
Chapter “Generic IoT Platform for Analytics in Agriculture” aims in creating a
universal platform used especially in the field of agriculture for data transmission
and reception from various sensors and plotting the statistical graphs on a dash-
board. Apart from this as a purpose of providing easy usage and extensibility, smart
devices can also be installed, which provide the plug-and-play capability of our
system.
Chapter “Mission SDGs Through Food Waste Management: Nature and
Approaches” attempts to assess and account the loss of economy for wastage of
food items from global canvas to national perspective. This study has also focused
on how to use IoT platform so that the food wastage can be reduced up to a
considerable amount both in the supply chain and even in the household practices.
This research work is based on secondary information like research papers, reports,
and results of other relevant studies. The chapter has attempted to develop and
devise a conceptual and strategic model where the IoT ecosystem can be incor-
porated to ensure real-time solutions and to curb on massive food wastage practices.
If the model is implemented and practiced with appropriate case-specific modifi-
cations and customizations, it would make the economy more efficient and address
different perspectives and dimensions of UN Sustainable Development Goals
(UNSDGs) to a larger extent primarily in the Indian context.
Chapter “Frost Prediction in Highland Crops Management Using IoT-Enabled
System and Multiple Regression” discusses the concepts of precision agriculture,
applications of the Internet of things (IoT) in agriculture, the design and challenges
of an IoT-enabled system for highland crop management, and utilized multiple
regression as a frost prediction technique. The remote sensing devices deployed in
the farm collect the frost climatic events including air velocity, temperature, and
humidity, and transmit the information to the Web server. The system provides
early warnings and frost forecast for the farmers using SMS. The system helped
farmers as a useful resource to conduct frost protection activities on the farm,
thereby reducing the frost harm to the crops.
We are sincerely thankful to the Almighty for supporting and standing at all
times with us, whether it is good or tough times and given ways to concede us.
Starting from the call for chapters till the finalization of chapters, all the editors have
given their contributions amicably, which is a positive sign of significant teamwork.
viii Preface

The editors are sincerely thankful to all the members of Springer, especially Prof.
Aninda Bose for the providing constructive inputs and allowing an opportunity to
edit this important book. We are equally thankful to reviewers who hailed from
different places in and around the globe, shared their support, and stood firm toward
quality chapter submission.

Bhubaneswar, India Prasant Kumar Pattnaik

Jabalpur, India Raghvendra Kumar
Kolkata, India Souvik Pal
Contents

IoT-Based Soil Health Monitoring and Recommendation System . . . . . . 1

Vaibhav Bhatnagar and Ramesh Chandra
Implementing IoT and Wireless Sensor Networks for Precision
Agriculture . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
D. D. Dasig Jr.
Bayesian Network for Development of Expert System in Pest
Management . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
Niranjan Singh and Neha Gupta
An IoT and Wireless Sensor Network-Based Technology
for a Low-Cost Precision Apiculture . . . . . . . . . . . . . . . . . . . . . . . . . . . 67
D. D. Dasig Jr. and J. M. Mendez
Adaptive Encryption Technique for Collaborative Cloud
Environments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93
Y. Harold Robinson, E. Golden Julie and A. Vegi Fernando
Role of Internet of Things and IT in Health Care . . . . . . . . . . . . . . . . . 119
Rohit Sharma and Rajendra Prasad Mahapatra
Adopting Big Data Analysis in the Agricultural Sector:
Financial and Societal Impacts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 131
Rasha Kashef
AI-Based Yield Prediction and Smart Irrigation . . . . . . . . . . . . . . . . . . 155
Deepak Sinwar, Vijaypal Singh Dhaka, Manoj Kumar Sharma
and Geeta Rani
IoT-Based Smart Tree Management Solution for Green Cities . . . . . . . 181
Bilal Shabandri, Sahith Reddy Madara and Piyush Maheshwari

ix
x Contents

Data Analysis and Prediction Using Big Data Analytics

in Agriculture . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 201
K. Aditya Shastry and H. A. Sanjay
Generic IoT Platform for Analytics in Agriculture . . . . . . . . . . . . . . . . . 225
Balasubramani Pradeep, R. Balasubramani, Jason Elroy Martis
and M. S. Sannidhan
Mission SDGs Through Food Waste Management:
Nature and Approaches . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 249
Arindam Chakrabarty and Uday Sankar Das
Frost Prediction in Highland Crops Management
Using IoT-Enabled System and Multiple Regression . . . . . . . . . . . . . . . 261
J. M. Mendez and D. D. Dasig
About the Editors

Prasant Kumar Pattnaik Ph.D (Computer Science), Fellow IETE, Senior

Member IEEE is a Professor at the School of Computer Engineering, KIIT Deemed
University, Bhubaneswar. He has more than a decade of teaching and research
experience. Dr. Pattnaik has published numbers of Research Papers in
peer-reviewed International Journals and Conferences. He also published many
edited book volumes in Springer and IGI Global Publication. His areas of interest
include Mobile Computing, Cloud Computing, Cyber Security, Intelligent Systems
and Brain Computer Interface. He is one of the Associate Editor of Journal of
Intelligent & Fuzzy Systems, IOS Press and Intelligent Systems Book Series Editor
of CRC Press, Taylor Francis Group.

Raghvendra Kumar is working as Associate Professor in Computer Science and

Engineering Department at L.N.C.T Group of College Jabalpur, M.P. India. He
received B. Tech, M.Tech and Ph.D. in Computer Science and Engineering, India,
and Postdoc Fellow from Institute of Information Technology, Virtual Reality and
Multimedia, Vietnam National University (VNU), Vietnam. He serves as Series
Editor Internet of Everything (IOE): Security and Privacy Paradigm, Green
Engineering and Technology: Concepts and Applications, publishes by CRC press,
Taylor & Francis Group, USA and Bio-Medical Engineering: Techniques and
Applications, Publishes by Apple Academic Press, CRC Press, Taylor & Francis
Group, USA. He also serves as acquisition editor for Computer Science by Apple
Academic Press, CRC Press, Taylor & Francis Group, USA.

Souvik Pal Ph.D, MCSI; MCSTA/ACM, USA; MIAENG, Hong Kong; MIRED,
USA; MACEEE, New Delhi; MIACSIT, Singapore; MAASCIT, USA is Associate
Professor at the Department of Computer Science and Engineering, Brainware
University, Kolkata, India. Dr. Pal received his B. Tech degree in Computer
Science and Engineering from West Bengal University of Technology, Kolkata.
He has received his M. Tech and PhD. degree in Computer Engineering from
KIIT University, Bhubaneswar, India. He has worked as Assistant Professor in
Nalanda Institute of Technology, Bhubaneswar, and JIS College of Engineering,

xi
xii About the Editors

Kolkata (NAAC “A” Accredited College). He has also worked as head of the
Computer Science Department in Elitte College of Engineering, Kolkata. Dr. Pal
has published several research papers in Scopus-indexed International journals and
conferences.
IoT-Based Soil Health Monitoring
and Recommendation System

Vaibhav Bhatnagar and Ramesh Chandra

Abstract Soil is the base of agriculture. Soil provides nutrients that increase the
growth of a crop. Some chemical and physical properties of soil, such as its moisture,
temperature and its pH, heavily affect the yield of a crop. These properties can be
sensed by the open-source hardware, and they can be used in the field. In this chapter,
a soil health monitoring system is proposed in which farmer will be able to monitor
soil moisture, soil temperature and soil pH in his android smart-phone. The farmer
will also get the recommendations of lime and sulphur on the basis of pH of the soil.
The area of study is Jaipur, Rajasthan (26.9124° N, 75.7873° E). The proposed system
is tested with the results taken from authorized laboratories. The proposed system is
validated with t-test having no significance difference between calibrated values and
laboratory-recorded values. The proposed system is implemented on android smart-
phone so it is useful for farmers, agriculture scientists, agriculture professionals and
IoT experts.

Keywords Internet of things · Fertilization · Soil health

1 Introduction

Soil is the upper most layer of the earth. Agriculture, which is the prime component
of livelihood, cannot be expected without soil. There are so many chemical and
physical properties of soil that affect the growth of crops such as soil moisture, soil
pH, soil temperature, soil electrical conductivity and its humidity. Water present in
the soil is known as soil moisture. Soil moisture acts as carrier of nutrients for crop
and plant growth. Moreover, it itself acts as nutrient for crops and plants; it also helps

V. Bhatnagar (B)
Department of Computer Science, S.S. Jain Subodh PG College, Jaipur, India
e-mail: [email protected]
R. Chandra
Cyber-Physical Systems Laboratory, Department of ICT and Natural Sciences, Faculty of
Information Technology and Electrical Engineering NTNU - Norwegian University of Science
and Technology, Ålesund, Norway

© Springer Nature Singapore Pte Ltd. 2020 1

P. K. Pattnaik et al. (eds.), Internet of Things and Analytics
for Agriculture, Volume 2, Studies in Big Data 67,
https://doi.org/10.1007/978-981-15-0663-5_1
2 V. Bhatnagar and R. Chandra

to maintain the soil temperature [1]. Soil temperature is referred as warmth in the
soil. It helps to maintain the life in the soil of germs. The ideal for soil organisms lies
between 25 and 350 °C [1]. Soil humidity is the water present in the soil in vapour
form. Soil humidity plays an important role in evapotranspiration that directly affects
crop’s water need. Soil pH is most prominent factor responsible for successful growth
of crops and plants. As it is commonly known that range of pH lies between 0 and
14, and distilled water has exactly 7 pH (neutral). The ideal pH for crop growth lies
between 6.0 and 7.0. If a sample of soil has pH less than 6.0, then it is acidic in nature
and needs to be raised with the help of different liming material, on the other hand
if sample of soil has pH greater, then it is alkaline in nature and needs to be lowered
with the help of sulphur and gypsum. Table 1a and b shows fertilizer requirement
for maintaining the pH level [2].
The objectives of the chapter are to show the current status of soil moisture, soil
temperature, humidity and soil pH in the smart-phone of farmer. To show the pH of
the soil, the pH sensor is needed to be calibrated with the soil which is the second
objective, and the last objective is to recommend the fertilizer on the basis of pH of
soil. The area of study is Jaipur, Rajasthan (26.9124° N, 75.7873° E).
This chapter is divided into five sections—it begins with introduction, the second
part consists of literature review and research motivation, third part consists of design
of proposed system, fourth part consists of results and discussion, and fifth part
consists of conclusion and future work.

Table 1 a Limestone needed to raise pH. b Sulphur needed to lowering pH

(a) Approximate amount of finely ground limestone needed to raise the pH of a 7-inch layer of
soil, tones per acre
Soil texture From pH 4.5 to 5.5 From pH 5.5 to 6.5
Sand and loam sand 0.5 0.6
Sandy loam 0.8 1.3
Loam 1.2 1.7
Silt loam 1.5 2.0
Clay loam 1.9 2.3
Muck 3.8 4.3
(b) Tons of sulphur needed per acre to lower pH to 6.5 (1 ton of sulphur)
Original pH Sandy soil Clay soil
8.5 0.7–1.0 1.0–1.3
8.0 0.5–0.7 0.7–1.1
7.5 0.2–0.3 0.4–0.5
IoT-Based Soil Health Monitoring and Recommendation System 3

2 Related Work

In the year 2017, Encinas et al. [3] presented a prototype model of distributed mon-
itoring system of different properties of aquaculture water quality, such as pH, tem-
perature and availability of oxygen. The proposed prototype model is useful for
improving environmental control, reduction of production cost, etc. Authors used
Arduino sensors, Zigbee technology, Arduino module, web services, mobile appli-
cation and database to design the system. The prototype is shown in Fig. 1 [3].
Using the Arduino module, sensors sense pH, temperature and availability of
oxygen and sensed value which are transmitted to external cloud with the help of
database and different web services. Smart-phones are connected through the cloud
by which these sensed data can be visualized. As the future work, alarm conditions
are to be applied in the system with help of artificial intelligence.
In the year 2009, Coyle et al. [4] designed and calibrated textile-based wearable
sensor for sports performance. These sensors are comfortable in wearing in that
manner a person can do exercise and can play a sport. To analyse the rehydration,
the fabric-based pH sensor is used. This pH sensor used sweat of the body to sense
the rehydration. This special sensor is depicted in Fig. 2 [4].
This above-mentioned pH sensor is calibrated with the help of artificial sweat
of range 4–8 pH. Authors calibrated three times and illustrated that this pH sensor
is sensitive to changes in pH of less 0.2 units and that the response is repeatable.
LilyPad Arduino is used to control the sensors attached to the body.

Fig. 1 Prototype model of distributed monitoring system of different properties of aquaculture

water quality
4 V. Bhatnagar and R. Chandra

Fig. 2 Special pH sensor for sweat measurement

Yamaguchi et al. [5] presented a sensor network-based agriculture project called

as e-kakashi project using different sensors such as camera, Zigbee module, temper-
ature and humidity sensor. It was basically designed for improvement of productivity
in farming system and reduction of vermin disease. This application was designed for
iOS operating system. Lee et al. [6] presented IoT-designed agricultural production
system for stabilizing the demand and supply of agricultural related products while
developing the required sensors and prediction system for the growth and produc-
tion amount of crops. The authors used HSDPA/Dual CDMA, TCP/UDP/IP/ICMP
Protocol Support, 2.4 GHz IEEE 802.15.4 compatible, temperature sensor, humidity
sensor, soil EC senor and pH sensor.

3 Architecture of Proposed System

The proposed system is two-layer architecture. First layer is data acquisition and
calibration layer, and second layer is fertilization recommendation layer. The con-
nectivity between these two layers is done by Bluetooth communication system
(IEEE 802.15) [7]. Figure 3 depicts the overview of proposed system.

3.1 Data Acquisition and Calibration Layer

In the proposed system, data acquisition and calibration layer is responsible for
acquiring raw data from the field using sensors, calibration of the input and sending
the data to fertilization recommendation layer. In the proposed system, Arduino Uno
[8] microprocessor is used. DHT11, SEN: 0161 and SEN: 0193 sensors are used in
the system.
IoT-Based Soil Health Monitoring and Recommendation System 5

Fig. 3 Overview of proposed system

3.2 Soil Temperature and Humidity Sensor

Here DHT11 [9] sensor is used. DHT11 sensor is a analogue sensor which is used for
sensing the soil temperature and soil humidity. Since library of DHT11 is directly
available, this sensor generates the calibrated values, so manual calibration is not
required. This sensor gives the values of both soil humidity and temperature. The
sketch code of this sensor shown in Figs. 4 and 5 illustrates the DHT11 sensor [9].

3.3 pH Sensor

SEN: 0161 pH sensor [10] is used. SEN: 0161 sensor is typical analogue sensor
which is used to sense the pH of soil. This sensor has been calibrated for soil of
Rajasthan. The sensor is calibrated from 0 to 12 values. SEN: 0161 sensor is shown
in Fig. 6 [10].
6 V. Bhatnagar and R. Chandra

Fig. 4 Sketch code of DHT11 sensors

Fig. 5 DHT11 sensor

IoT-Based Soil Health Monitoring and Recommendation System 7

Fig. 6 SEN: 0161 sensor

The following steps are used for calibration:

• Selection of Materials and Instruments: To perform the calibration the pH rapid
tester (Yeslike Digital PH Meter), oxalic acid [11] is used as acid agent, and
potassium bicarbonate [12] is used as alkaline agent. pH buffer tablets of 7.00 are
used along with 100 ml water with the help of measuring cylinder. A total of 21
samples ranging 0–12 pH are drawn for calibration.
• Capturing the Reading: The next step is to read the sensor values from the
samples that are arranged in last step. For acquiring the inputs, Table 2 depicts the
requirements.
SD-card data logger [13] is used to save the values sensed by both the sensors.
The reason behind using data logger is that the readings of sensors fluctuate with

Table 2 Requirement for acquiring input

S. No. Name of device Use
1 Arduino Uno Microcontroller
2 D-cable Connect Arduino to computer
3 SD-card data logger To save data sensed by the sensors
4 Jumper wire Male–male, female–male, female–female
5 Sketch IDE for development
6 Breadboard For connecting two or more input sensors
8 V. Bhatnagar and R. Chandra

certain range of values. In this experiment, 500 readings of every experiment are
being recorded. Same Sketch is used to develop the code for sensing and recording
the values of both the sensors.
All the values are saved in.CSV file that are convenient to open in MS Excel. This
sketch code has been executed for 21 times, and 500 values have been saved. Mean
values of all 500 values of 21 samples are depicted in Table 3.
• Selection of Soil Samples and Laboratory Testing: The next step is to collect the
soil samples for development of decision support system. The soil samples have
been collected from Jaipur Rural (26.9124° N, 75.7873° E), Sambhar (26.9096° N,
75.1859° E), Kota (25.2138° N, 75.8648° E) and Jhalawar (24.5973° N, 76.1610°
E) regions. Soil pH of these samples was identified in authorized soil testing lab.
The pH of soil samples is shown in Table 4.
• Reading the Values of Soil from Sensor: The next step is to read the sensor
value from soil. All the six soil samples are dissolved in water, 100 readings were
taken from each sample, and average has been calculated. The values are shown
in Table 5.
Now the next step is to match these values with actual values received in above
tables. These values are shown in Table 6 and Fig. 7.

Table 3 Mean values of pH

S. No. pH value Mean value received
sensor
1 7 520
2 7.5 480
3 8 440
4 8.5 400
5 9 360
6 9.5 320
7 10 300
8 10.5 310
9 11 305
10 11.5 305
11 12 300
12 6.5 565
13 6 610
14 5.5 655
15 5 700
16 4.5 745
17 4 785
18 3.5 827
19 3 872
20 2.5 871
21 2 873
IoT-Based Soil Health Monitoring and Recommendation System 9

Table 4 pH of soil samples

S. No. Soil type Soil location Soil pH
1 Clay Jhalawar 7.9
2 Clay (yellow) Kota 8.2
3 Clay (black) Jhalawar 8.0
4 Loam Sambhar 8.0
5 Silt sand Kalwar, Jaipur 8.2
6 Pure sand Jagatpura 7.8

Table 5 Mean values of pH of soil taken from pH sensor

S. No. Soil type Mean value
1 Clay 445
2 Clay (yellow) 435
3 Clay (black) 439
4 Loam 437
5 Silt sand 432
6 Pure sand 452

Table 6 Laboratory values and sensor values of pH of soil

S. No. Soil type Actual pH Sensor value Nearest pH value from
1 Clay 7.9 445 8
2 Clay (yellow) 8.2 435 8
3 Clay (black) 8.0 439 8
4 Loam 8.0 437 8
5 Silt sand 8.2 432 8
6 Pure sand 7.8 452 8

• Setting the Values: these values are set with the pH values by taking interval of
10. These values will be sent to the smart-phone and this table will be saved in the
database of smart-phone (SQLite). The set values are shown in Table 7.

3.4 Soil Moisture

Here capacitance-based soil moisture sensor is used, i.e. SEN: 0193 [14]. This sensor
is also a analogue sensor which is used to sense soil moisture. It gives value from
0 to 1023; therefore, calibration is required. The calibration of this sensor is also
done with the same soil samples and with same data logger. In order to prepare the
10 V. Bhatnagar and R. Chandra

Fig. 7 Laboratory values and sensor values of pH of soil

Table 7 Set values of pH of soil

S. No. pH value Range
1 9.5 and greater 320 below
2 9.1–9.4 321–349
3 9 350–370
4 8.6–8.9 371–389
5 8.5 390–410
6 8.1–8.4 411–429
7 8 430–450
8 7.6–7.9 451–469
9 7.5 470–490
10 7.1–7.4 491–509
11 7 510–530
12 6.6–6.9 531–554
13 6.5 555–575
14 6.6–6.1 576–599
15 6 600–620
16 5.6–5.9 621–644
17 5.5 645–665
IoT-Based Soil Health Monitoring and Recommendation System 11

first sample, a random soil sample of large amount is collected from the field of
more than 20 kg in weight. This 20 kg of soil is put in the oven so that the moisture
already present inside the soil is evaporated by using gravimetric method [15]. Now
from this sample of 20 kg soil, 500 g soil has been taken. Now this sample has 0%
moisture, and the soil and sensors are inserted for reading the values. Further again,
500 g of soil has been taken from the collected sample, and 25 ml water is added to
this sample. Now this sample has 5% moisture content, and sensors are put in the
soil for reading the values. This same process has been carried out for 21 times. The
last sample has 500 g of soil with 500 ml water content and has 100% moisture, and
sensors are put for reading the values as shown in Tables 8, 9 and 10.
After the calibration, the physical properties sensed by different sensors would be
sent to the smart-phone using Bluetooth communication [16] with the help HC-05.
HC-05 is a Bluetooth chip that will be used to bridge the gap between data acquisition
and calibration layer and fertilization recommendation layer. This Bluetooth chip is

Table 8 Soil moisture calibration of loam soil

S. No. Size of sample Water content Moisture Sensor value Range value
(g) added content, %
water content
1 500 Nil 0 560 1000–560
2 500 25 ml 5 500 559–500
3 500 50 ml 10 400 499–400
4 500 75 ml 15 349 399–349
5 500 100 ml 20 342 348–342
6 500 125 ml 25 280 341–280
7 500 150 ml 30 279 279–279
8 500 175 ml 35 274 278–277
9 500 200 ml 40 269 276–269
10 500 225 ml 45 265 268–265
11 500 250 ml 50 254 264–264
12 500 275 ml 55 253 263–253
13 500 300 ml 60 251 252–251
14 500 325 ml 65 250 250–250
15 500 350 ml 70 248 249–248
16 500 375 ml 75 248 247–248
17 500 400 ml 80 247 247–247
18 500 425 ml 85 247 247–247
19 500 450 ml 90 240 246–240
20 500 475 ml 95 240 246–240
21 500 500 ml 100 240 246–240
12 V. Bhatnagar and R. Chandra

Table 9 Soil moisture calibration of clay soil

S. No. Size of sample Water content Moisture Sensor value Range value
(g) added content, %
water content
1 500 Nil 0 555 1000–555
2 500 25 ml 5 506 554–506
3 500 50 ml 10 449 505–449
4 500 75 ml 15 436 448–436
5 500 100 ml 20 430 435–429
6 500 125 ml 25 425 428–425
7 500 150 ml 30 420 424–420
8 500 175 ml 35 401 419–401
9 500 200 ml 40 363 400–363
10 500 225 ml 45 360 362–360
11 500 250 ml 50 352 359–352
12 500 275 ml 55 301 351–301
13 500 300 ml 60 285 300–285
14 500 325 ml 65 284 284–284
15 500 350 ml 70 282 283–282
16 500 375 ml 75 282 282–0
17 500 400 ml 80 282 282–0
18 500 425 ml 85 282 282–0
19 500 450 ml 90 282 282–0
20 500 475 ml 95 282 282–0
21 500 500 ml 100 282 282–0

connected with Arduino which will send the data to the smart-phone. In this case,
android-based smart-phone is used. Figure 8 depicts the sketch code.
The experimental setup is shown in Fig. 9.

3.5 Fertilization Recommendation Layer

The next layer is fertilization recommendation layer which is responsible for depict-
ing physical properties of soil and lime requirement to the farmer. This layer is
developed in the form of android-based smart-phone application. The physical prop-
erties of soil sensed by sensors are depicted here using Bluetooth available in the
smart-phone. As discussed in the above section, soil temperature and humidity sen-
sor are pre-calibrated so direct sensed values are used. Soil moisture sensor and pH
IoT-Based Soil Health Monitoring and Recommendation System 13

Table 10 Soil moisture calibration of sand soil

S. No. Size of sample Water content Moisture Sensor value Range value
(g) added content, %
water content
1 500 Nil 0 578 1000–578
2 500 25 ml 5 405 576–405
3 500 50 ml 10 348 404–348
4 500 75 ml 15 343 347–343
5 500 100 ml 20 340 342–340
6 500 125 ml 25 309 339–309
7 500 150 ml 30 301 308–301
8 500 175 ml 35 293 300–293
9 500 200 ml 40 284 292–284
10 500 225 ml 45 275 283–275
11 500 250 ml 50 273 274–273
12 500 275 ml 55 271 272–271
13 500 300 ml 60 270 270–270
14 500 325 ml 65 270 270–270
15 500 350 ml 70 270 270–270
16 500 375 ml 75 269 269–269
17 500 400 ml 80 268 268–268
18 500 425 ml 85 268 268–268
19 500 450 ml 90 261 267–261
20 500 475 ml 95 260 260–260
21 500 500 ml 100 261 260–260

sensor are already calibrated, and these calibrated values are already saved in SQLite
database. Two fertilizers are recommended on the basis of pH, i.e. sulphur and lime-
stone. The standard recommendations for fertilizer use have already been discussed
in Sect. 1 (Introduction). The flow diagram of fertilization recommendation layer is
shown in Fig. 10. The pseudo-code below gives the sulphur and lime requirements
on the basis of pH and land area given by the farmer. The average value is taken for
sulphur and lime requirements as not a specific value is recommended but a range is
recommended.
14 V. Bhatnagar and R. Chandra

Fig. 8 Sketch code of data calibration and acquisition layer

IoT-Based Soil Health Monitoring and Recommendation System 15

Fig. 9 Experimental setup of proposed system with pH sensor, soil moisture sensor, soil temperature
sensor and Bluetooth chip
16 V. Bhatnagar and R. Chandra

Fig. 10 Flow chart of fertilization recommendation system

IoT-Based Soil Health Monitoring and Recommendation System 17

Begin
1. Read type_soil \\ To be Given by Farmer
2. Read land_sizeB \\ To be Given by Farmer in Bigha
3. Read pH_r \\ Raw value received from pH Sensor
4. Set land_sizeA :- land_sizeB * 0.25 \\ Convert Bigha into Acre
5. Set pH :- pH_r \\ Match raw sensor value with Database
6. If (type_soil = “Sand”) then
6.1. If(pH>=4.5 && pH<5.5) then
6.1.1. Lime_ Requirement :- 0.5 * land_sizeA
6.1.2. Else if (pH>=5.5 && pH<=6.5) then
6.1.3. Lime_ Requirement :- 0.6 * land_sizeA
7. Else if (type_soil = “Loam”) then
7.1. If(pH>=4.5 && pH<5.5) then
7.1.1. Lime_ Requirement :- 1.2 * land_sizeA
7.1.2. Else if (pH>=5.5 && pH<=6.5) then
7.1.3. Lime_ Requirement :- 1.7 * land_sizeA
8. Else if (type_soil = “Clay”) then
8.1. If(pH>=4.5 && pH<5.5) then
8.1.1. Lime_ Requirement :- 1.9 * land_sizeA
8.1.2. Else if (pH>=5.5 && pH<=6.5) then
8.1.3. Lime_ Requirement :- 2.3* land_sizeA
End IF
9. IF (type_soil = “Sand” && pH=8.5) then
9.1. Sulphur_Requirement :- 0.8* land_sizeA
9.2. Else IF (type_soil = “Clay” && pH=8.5)
9.3. Sulphur_Requirement :- 1.2* land_sizeA
End IF
10. IF (type_soil = “Sand” && pH=8.0) then
10.1.Sulphur_Requirement :- 0.6* land_sizeA
10.2.Else IF (type_soil = “Clay” && pH=8.0)
10.3.Sulphur_Requirement :- 0.9* land_sizeA
End IF
11. IF (type_soil = “Sand” && pH=7.5) then
11.1.Sulphur_Requirement :- 0.25* land_sizeA
11.2.Else IF (type_soil = “Clay” && pH=7.5)
11.3.Sulphur_Requirement :- 0.45* land_sizeA
End IF
End

4 Results and Discussion

The proposed system is validated with authorized soil laboratory (shown in Appendix
E). New soil samples are taken from the same place and tested again with different
parameters. The values are shown in Table 11 and Fig. 11.
18 V. Bhatnagar and R. Chandra

Table 11 Comparison between calibrated values and laboratory values

S. No. Sample soil Time Sensor value Calibrated value Laboratory value
(%)
1 Sandy soil 9.00 A.M. 341 20 23
2 Sandy soil 1.00 P.M. 291 35 31
3 Sandy soil 6.00 P.M. 390 10 9
4 Loam soil 9.00 A.M. 293 25 19
5 Loam soil 1.00 P.M. 279 30 28
6 Loam soil 6.00 P.M. 458 10 13
7 Clay soil 9.00 A.M. 510 5 7
8 Clay soil 1.00 P.M. 526 5 9
9 Clay soil 6.00 P.M. 546 5 11

Fig. 11 Comparison between calibrated values and laboratory values

Table 12 Comparison of calibrated values and laboratory values of pH sensor

S. No. Soil type Actual pH Sensor value Nearest pH value from
1 Clay 7.9 451 8
2 Loam 8.0 430 8
3 Pure sand 7.8 449 8
IoT-Based Soil Health Monitoring and Recommendation System 19

To validate the calibrated value of soil moisture sensor from laboratory-tested val-
ues, Two-tailed paired t-test is applied [17] between calibrated values and laboratory-
tested values. The t-test can determine whether the means of two sets of data are sig-
nificantly different from each other. The value of T is obtained with the help of MS
Excel, and the value is 0.688, that is less than tabular value of T at eight degrees of
freedom at 5% level of significance 1.860; there is no significant difference between
calibrated value and laboratory value. Therefore, it can be stated that the proposed
model can be used in the field.
The calibration values of pH sensor are also validated with laboratory values. New
soil samples are taken from the same place. The values are shown in Table 12. To
validate the calibrated value of pH sensor from laboratory-tested values, two-tailed
paired t-test is applied between calibrated values and laboratory-tested values. The
value of T is obtained with the help of MS Excel, and the value is 0.225, that is

Fig. 12 Output of proposed system

20 V. Bhatnagar and R. Chandra

less than tabular value of T at eight degrees of freedom at 5% level of significance

1.860; there is no significant difference between calibrated value and laboratory
value. Therefore, it can be stated that the proposed model can be used in the field.
The proposed output is shown Fig. 12.

5 Conclusion and Future Work

Soil is essential for life. It habitat for many insects, act as a filtration system for
surface water and maintenance of atmospheric gases. It is an important component
for agriculture. It is responsible for crop yield. Crops and plants get irrigation and
fertilization through the soil only. There are some physical and chemical proper-
ties of soil that affect the crop yield, such as soil moisture, soil temperature, soil
humidity, electrical conductivity, organic compounds and soil pH. These properties
are required to be monitored on regular basis by the farmers, so in this chapter, IoT
sensor-based system is proposed that monitors the soil temperature, humidity, pH
and moisture and based on pH, lime and sulphur are recommended. The proposed
system is implemented on android smart-phone. Three sensors DHT11, SEN: 0161
and SEN: 0193 for soil humidity and temperature, pH, and moisture, respectively,
are used. pH sensor and soil moisture are calibrated for the soil of Rajasthan of three
categories of loam, sand and clay. The proposed system is tested with authorized lab-
oratories and validated using t-test and found that there is no significance difference
between calibrated values and laboratory-tested values. The proposed system is ben-
eficial for farmers, agriculture scientist, chemists, agriculture professionals and IoT
experts. This work can be extended with more geographical area and incorporation
of other properties of soil, such as electrical conductivity and NPK present in soil.

References

1. Topic: Soil Moisture, available at http://www.agriinfo.in/?page=topic&superid=4&

topicid=274, last visited on 1 Mar 2019
2. Topic: pH Recommendations, available athttp://vric.ucdavis.edu/pdf/Soil/ChangingpHinSoil.
pdf, last visited on 1 Mar 2019
3. Encinas, C., et al.: Design and implementation of a distributed IoT system for the monitoring
of water quality in aquaculture. In: 2017 Wireless Telecommunications Symposium (WTS).
IEEE (2017)
4. Coyle, S., et al.: BIOTEX—Biosensing textiles for personalised healthcare management. IEEE
Trans. Inf. Technol. Biomed. 14(2), 364–370 (2010)
5. Yamaguchi, N., et al.: E-kakashi project, an agri sensor network using ad hoc network tech-
nology. In: 2011 Proceedings of SICE Annual Conference (SICE). IEEE (2011)
6. Lee, M., Hwang, J., Yoe, H.: Agricultural production system based on IoT. In: 2013 IEEE 16th
International Conference on Computational Science and Engineering (CSE). IEEE (2013)
7. Molisch, A.F., et al.: IEEE 802.15. 4a channel model-final report. IEEE P802 15(04), 0662
(2004)
IoT-Based Soil Health Monitoring and Recommendation System 21

8. Badamasi, Y.A.: The working principle of an Arduino. In: 2014 11th International Conference
on Electronics, Computer and Computation (ICECCO). IEEE (2014)
9. Tianlong, N.: Application of single bus sensor DHT11 in temperature humidity measure and
control system. Microcontrollers Embed. Syst. 6, 026 (2010)
10. Topic: PH SEN: 0161, available at https://wiki.dfrobot.com/PH_meter_SKU__SEN0161, last
visited on 10 Mar 2019
11. Hodgkinson, A.: Oxalic Acid in Biology and Medicine. New York, p 416 (1977)
12. Nitta, I., Tomiie, Y., Koo, C.H.: The crystal structure of potassium bicarbonate, KHCO3. Acta
Crystallogr. 5(2), 292–292(1952)
13. Mahzan, N.N., et al.: Design of data logger with multiple SD cards. In: 2013 IEEE Conference
on Clean Energy and Technology (CEAT). IEEE (2013)
14. Topic: PH SEN: 0161, available at https://wiki.dfrobot.com/Capacitive_Soil_Moisture_
Sensor_SKU_SEN0193, last visited on 10 Mar 2019
15. Vance, Eric D., Brookes, Peter C., Jenkinson, David S.: An extraction method for measuring
soil microbial biomass C. Soil Biol. Biochem. 19(6), 703–707 (1987)
16. Pahuja, Ritika, Kumar, Narender: Android mobile phone controlled bluetooth robot using 8051
microcontroller. Int. J. Sci. Eng. Res. 2(7), 14–17 (2014)
17. Ruxton, Graeme D.: The unequal variance t-test is an underused alternative to Student’s t-test
and the Mann-Whitney U test. Behav. Ecol. 17(4), 688–690 (2006)
Implementing IoT and Wireless Sensor
Networks for Precision Agriculture

D. D. Dasig Jr.

Abstract The burgeoning effects of climate change to the agricultural systems have
resulted in reduced agricultural productivity. These extreme weather events, changes,
and increased in temperatures and precipitation patterns can be countervailed by
deploying precision farming technologies in this milieu of a cyber-physical system.
Among the precision farming technologies is the application of Internet of things
(IoT) devices and sensors in farming management by observation, measurement,
response, and monitoring of the food and horticulture crops inter- and intra-field
variability. This chapter presents the rudimentary and recent advances in Agriculture
4.0. This chapter also afforded the deployment of an Internet of things (IoT)-based
precision agriculture technology using wireless sensor networks (WSN) in crop man-
agement. An online crop management system (OCMS) was developed to complement
the IoT platform layer. The application aimed at providing the farm environmental
conditions including the temperature, water level, pH value, and dissolved oxygen
measurements from the paddy field with food crops and horticulture crops being
grown. Further studies could be done by analyzing the inter- and intra-field variabil-
ities such as soil herbicide weed control, side dressing, late blight control, and haulm
killing.

Keywords Agricultural system · Internet of things · Precision farming · Tomato

crop management · Wireless sensor network

1 Overview

Fundamental to creating a functional economy and sound societies is the agricultural

and food production, hence, sustainable agriculture. Today, the Internet of things
(IoT) progresses toward becoming limestone on the planet, where every object
or everything is interconnected and interacting with a specific task or functions.

D. D. Dasig Jr. (B)

Graduate Studies, College of Science and Computer Studies, De La Salle
University—Dasmarinas, DBB-B, 4115 West Ave, Dasmariñas, Cavite, Philippines
e-mail: [email protected]

© Springer Nature Singapore Pte Ltd. 2020 23

P. K. Pattnaik et al. (eds.), Internet of Things and Analytics
for Agriculture, Volume 2, Studies in Big Data 67,
https://doi.org/10.1007/978-981-15-0663-5_2
24 D. D. Dasig Jr.

In this chapter, the issues and advances in the agricultural sector are discussed. The
effects of Industry 4.0 on ramifications and industrial modifications in the industrial
and agricultural production systems have open new paradigm and opportunities for
the researchers and scientists to conduct empirical investigations and technological
innovations to suppress and address the increasing world population to feed, and the
changing demands of the society.
There exists a growing demand for ubiquitous connectivity as the world forward
with the Fourth Industrial Revolution (4IR). The 4IR has changed the many faces
of society, hence, Agriculture 4.0 era. The recent advances in the primary sector of
the economy such as in agriculture strike the notion of more productive agriculture,
modern, and innovative.
These characteristics associated with Agriculture 4.0 have coined smart agricul-
ture, diversified farming, digital farm, disruptive models in agriculture, and many
others. Agriculture 4.0 will have a pivotal role in agribusiness and agriculture eco-
nomics. These agricultural technologies will help improve food production, reduced
costs, and promote process efficiency.

2 Introduction

The primary sector of the economy including the agriculture sector plays a vital role
in the national and economic development of a country [1]. The activities being asso-
ciated to the primary sector of the economy includes the extractions and harvesting
of basic goods and raw materials from the earth such as agriculture and forestry, fish-
ing, quarrying, mining, gathering, and hunting [2]. Agriculture is the science with
regard to cultivating or farming, including cultivation of the soil for the growing and
management of crops and the rearing of animals to produce and provide food, fleece,
and other variety of agricultural produce. Agricultural development and an increase
in agricultural production are crucial to broader national and economic development
[3].
Agricultural crops production is primarily important for human use and domestic
consumption. Crop production is highly dependent and affected by the type of soil,
pests, irrigation, weeds, bacterial endophytes, residues, and yields. In recent years,
agricultural crop production has drastically reduced due to short episodes of high
temperatures during the reproductive period [4]. Also, a decreased in production
was attributed to the extent of application of fertilizers since and intensive high-yield
agriculture is heavily dependent on fertilizers [5].
In some studies, it was posited the water salinity in an irrigated agriculture, and
the problem of solidity [6] and the endophytic bacteria latently residing in the plant
tissue have impacted crops production [7], although agriculturists recommended the
use of cover crops to improve the soil fertility and crop performance [8]. The need to
remodel crop management should be done to improve the growth and development
of the crop, the expected yields through modifications in the management practices
Implementing IoT and Wireless Sensor Networks for … 25

of irrigation, biological interactions, countervail diseases, and weeds and pests using
environment-friendly solutions.
Most studies have concurred deploying well-advanced technologies for future
development in the world food, nutrition, and agriculture [9] to sustain the supply
and demand. In most agricultural countries, lower agricultural productivity is asso-
ciated with traditional farming methods and climate [10]. The burgeoning and long-
term effects of climate change to the agricultural systems have resulted in reduced
to adverse agricultural productivity, a problem burnt by agricultural-dependent com-
munities and countries [11]. These extreme weather events, changes, and increased
in temperatures and precipitation patterns can be countervailed by utilizing quanti-
tative tools and techniques [12], climate model forecasting, and deploying precision
farming technologies in this milieu of a cyber-physical system.
Among the precision farming technologies is the application of Internet of things
(IoT) devices and sensors in farming management by observation, measurement,
response and monitoring of the crops inter- and intra-field variability. This study
presents the deployment of wireless sensor networks (WSN) in crop management to
observe, measure, and monitor the farm temperature, humidity, and pressure using
P89V51RD2 microcontroller, ZigBee module XBee module IEEE 802.15.4 standard,
and RF transceiver.

3 Agroecology and Sustainable Agriculture

Fundamental to creating a functional economy and sound societies is the agricul-

tural and food production, hence, sustainable agriculture. Sustainable agriculture is
a prominent and a very vague discourse [13] defined as cultivating or the conduct of
farming in a modest and environment-friendly approach to meet the present needs
of society without compromising the needs of tomorrow’s generation, given their
comprehension of the relationships and interactions of the biological system.
Today, the agricultural sector should transform from a traditional farming system
to e-agroecology to achieve sustainable agriculture. This e-agroecology integrates
information and communications technology (ICT) to the ecological processes in
agricultural production systems. The e-agroecology recognizes the importance of
ICT in agriculture. Agroecology is the investigation of natural processes related to
farming techniques as the ecology of food systems [14]. This farming approach recog-
nizes environmental, economic, and social dimensions centered on food production
which allows the future generations to be able to feed themselves [15, 16].
According to the study which utilized participatory learning, aside from having
well-advanced and top of the line technologies in the agricultural production sys-
tems, there is a need to have experts in the agricultural fields who are passionate
to share their knowledge, expertise, and empower communities, and agricultural
stakeholders [17]. Participatory learning in agriculture developed the community
of farmers and agriculturists equipped with ecological processes and technological
skills to manage engineering and ICT-based solutions toward sustainable agriculture.
26 D. D. Dasig Jr.

Agriculturists, scientists, and engineers should work together to improve the design,
development, production strategies, and better decision-making.

3.1 Challenges and Developments in Sustainable Agriculture

Fundamental to global demands, agricultural work has always had been considered
the most prevalent employment opportunities [18]. Agriculture, however, threatens
the water supply since it is the major contributors to wastewater. Wastewater from
agriculture is a global issue which has directly and indirectly adversely impacted the
ecosystem and biological diversity. According to the United Nations Environment,
water is principal and fundamental to life on our planet, yet this valuable asset has
progressively been sought after and under risk [19] and water shortages in many
regions [20]. Water covers multiple thirds of the planet, and our agricultural sector
has always had produced wastewater including those chemicals from industrialized
agricultural spaces which flows directly to waterways, lakes, wetlands, and springs.
Another persistent issue in agriculture is agricultural biotechnology. Although
its well-advanced and scientifically investigated application has presented several
promising benefits, it also brought a new dimension of challenges to agricultural
sector [21]. Land clearing, land conversions, and industrial agricultural production
have been seen as a factor to radiative forcing of climate because of secretions of
greenhouse gasses, and ozone on crops and agro-ecosystems [22, 23].

3.1.1 Industrial Agriculture

Industrial agriculture is now the prevailing food production system; a large-scale

production system heavily utilizes and promotes substantial utilization of compound
composts, chemical fertilizers, and pesticides. Industrial agriculture posed soil deple-
tion a made it vulnerable to soil erosion. This kind of farming technique develops
“dead zones” or the oxygen-deprived spaces.
Industrial agriculture posed health risks as a source of air contamination, one of
the reasons for agricultural production laborers respiratory and cardiovascular con-
ditions. Although industrialized farming has been effective in creating huge amounts
of sustenance, the eventual fate of food production is in peril because of loss of agri-
cultural lands and reduced variations in the production of crops and livestock even
more.

3.1.2 Conservation Agriculture

Today, several agricultural production systems have been integrated into the frame-
work of large-scale food production including conservation agriculture. In this con-
text, it is characterized as a sustainable agriculture production framework including
Implementing IoT and Wireless Sensor Networks for … 27

land cultivating practices with an embedded adaptation technique based on the yields
and Ambiental and environmental conditions of the region or locality. CA is utilized
as a soil management practice to shield the soil from disintegration, degradation, and
erosion [24], as well as to improve its quality.
CA advocated the minimal soil disturbance, permanent soil cover combined with
rotations as a sustainable agricultural production system [25]. CA will help the agri-
culture sector to efficiently and effectively provide the need of the human population
while making the agricultural land productive [26] to help economy advance toward
agricultural sustainability and biodiversity conservation [27, 28].

3.1.3 Climate-Smart Agriculture

Another transformative approach in the agricultural production system is sustain-

able intensification and climate-smart agriculture (CSA). Climate-smart agricul-
ture (CSA) is a methodology that aides’ activities expected to change and recalibrate
agricultural systems to successfully bolster improvement and guarantee sustenance
security in a changing climate. CSA approach is characterized as one of the rural prac-
tices that economically viable and helped increment the profitability in agriculture
with space for flexibility while diminishing ozone-depleting substance emissions.
Climate-smart agriculture can be characterized as a transformative approach
toward agricultural efficiency and livelihoods by transforming and reorienting agri-
cultural systems [29]. Climate-smart agriculture (CSA) compliments sustainable
intensification which essentially important in adapting to climate change [30],
increased in food security and adaptive climate mitigation capacity in a sustain-
able way [31, 32] while providing the increasing demands of food, fuel, and fiber
[33].

3.1.4 The e-Agroecology

Due to the increasing threat to the natural resources brought by modern agriculture,
there have been approaches and diversified farming methods applied by the rural
social movements to attain sustainable agriculture [34, 35]. One of the frameworks
used to address the issues and challenges in agricultural production is the agroecol-
ogy. Primarily, agroecology was used in crop production [15, 36]. Agroecological
approach dealt with farming systems to preserve and feed future generations consid-
ering the economic, social, and environmental dimensions [15, 37].
In this study, the e-agroecology was proposed to integrate technological ramifi-
cations into the agricultural production systems centered on food production consid-
ering the environmental, social, economic, and technological dimensions (ESET) as
illustrated in Fig. 1.
This farming approach will utilize the cyber-physical age instrumentation and
communications techniques including the Internet of things (IoT) devices and wire-
less sensor networks.
28 D. D. Dasig Jr.

Fig. 1 e-Agroecology
(ESET) model

3.1.5 Precision Agriculture

The industrial ramifications of the Fourth Industrial Revolution (4IR) have reshaped
the society with the technological advances, disruptive technologies, and emerging
business models. It has disrupted the conventional industries with new technical
models and technologies. In the agricultural sector, the use of drones, the Internet
of things (IoT), and blockchain have been notably considered as the agritech sector
welcomes Agriculture 4.0. The use of these new technologies to farming management
is referred to as precision agriculture. Aside from the deployment of technological
innovations in the farming and crop management, precision agriculture is seen as a
technified model to farming management to increase crops productions and making
sure of environmental, social, and economic sustainability.
The agricultural sector has been primed to develop and produce food to feed the
world. Over the years, there have been innovations in the agritech using emerging
technologies including aeroponics and hydroponics. These methods utilized engi-
neering and information technology the same as precision agriculture. Precision
agriculture is a farm or crop management approach utilizing information technology
and engineering to ensure the optimal health of the soil and increase productivity.
With the use of information technology services, the farmers could be able to conduct
the processes of observing and measuring as well as able to immediately respond
to the inter- and intra-field crops variability. Precision agriculture has been seen to
boost the agricultural sector as an effective method in farm and crop management.
Notably, in the mid of 1980s, precision agriculture was initiated to blend with
agriculturists and farmers way of doing the farming processes with the available and
emerging technologies [38]. Precision agriculture is an inventive approach, coordi-
nated and universally standardized method intends to further the effectiveness of
technologies and soil use and to decrease the vulnerability of decisions required
to control minor departure from farms [33]. In deploying technology for precision
farming, it would involve a well-designed technology utilizing sensors, informa-
tion systems, improved apparatus, and information management system [39, 40].
These spatially planned and deployed sensing devices will be the gathering tools for
sensing the environmental and Ambiental conditions of the farms, hence improving
the farmer management system and production, as well as ecological quality [41].
Precision agriculture also advocated farm-site management from a specific
approach to large-scale site management to improve soil quality, pesticides distribu-
tion to support farmers to increase the profitability of yield creation, improve crops
Implementing IoT and Wireless Sensor Networks for … 29

and farm produce quality, and protect the environment [42]. Other precision agri-
culture technologies used optical lists from hyperspectral remote detecting in the
evaluation of vegetation biophysical factors both in forestry and agriculture manage-
ment [43]. Remote sensing techniques and detection applications in precision agri-
culture began with sensors for soil natural issue and have rapidly diversified toward
satellite, flying, and handheld or tractor-mounted sensors [44]. Spatial estimations
of yield utilizing mechanical advances like in a hurry yield monitoring systems [45],
while the use of the airborne remote-detecting system has been recognized worldwide
as a promising strategy for identifying and mapping weeds in yields [46].

4 Agriculture 4.0

With the onslaught of the Fourth Industrial Revolution (4IR), there have been ram-
ifications and disruption to the sectors of the economy. The 4IR can be character-
ized as the era of cyber-physical systems where automation and advancements to
industries were shaped and heavily influenced by information technology. With the
automation and machine-driven production and technology-driven processes, there
have been an increased in technology proficiency requirements, higher demand of
engineers and information technology professionals, and enabled service-centric and
customer-centric organizations.
The 4IR has brought emerging and innovations to current technologies includ-
ing Internet of things, artificial intelligence and machine learning, robotics and
automation, virtual, and commercial reality. These modifications and ramifications
are attributed to advancements in the process and technology in the production to
better the lives of humanity. The Fourth Industrial Revolution or commonly known
as Industry 4.0 has changed the society with technological concepts and solutions,
top of the line technologies to answer the recurring problems [47, 48]. It was opined
to realize mesh economy of scale and the economy of scope.
With the Industry 4.0, the agricultural sector has been seen to be innovative to be
able to adapt to the changing demands of society today with the concept of Agriculture
4.0. Agriculture 4.0 has invigorated agricultural systems to address challenges and
issues in soil management, fertilizer, pests, crop management, yields, plant stresses,
etc. [48, 49]. The synergies with combination and cooperative energies of the most
recent ICT advancements such as embedded system, real-time system, telecommu-
nication, web automation, mobile communications and ICT infrastructure, remote
sensing, and wireless communications have pinned into the agricultural sector geared
toward smart agriculture and precision agriculture [49–51].
Agriculture 4.0 was coined as the function and work of agricultural engineering
in the design, development, and deployment of systems including drones, Internet
of things (IoT), smart and vertical farming, artificial intelligence, robotics, and solar
energy [52] to advocate low-cost farming and increasing yields and productivity [53].
In the study for agro-technology, it was emphasized that Agriculture 4.0 utilizes a
30 D. D. Dasig Jr.

decision support system to address climate and weather forecasting, estimation of

complex, and ill effects of climate to crops and soil [54].
The Agriculture 4.0 has opened opportunities toward the application of drones in
farm and crop management, using TCP/IP protocols, image processing for plant dis-
ease identification and management, machine learning and classification techniques,
and a complete data-driven agriculture value chain. The uses of the Internet of things
and machine learning have produced an optimal management approach in agricul-
tural production system [50]. Researchers concur that Industry 4.0 has boosted the
economies and industrial upheaval. These advances in the distributed computing
were pivotal to foster the design, development, and deployment of smart agricul-
tural systems, precision farming [55]. Industry 4.0 has immensely revolutionized the
agribusiness models and transformed the agricultural sector [55, 56].
In the context of Agriculture 4.0, the agri-food value chain considers and intercon-
nected stakeholders [57] who advocate an enhanced productivity, and with the support
of policy-makers in the technology revolution [58], this era will achieve higher pro-
ductivity in agriculture, better marketing strategies using technology, reduced food
wastes, and greater eco-efficiency [57–59].

5 Internet of Things (IoT)

The Internet of things (IoT) is one of the ramifications in this Industry 4.0 era. IoT is
a system of interrelated and connected computing devices, a cyber-physical system,
and digital machines with unique attributes and identifiers with the ability to transfer
data over a network. Internet of things is a system which combines the Internet
and the emerging and future technologies [60]. Because of the Internet of things
(IoT), the ability to gather, transfer, and process data without human intervention
has been made possible; hence, human–human and human–computer interactions
(HCI) have been eliminated from the architecture [60, 61]. IoT is characterized as
a global Internet-based architecture which utilizes IoT devices, IoT gateway, and
IoT platform layers [61, 62] with the networked interconnection of communicating
things equipped with ubiquitous intelligence [62, 63].
The Internet of things (IoT) as a technological paradigm is also considered as
the Industrial Internet because of its applications to the industries and realm of the
physical world [64] to provide secured and powerful industrial systems [62, 64,
65]. Internet of things (IoT) also known as the Internet of everything is a system
that utilizes spatially planned and distributed remote sensing devices and actuators,
wireless, and mobile devices with an embedded system to seamlessly collect and
transfer data over an open access system in the IoT environment [66–68]. With the
technological revolution of 4IR, Internet of things has been seen to provide numerous
opportunities to industrial and agricultural sectors given the support of academic and
technical communities toward sophisticated and a more transparent and secured
ubiquitous world [69].
Implementing IoT and Wireless Sensor Networks for … 31

Fig. 2 IoT architecture

The Internet of things as an Internet-based technical architecture utilizes the IoT

architecture as illustrated in Fig. 2. The IoT architecture affords the three layers
including the IoT device layer, IoT gateway layer, and the IoT platform layer. Each
layer of this architecture is characterized according to their functions to collect and
transfer data from the source to the physical world of the IoT solution.
The Internet of the future or the Internet of things (IoT) has been seen to have
grown rapidly in recent years. It has become the domain of discourse and research
focus to many which embed IoT devices such as RFID, the sensors and actuators
into the design and development of an IoT architecture IoT solution [64]. The expe-
ditious developments of the Internet of things and the cyber-physical systems have
even increased the demands of the IoT and smart devices [65] as these devices are
main system components to serve as the bridge between the physical and virtual
environments [70].
Looking at the IoT architecture’s IoT device layer, the IoT devices are spatially
distributed purposely embedded in the design for sensing and actuating the physical
environment such as Ambiental conditions [71] and collecting information [72] from
the locale for transfer to the IoT gateway layer using a protocol [65, 70]. These IoT
devices are also connected to multiple devices to connect to the Internet to facilitate
communication to the cloud server and utilize the cloud computing services in the IoT
platform layer [73]. These IoT devices are geographically deployed in heterogeneous
platforms, topographically dispersed over heterogeneous stages, traversing various
administration and functional areas [71].
In the Internet of things (IoT), the IoT gateway layer is considered as the software
program or a physical device that bridges the connection between the IoT device
layer and IoT platform layer. This layer is responsible for connecting the sensing and
actuating devices and the heterogeneous platforms or applications in the IoT platform
layer. It is exemplified by providing additional and improved data security for the
intelligent system or network as well as the data that it transports [74]. Primarily,
IoT gateway is the enabler of local communication or a short-range communication
between the sensors, actuators, and other IoT devices commonly applied in industrial
and production systems, public security system, smart communities and buildings,
logistics [74], and even in agriculture.
In a design of an IoT gateway-centric architecture, the IoT architecture was used to
develop novel M2M services by providing real-time interactions between the smart
things and the mobile clients [75]. Another implementation of smart IoT gateway has
emphasized the applicability of join point of a heterogeneous network [76], Zigbee
32 D. D. Dasig Jr.

Adhoc network, cable network, and wireless LAN to countervail design issues and
demand for ubiquitous connectivity [77].
The third layer of the IoT architecture is the IoT platform layer. It is a multi-layer
technology or commonly referred to as middleware which facilitates data commu-
nication, flow, device management, and application support and management. IoT
platform is the enabler for provisioning of the IoT solution services, management of
IoT services, and the automation of connected things and devices. The Internet of
things universe has increased the demands of IoT platforms to reduce IoT solutions
development cost, risks, and scope. These commercially available and open sources
IoT platforms have even help to foster IoT solutions development’s time-to-market.
These IoT platforms maybe an end-to-end solution, connectivity, cloud platform, or
data platform.
These IoT platforms play a pivotal role in the IoT solution development for indus-
trial applications since it enables value for a sector or business and its customer. In
the context of the agricultural sector, IoT platforms help developers and farmers to
select appropriate platform and services which heavily rely on the type of attributes
and defined system functions. IoT platform allows solutions users to be able to collect
the needed data, store and manage the same data, and process into actionable insights
using analytical models and visualization techniques for a data-driven platform [78].
In the study enabling IoT ecosystem in the deployment of the Internet of the future
applications, and ramifications, the researchers explored the functional design of the
system using a platform-specific API and IoT solutions information models [79].
The proliferations of ICT and engineering solutions have our cities transformed
into smart cities and smart communities utilizing IoT hubs [80]. These IoT hubs sub-
scribed to the cloud-based platform or cloud computing services with the advent of
Platform-as-a-Service (PaaS) architecture [78, 80] to support IoT hubs for the smart
cities and smart communities’ services. Among other revolutionary IoT platforms
to manage IoT devices and applications in Industry 4.0 is the blockchain technol-
ogy’s bitcoin deployed to cashless communities, financial, and banking sector [81].
In digital agriculture, a “Phenonet” was proposed by enhancing an IoT platform
semantically enhanced IoT platform to address issues of interoperability, IoT silos,
among others [81, 82].

6 The Wireless Sensor Network

In the wireless sensor network, the main building block is the sensor node. A sensor
node or a “mote” is a node in a sensor network. A sensor node will observe the
climatic and environmental conditions, process, and gather the sensory information,
and it communicates with other nodes in the network of sensors. A sensor node in
a sensor network consists of the sensor unit, processing, communication, and power
units. A wireless sensor network is a group of geographically deployed and dispersed
sensors for gathering, monitoring, and recording the environmental and Ambiental
conditions [83], then organized these collected data at a central location.
Implementing IoT and Wireless Sensor Networks for … 33

A wireless sensor network has immensely reinforced numerous industrial and

application-specific use since it has been re-engineered into its smaller, cheaper,
and intelligent features [83] as an integrated micro-devices and micro-technology,
digital electronics, wireless, and telecommunication devices [83, 84]. This compact
design and feature-rich embedded in the sensor and actuators have been given impor-
tance in the areas of research, embedded system, real-time system, measurements,
and network implementations [85]. The sensor nodes’ sensing capabilities, power
management, routing, and transfer protocols are specifically designed based on the
purpose of the application [85, 86].
Technical communities and researchers have deployed wireless sensor networks
using different design, protocols, methods, algorithm, estimations, and enhancements
to align with specific system characteristics. A T-MAC WSN was developed to deploy
a network with low message rate and lower sensitivity to latency [87]. Another
study advocated Energy-LEACH protocol to improve the choice method of cluster
head [88]. Another study on estimation of convex position was proposed to estimate
the sensor network unknown node position [89]. A medium access control (MAC)
protocol was proposed for a WSN that allows the network of devices to collaborate for
common applications [90]. A low-power transceiver was effectively used to advance
an energy-efficient WSN using hierarchical cluster algorithm [91].
Another WSN deployed to a vineyard setting utilized a 65-node multi-hop net-
work to identify ways and identify the farm conditions [92]. Another system deployed
to monitor different parameters of distributed crop aimed at improving the agricul-
tural production by utilizing a multiplicity of emerging and innovative technologies
[93]. Another innovation in WSN was a pervasive wireless sensor network consists
of connected cheap and simple devices capable of self-configuration network to
transform agriculture utilizing precision technologies for agriculture [94]. A Zigbee-
based technology was used in precision agriculture with a sensor network based on
the IEEE 802.15.4 standard aimed at collecting the data of climatologically and other
environmental properties [15, 94].

7 The System

This study presents an IoT-based precision agriculture technology using wireless

sensor networks (WSN). The technology aimed at collecting, processing the Ambi-
ental and environmental condition on the farm and provides notifications and rec-
ommended courses of actions when needed.
The technology removes the human-to-human and human-to-computer interac-
tion utilizing the interconnected sensor nodes deployed spatially over the distributed
mixed crop farm with food and horticulture crops (rice, corn, fruits, and vegetables).
The next sections discuss the design and detail of the system.
34 D. D. Dasig Jr.

7.1 IoT-Architectured Precision Agriculture Technology

In this study, the IoT device layer includes the sensors and IoT gateway layer, and
the IoT platform layer. The architecture of the developed technology is illustrated
in Fig. 3. All system components are connected to a P89V51RD2 microcontroller
including the sensors (for four sensor nodes), Zigbee module, and RF transceiver.
The IoT architecture was employed in the design and development of the technology
with the IoT device layer, IoT gateway layer, and IoT platform layer. In the IoT
device layer, the sensor node is composed of temperature sensor, pH sensor, dissolved
oxygen sensor, and the water level sensor. The sensor node will observe the climatic
and environmental conditions in the farm where food and horticulture crops are
grown.
The sensing devices will process the gathered sensory information, and it will be
communicated to the other nodes or the platform layer using the TCP/IP and Zigbee
protocol. These sensors were deployed in the paddy field or an arable land primarily
used to grow aquatic and semi-aquatic rice. A wireless sensor network and Zigbee,
and TCP/IP protocol were used in the IoT gateway layer.
The cloud computing services were used with an online crop management sys-
tem—a web and mobile compatible application. The data observed and measured is
collected and displayed in a dashboard which signals the system to respond based
on the needed parameterized measurements.

Fig. 3 Block diagram of precision agriculture technology

Implementing IoT and Wireless Sensor Networks for … 35

7.2 The Precision Agriculture Wireless Sensor Network

In this design, the sensor unit is composed of temperature, pH, dissolved oxygen, and
the water level sensors. In measuring the water temperature, a thermometer sensing
technology was used, and a glass electrode method was utilized for pH value. On
the other hand, a membrane electrode technique was utilized for a dissolved oxygen
sensor and a localized sensor designed out of an operational amplifier as a functional
water level sensor.
In this system, there were four (4) sensor nodes spatially deployed in the farm of
food and horticulture crops as illustrated in Fig. 4. Each sensor unit has a temperature
sensor, pH sensor, dissolved oxygen sensor, and water level sensors. These sensors
will also gather sensory information from the paddy field to measure and monitor the
value of the parameter specified for each sensor. In measuring the water temperature
in the field of aquatic and semi-aquatic rice, a thermometer sensing technique is used.
For the pH value of water in the field an electrode, a method was used to identify

Fig. 4 Wireless sensor network (WSN)

36 D. D. Dasig Jr.

and measure the current pH level of the water or acidity. This is to make sure that the
plants or crops will have optimal and healthy conditions, although some of the crops
thrive to survive even the pH value is below the range, 7; however, these plants and
crops are heavily affected.
For aquatic and semi-aquatic crops, the dissolved oxygen in their habitat is primed
for their survival. The membrane electrode technique was utilized for dissolved oxy-
gen sensing and measurement. For the water level sensor, there has been a localized
sensor designed out of an operational amplifier as a functional water level sensor.

7.3 Online Crop Management System

In this study, an online crop management system (OCMS) was developed as part
of the IoT platform layer. The system aimed at providing the farm environmental
conditions including the temperature, water level, pH value, and dissolved oxygen
measurements from the arable farm with food crops and horticulture crops being
grown. The OCMS will help the farmers to the optimized farm operation and man-
agement, data management, file sharing, analyzing the farm Ambiental data, and for
analytical reporting management.
The online crop management system (OCMS) graphical user interface as illus-
trated in Fig. 5 provides descriptive visualization in terms of temperature in (°C), pH
or water acidity in moles per liter, dissolved oxygen in milligrams per liter (mg/L),
and water level measurements (ft).

Fig. 5 Graphical user interface of online crop management system

Implementing IoT and Wireless Sensor Networks for … 37

8 Results and Discussion

The succeeding subsections will discuss the data gathered from the 6-day sensory
observations during the field tests conducted in the farm with food crops and horti-
culture crops. The gathered data includes temperature in (°C), pH or water acidity
in moles per liter, dissolved oxygen in milligrams per liter (mg/L), and water level
measurements (ft).

8.1 Farm/Water Temperature

As illustrated in Table 1, the average water temperature in the farm was within the
normal temperature of 25 °C except during the fifth day of testing. It also shows that
field or the water temperature is higher than the air temperature. In farm management,
when the water temperature is above 25 °C, there is a need for moisture to about five
mm per day.
It was warm and stems elongation growth stage during the testing days; hence,
each of the food and horticulture crops will utilize water as well in the paddy field.
In this soil system of the paddy fields where the food crops and horticulture crops
are grown, the soil temperature change also happens due to the competition in the
ecosystem on water consumption [95].

8.2 pH Level

The pH level is the negative logarithm of hydrogen ions concentration in water.

The pH is a scale being used to specify if the solution or water is basic or acidic.
Accordingly, the pH in water is heavily influenced by the concentration of carbon
dioxide. The farmers sometimes utilized dolomitic limestone and phytoremediation
so that the pH level can be adjusted.

Table 1 Average temperature

Test day Temperature (°C)
Soil/water Air
1 24 22
2 24 23
3 24 22
4 25 23
5 26 24
6 25 23
°C Celsius
38 D. D. Dasig Jr.

Table 2 Average pH level

Test day pH level (moles/L)
a.m. p.m.
1 6.5 6.5
2 6.4 6.4
3 6.5 6.4
4 6.3 6.3
5 6.4 6.3
6 6.4 6.4
a.m. 12:00 a.m.–11:59 a.m. (PHT)
p.m. 12:00 p.m.–11:59 a.m. (PHT)

The plot where testing has been conducted was a paddy field with a fertile riverine
alluvial soil, and however, due to irrigation, drainage and daily data of precipitation,
its desirability to ideally grow food crops such as rice and other horticulture crops
are paralyzed due to acidity [96]. Based on the testing, there have been no recorded
entries for the mean of pH level. This implies that the pH level of soil/water is within
the desirable soil pH range between 5.5 and 6.5 as illustrated in Table 2.

8.3 Dissolved Oxygen

In the wet rice farming and food crop production such as rice, water quality is also
an important indicator. The water flows to the plot during rain or irrigation schedule.
These farming activities are important to make sure that the semi-aquatic organisms
can respire to survive. Oxygen dissolves in the surface water due to the process of
aerating action of winds.
Dissolved oxygen (DO) is the amount of gaseous oxygen (O2 ) that has been
dissolved water expressed in milligrams per liter (mg/L). Food crop such as rice
can grow in bunded fields or paddy ones. Both fields need continuous flooding or
irrigation to make sure that there is sufficient water to avoid supersaturation that can
cause gas bubble disease to aquatic and semi-aquatic organisms.
The very high concentration of DO may pose lethal effects to the aquatic and
semi-aquatic organisms and might deliberately destroy the soil. Table 3 illustrates
that there was a variation; however, these values are within the range of desirable
values for growing food crops. It also shows that during the testing, the paddy field
was conducive for growing wet rice as a food crop and other horticulture crops.
Implementing IoT and Wireless Sensor Networks for … 39

Table 3 Average dissolved

Test day Dissolved oxygen (mg/L)
oxygen
a.m. p.m.
1 5.0 5.1
2 5.2 5.1
3 5.2 5.1
4 5.3 5.4
5 5.3 5.3
6 5.4 5.5
a.m. 12:00 a.m.–11:59 a.m. (PHT)
p.m. 12:00 p.m.–11:59 a.m. (PHT)

8.4 Water Level

In growing food crops, a recommended continuous flooding or scheduled irrigation

should be best provided after transplanting to provide the best growth and con-
ducive environment. A desirable level of 3 cm water level is being maintained and
is increased according to environmental conditions, or an increase in plant height.
Water level and continuous flooding are also important to make sure weeds and other
pests will not flourish in the paddy field. Depending on the type of rice being grown,
the water level should also be monitored to make sure that the system could support
the rice growth stages of vegetative, reproductive, and ripening.
As illustrated in Table 4, during the tests, the water level was sent to the analytical
reporting system by the sensors were higher during the first day. These water levels
from two areas/plots were vouched by the farming activities since the first day of
the testing was conducted a few hours after the scheduled weekly irrigation. During
irrigation, the water level is maintained at 5 cm from the ground, and during drainage,
3 cm from the ground is maintained.

Table 4 Average water level

Test day Plot 1 Plot 2
Water level (cm) Water level (cm)
a.m. p.m. a.m. p.m.
1 4.4 4.3 4.4 4.4
2 4.0 3.9 4.3 4.2
3 3.7 3.5 4.2 4.1
4 3.5 3.5 4.0 3.8
5 3.5 3.4 3.7 3.6
6 3.4 3.4 3.4 3.4
40 D. D. Dasig Jr.

9 Conclusion and Recommendation

This chapter aimed at providing the rudimentary and ramifications of Industry 4.0 in
the agricultural sector. It presented the issues and challenges and recent advancements
in the agriculture during the Agriculture 4.0 era. This chapter also presented the
design, development, and deployment of an application of Internet of things (IoT)-
based precision agriculture technology using wireless sensor networks (WSN) in
food and horticulture crop management by collecting and processing the Ambiental
and environmental condition on the farm. The same application provides notifications
and recommended courses of actions when needed.
In this chapter, it was elaborated in the design and a development phase that
the IoT architecture was utilized by the author which includes the IoT device layer
including the sensors and IoT gateway layer, and the IoT platform layer. In this
precision agriculture technology, there were four (4) sensor nodes spatially deployed
in the farm of food and horticulture crops. Each sensor unit has a temperature sensor,
pH sensor, dissolved oxygen sensor, and water level sensors. These sensors will also
gather sensory information from the paddy field to measure and monitor the value
of the parameter specified for each sensor.
Also, an online crop management system (OCMS) was developed as part of the
IoT platform layer. The application aimed at providing the farm environmental con-
ditions including the temperature, water level, pH value, and dissolved oxygen mea-
surements from the arable farm with food crops and horticulture crops being grown.
The OCMS will help the farmers to the optimized farm operation and management,
data management, file sharing, analyzing the farm Ambiental data, and for analytical
reporting management on temperature in (°C), pH or water acidity in moles per liter,
dissolved oxygen in milligrams per liter (mg/L), and water level measurements (cm).
Further studies could be done by analyzing the inter- and intra-field variabilities
such as soil herbicide weed control, side dressing, late blight control, and haulm
killing.

References

1. Matsuyama, K.: Agricultural productivity, comparative advantage, and economic growth. J.

Econ. Theor. 58(2), 317–334 (1992)
2. Kenessey, Z.: The primary, secondary, tertiary and quaternary sectors of the economy. Rev.
Income Wealth 33(4), 359–385 (1987)
3. Alexandratos, N.: World food and agriculture: outlook for the medium and longer term. Proc.
Natl. Acad. Sci. 96(11), 5908–5914 (1999)
4. Teixeira, E.I., Fischer, G., Van Velthuizen, H., Walter, C., Ewert, F.: Global hot-spots of heat
stress on agricultural crops due to climate change. Agric. For. Meteorol. 170, 206–215 (2013)
5. Matson, P.A., Parton, W.J., Power, A.G., Swift, M.J.: Agricultural intensification and ecosystem
properties. Science 277(5325), 504–509 (1997)
6. Rhoades, J.D., Loveday, J.: Salinity in irrigated agriculture. Agronomy 30, 1089–1142 (1990)
7. Hallmann, J., Quadt-Hallmann, A., Mahaffee, W.F., Kloepper, J.W.: Bacterial endophytes in
agricultural crops. Can. J. Microbiol. 43(10), 895–914 (1997)
Implementing IoT and Wireless Sensor Networks for … 41

8. Teasdale, J.R.: Contribution of cover crops to weed management in sustainable agricultural

systems. J. Prod. Agric. 9(4), 475–479 (1996)
9. Bruinsma, J.: World agriculture: towards 2015/2030: an FAO study. Routledge (2017)
10. Adams, R.M., Hurd, B.H., Lenhart, S., Leary, N.: Effects of global climate change on agricul-
ture: an interpretative review. Climate Res. 11(1), 19–30 (1998)
11. Rosenzweig, C., Parry, M.L.: Potential impact of climate change on world food supply. Nature
367(6459), 133 (1994)
12. Adams, R.M., Rosenzweig, C., Peart, R.M., Ritchie, J.T., McCarl, B.A., Glyer, J.D., Allen Jr.,
L.H.: Global climate change and US agriculture. Nature 345(6272), 219 (1990)
13. Velten, S., Leventon, J., Jager, N., Newig, J.: What is sustainable agriculture? A systematic
review. Sustainability 7(6), 7833–7865 (2015)
14. Francis, C., Lieblein, G., Gliessman, S., Breland, T.A., Creamer, N., Harwood, R., Wiedenhoeft,
M.: Agroecology: the ecology of food systems. J. Sustain. Agric. 22(3), 99–118 (2003)
15. Wezel, A., Bellon, S., Doré, T., Francis, C., Vallod, D., David, C.: Agroecology as a science,
a movement and a practice. A review. Agron. Sustain. Dev. 29(4), 503–515 (2009)
16. De Schutter, O. Agroecology and the Right to Food. Report presented at the 16th session of
the United Nations Human Rights Council [A/HRC/16/49], vol. 8 (2011)
17. Pretty, J.N.: Participatory learning for sustainable agriculture. World Dev. 23(8), 1247–1263
(1995)
18. Frank, A.L., McKnight, R., Kirkhorn, S.R., Gunderson, P.: Issues of agricultural safety and
health. Annu. Rev. Public Health 25, 225–245 (2004)
19. About Water. Retrieved from https://www.unenvironment.org/explore-topics/water/about-
water
20. Pimentel, D., Berger, B., Filiberto, D., Newton, M., Wolfe, B., Karabinakis, E., Nandagopal,
S.: Water resources: agricultural and environmental issues. Bioscience 54(10), 909–918 (2004)
21. Shoemaker, R.A., Harwood, J.L., Day-Rubenstein, K.A., Dunahay, T., Heisey, P.W., Hoff-
man, L.A., Klotz-Ingram, C., Lin, W.W., Mitchell, L., McBride W.D., Fernandez-Cornejo, J.:
Economic Issues in Agricultural Biotechnology (No. 1474-2016-120826) (2001)
22. Dioxide, A.C.: Status and issues concerning agricultural emissions of greenhouse gases. Agric.
Dimensions Glob. Clim. Change 229 (1993)
23. Fuhrer, J., Booker, F.: Ecological issues related to ozone: agricultural issues. Environ. Int.
29(2–3), 141–154 (2003)
24. Giller, K.E., Witter, E., Corbeels, M., Tittonell, P.: Conservation agriculture and smallholder
farming in Africa: the heretics’ view. Field Crops research 114(1), 23–34 (2009)
25. Hobbs, P.R., Sayre, K., Gupta, R.: The role of conservation agriculture in sustainable agricul-
ture. Philos. Trans. R. Soc. B: Biol. Sci. 363(1491), 543–555 (2007)
26. Tscharntke, T., Clough, Y., Wanger, T.C., Jackson, L., Motzke, I., Perfecto, I., Vandermeer, J.,
Whitbread, A.: Global food security, biodiversity conservation and the future of agricultural
intensification. Biol. Conserv. 151(1) (2012)
27. Pretty, J.N., Noble, A.D., Bossio, D., Dixon, J., Hine, R.E., Penning de Vries, F.W., Morison,
J.I.: Resource-Conserving Agriculture Increases Yields in Developing Countries (2006)
28. Bennett, A.F., Radford, J.Q., Haslem, A.: Properties of land mosaics: implications for nature
conservation in agricultural environments. Biol. Cons. 133(2), 250–264 (2006)
29. Lipper, L., Thornton, P., Campbell, B.M., Baedeker, T., Braimoh, A., Bwalya, M., Hottle, R.:
Climate-smart agriculture for food security. Nat. Clim. Change 4(12), 1068 (2014)
30. Campbell, B.M., Thornton, P., Zougmoré, R., Van Asten, P., Lipper, L.: Sustainable intensi-
fication: what is its role in climate smart agriculture? Curr. Opin. Environ. Sustain. 8, 39–43
(2014)
31. Neufeldt, H., Jahn, M., Campbell, B.M., Beddington, J.R., DeClerck, F., De Pinto, A., LeZaks,
D.: Beyond climate-smart agriculture: toward safe operating spaces for global food systems.
Agric. Food Secur. 2(1), 12 (2013)
32. Harvey, C.A., Chacón, M., Donatti, C.I., Garen, E., Hannah, L., Andrade, A., Clement, C.:
Climate-smart landscapes: opportunities and challenges for integrating adaptation and mitiga-
tion in tropical agriculture. Conserv. Lett. 7(2), 77–90 (2014)
42 D. D. Dasig Jr.

33. Steenwerth, K.L., Hodson, A.K., Bloom, A.J., Carter, M.R., Cattaneo, A., Chartres, C.J., Jenk-
ins, B.M.: Climate-smart agriculture global research agenda: scientific basis for action. Agric.
Food Secur. 3(1), 11 (2014)
34. Rosset, P.M., Martínez-Torres, M.E.: Rural social movements and agroecology: context, theory,
and process. Ecol. Soc. 17(3) (2012)
35. Altieri, M.A.: Agroecology: principles and strategies for designing sustainable farming sys-
tems. Agroecology in Action (2000)
36. Altieri, M.A.: Agroecology: The Scientific Basis of Alternative Agriculture (1983)
37. De Schutter, O.: Agroecology and the right to food. Report Presented at the 16th Session of
the United Nations Human Rights Council [A/HRC/16/49], p. 8 (2011)
38. Robert, P.C.: Precision agriculture: a challenge for crop nutrition management. In Progress
in Plant Nutrition: Plenary Lectures of the XIV International Plant Nutrition Colloquium,
pp. 143–149. Springer, Dordrecht (2002)
39. Gebbers, R., Adamchuk, V.I.: Precision agriculture and food security. Science 327(5967),
828–831 (2010)
40. Stafford, J.V.: Implementing precision agriculture in the 21st century. J. Agric. Eng. Res. 76(3),
267–275 (2000)
41. Pierce, F.J., Nowak, P.: Aspects of precision agriculture. Adv. Agron. 67, 1–85. Academic Press
(1999)
42. Adamchuk, V.I., Hummel, J.W., Morgan, M.T., Upadhyaya, S.K.: On-the-go soil sensors for
precision agriculture. Comput. Electron. Agric. 44(1), 71–91 (2004)
43. Haboudane, D., Miller, J.R., Tremblay, N., Zarco-Tejada, P.J., Dextraze, L.: Integrated narrow-
band vegetation indices for prediction of crop chlorophyll content for application to precision
agriculture. Remote Sens. Environ. 81(2–3), 416–426 (2002)
44. Mulla, D.J.: Twenty five years of remote sensing in precision agriculture: key advances and
remaining knowledge gaps. Biosys. Eng. 114(4), 358–371 (2013)
45. Basso, B., Ritchie, J.T., Pierce, F.J., Braga, R.P., Jones, J.W.: Spatial validation of crop models
for precision agriculture. Agric. Syst. 68(2), 97–112 (2001)
46. Lamb, D.W., Brown, R.B.: Pa—precision agriculture: remote-sensing and mapping of weeds
in crops. J. Agric. Eng. Res. 78(2), 117–125 (2001)
47. Schellberg, J., Hill, M.J., Gerhards, R., Rothmund, M., Braun, M.: Precision agriculture on
grassland: applications, perspectives and constraints. Eur. J. Agron. 29(2–3), 59–71 (2008)
48. Dombrowski, U., Wagner, T.: Mental strain as field of action in the 4th industrial revolution.
Procedia Cirp 17, 100–105 (2014)
49. Abramovici, M., Göbel, J.C., Neges, M.: Smart engineering as enabler for the 4th industrial
revolution. In: Integrated Systems: Innovations and Applications, pp. 163–170. Springer, Cham
(2015)
50. Falkenthal, M., Breitenbücher, U., Képes, K., Leymann, F., Zimmermann, M., Christ, M.,
Kempa-Liehr, A.W.: Opentosca for the 4th industrial revolution: automating the provisioning
of analytics tools based on apache flink. In: Proceedings of the 6th International Conference
on the Internet of Things, pp. 179–180. ACM (2016, November)
51. Dasig, Jr, D.: User experience of embedded system students on Arduino and field programmable
gate array (FPGA). In: Proceeding of the Second International Conference on Advances in
Applied Science and Environmental Engineering-ASEE, pp. 124–128 (2014, February)
52. Yahya, N.: Agricultural 4.0: its implementation toward future sustainability. In: Green Urea,
pp. 125–145. Springer, Singapore (2018)
53. Weltzien, C.: Digital agriculture or why agriculture 4.0 still offers only modest returns.
Landtechnik 71(2), 66–68 (2016)
54. Hoogenboom, G., Jones, J.W., Wilkens, P.W., Porter, C.H., Batchelor, W.D., Hunt, L.A., Boote,
K.J., Singh, U., Uryasev, O., Gijsman, A. J.: Decision support system for agrotechnology
transfer version 4.0. University of Hawaii, Honolulu, HI (CD-ROM) (2004)
55. Falkenthal, M., Breitenbücher, U., Christ, M., Endres, C., Kempa-Liehr, A. W., Leymann, F.,
Zimmermann, M.: Towards function and data shipping in manufacturing environments: how
cloud technologies leverage the 4th industrial revolution. In: Proceedings of the 10th Advanced
Summer School on Service Oriented Computing, pp. 16–25 (2016)
Implementing IoT and Wireless Sensor Networks for … 43

56. Zysman, J.: The 4th service transformation: the algorithmic revolution. Berkeley Roundtable
on the International Economy (2006)
57. Corallo, A., Latino, M.E., Menegoli, M.: From industry 4.0 to agriculture 4.0: a framework
to manage product data in agri-food supply chain for voluntary traceability. Int. J. Nutr. Food
Eng. 12(5), 146–150 (2018)
58. Rose, D., Chilvers, J.: Agriculture 4.0: responsible innovation in an era of smart farming. Front.
Sustain. Food Syst. 2, 87 (2018)
59. De Clercq, M., Vats, A., Biel, A.: Agriculture 4.0: The Future of Farming technology. In:
Proceedings of the World Government Summit, Dubai, UAE, pp. 11–13 (2018)
60. Kortuem, G., Kawsar, F., Sundramoorthy, V., Fitton, D.: Smart objects as building blocks for
the internet of things. IEEE Internet Comput. 14(1), 44–51 (2009)
61. Li, S., Da Xu, L., Zhao, S.: The internet of things: a survey. Inf. Syst. Front. 17(2), 243–259
(2015)
62. Lee, I., Lee, K.: The Internet of Things (IoT): applications, investments, and challenges for
enterprises. Bus. Horiz. 58(4), 431–440 (2015)
63. Xia, F., Yang, L.T., Wang, L., Vinel, A.: Internet of things. Int. J. Commun Syst 25(9), 1101
(2012)
64. Meyer, S., Ruppen, A., Magerkurth, C.: Internet of things-aware process modeling: integrating
IoT devices as business process resources. In: International Conference on Advanced Informa-
tion Systems Engineering, pp. 84–98. Springer, Berlin, Heidelberg (2013, June)
65. Wurm, J., Hoang, K., Arias, O., Sadeghi, A.R., Jin, Y.: Security analysis on consumer and
industrial IoT devices. In: 2016 21st Asia and South Pacific Design Automation Conference
(ASP-DAC), pp. 519–524. IEEE (2016, January)
66. Zanella, A., Bui, N., Castellani, A., Vangelista, L., Zorzi, M.: Internet of things for smart cities.
IEEE Internet of Things J. 1(1), 22–32 (2014)
67. Da Xu, L., He, W., Li, S.: Internet of things in industries: a survey. IEEE Trans. Industr. Inf.
10(4), 2233–2243 (2014)
68. Miorandi, D., Sicari, S., De Pellegrini, F., Chlamtac, I.: Internet of things: vision, applications
and research challenges. Ad Hoc Netw. 10(7), 1497–1516 (2012)
69. Stankovic, J.A.: Research directions for the internet of things. IEEE Internet of Things J. 1(1),
3–9 (2014)
70. Ronen, E., Shamir, A.: Extended functionality attacks on IoT devices: The case of smart lights.
In: 2016 IEEE European Symposium on Security and Privacy (EuroS&P), pp. 3–12. IEEE
(2016, March)
71. Blaauw, D., Sylvester, D., Dutta, P., Lee, Y., Lee, I., Bang, S., Kim, Y., Kim, G., Pannuto, P.,
Kuo, Y.-S., Yoon, D., Jung, W., Foo, Z., Chen, Y.-P., Oh, S., Jeong, S., Choi, M.: IoT design
space challenges: Circuits and systems. In: 2014 Symposium on VLSI Technology (VLSI-
Technology): Digest of Technical Papers, pp. 1–2. IEEE (2014, June)
72. Biswas, A.R., Giaffreda, R.: IoT and cloud convergence: opportunities and challenges. In: 2014
IEEE World Forum on Internet of Things (WF-IoT), pp. 375–376). IEEE (2014, March)
73. Tewari, A., Gupta, B.B.: Cryptanalysis of a novel ultra-lightweight mutual authentication pro-
tocol for IoT devices using RFID tags. J. Supercomputing 73(3), 1085–1102 (2017)
74. Zhu, Q., Wang, R., Chen, Q., Liu, Y., Qin, W.: Iot gateway: Bridgingwireless sensor networks
into internet of things. In: 2010 IEEE/IFIP International Conference on Embedded and Ubiq-
uitous Computing, pp. 347–352. IEEE (2010, December)
75. Datta, S.K., Bonnet, C., Nikaein, N.: An IoT gateway centric architecture to provide novel
M2M services. In: 2014 IEEE World Forum on Internet of Things (WF-IoT), pp. 514–519.
IEEE (2014, March)
76. Dan, L.I.U., Xin, C., Chongwei, H., Liangliang, J.: Intelligent agriculture greenhouse envi-
ronment monitoring system based on IOT technology. In: 2015 International Conference on
Intelligent Transportation, Big Data and Smart City, pp. 487–490. IEEE (2015, December)
77. Guoqiang, S., Yanming, C., Chao, Z., Yanxu, Z.: Design and implementation of a smart IoT
gateway. In: 2013 IEEE International Conference on Green Computing and Communications
and IEEE Internet of Things and IEEE Cyber, Physical and Social Computing, pp. 720–723.
IEEE (2013, August)
44 D. D. Dasig Jr.

78. Vasisht, D., Kapetanovic, Z., Won, J., Jin, X., Chandra, R., Kapoor, A., Sinha, S., Sudarshan,
M., Stratman, S.: FarmBeats: an IoT platform for data-driven agriculture. In: 14th {USENIX}
Symposium on Networked Systems Design and Implementation ({NSDI} 17), pp. 515–529
(2017)
79. Bröring, A., Schmid, S., Schindhelm, C.K., Khelil, A., Käbisch, S., Kramer, D., Teniente, E.:
Enabling IoT ecosystems through platform interoperability. IEEE Softw. 34(1), 54–61 (2017)
80. Lea, R., Blackstock, M.: City hub: a cloud-based IoT platform for smart cities. In: 2014 IEEE
6th International Conference on Cloud Computing Technology and Science, pp. 799–804.
IEEE (2014, December)
81. Huh, S., Cho, S., Kim, S.: Managing IoT devices using blockchain platform. In: 2017 19th
International Conference on Advanced Communication Technology (ICACT), pp. 464–467.
IEEE (2017, February)
82. Jayaraman, P.P., Palmer, D., Zaslavsky, A., Georgakopoulos, D.: Do-it-yourself digital agricul-
ture applications with semantically enhanced IoT platform. In: 2015 IEEE Tenth International
Conference on Intelligent Sensors, Sensor Networks and Information Processing (ISSNIP),
pp. 1–6. IEEE (2015, April)
83. Yick, J., Mukherjee, B., Ghosal, D.: Wireless sensor network survey. Comput. Netw. 52(12),
2292–2330 (2008)
84. Akyildiz, I.F., Su, W., Sankarasubramaniam, Y., Cayirci, E.: Wireless sensor networks: a survey.
Comput. Netw. 38(4), 393–422 (2002)
85. Mao, G., Fidan, B., Anderson, B.D.: Wireless sensor network localization techniques. Comput.
Netw. 51(10), 2529–2553 (2007)
86. Al-Karaki, J.N., Kamal, A.E.: Routing techniques in wireless sensor networks: a survey. IEEE
Wirel. Commun. 11(6), 6–28 (2004)
87. Van Dam, T., Langendoen, K.: An adaptive energy-efficient MAC protocol for wireless sensor
networks. In: Proceedings of the 1st International Conference on Embedded Networked Sensor
Systems, pp. 171–180. ACM (2003, November)
88. Xiangning, F., Yulin, S.: Improvement on LEACH protocol of wireless sensor network. In:
2007 International Conference on Sensor Technologies and Applications (SENSORCOMM
2007), pp. 260–264. IEEE (2007, October)
89. Doherty, L., El Ghaoui, L.: Convex position estimation in wireless sensor networks. In: Proceed-
ings IEEE INFOCOM 2001. Conference on Computer Communications. Twentieth Annual
Joint Conference of the IEEE Computer and Communications Society (Cat. No. 01CH37213),
vol. 3, pp. 1655–1663. IEEE (2001, April)
90. Ye, W., Heidemann, J., Estrin, D.: Medium access control with coordinated adaptive sleeping
for wireless sensor networks. IEEE/ACM Trans. Netw. (ToN) 12(3), 493–506 (2004)
91. Beckwith, R., Teibel, D., Bowen, P.: Report from the field: results from an agricultural wireless
sensor network. In: 29th Annual IEEE International Conference on Local Computer Networks,
pp. 471–478. IEEE (2004, November)
92. Garcia-Sanchez, A.J., Garcia-Sanchez, F., Garcia-Haro, J.: Wireless sensor network deploy-
ment for integrating video-surveillance and data-monitoring in precision agriculture over dis-
tributed crops. Comput. Electron. Agric. 75(2), 288–303 (2011)
93. Wark, T., Corke, P., Sikka, P., Klingbeil, L., Guo, Y., Crossman, C., Bishop-Hurley, G.: Trans-
forming agriculture through pervasive wireless sensor networks. IEEE Pervasive Comput. 6(2),
50–57 (2007)
94. Keshtgari, M., Deljoo, A.: A wireless sensor network solution for precision agriculture based
on zigbee technology (2011)
95. Van Bodegom, P.M., Stams, A.J.M.: Effects of alternative electron acceptors and temperature
on methanogenesis in rice paddy soils. Chemosphere 39(2), 167–182 (1999)
96. Watanabe, H., Takagi, K., Vu, S.H.: Simulation of mefenacet concentrations in paddy fields by
an improved PCPF-1 model. Pest Manag. Sci. Formerly Pestic. Sci. 62(1), 20–29 (2006)
Bayesian Network for Development
of Expert System in Pest Management

Niranjan Singh and Neha Gupta

Abstract Timely availability of decision support to the farmers on ‘whether and

what management option is required’ is imperative for effective pest management.
In the absence of knowledge and expertise, farmers are over-dependent on pesticide
dealers for pest management decision support in the country, which results in exces-
sive, injudicious and irrational use of chemicals for the pest control and thus not
only degrades the environment but also affects human health. For decades, the pest
economic threshold level (ETL) has been the basis to select appropriate pest manage-
ment option for decision support to the farmers, used in ICT-based expert systems
for decision support in agriculture. This process requires quantitative information
about pest activity, which needs to be scientifically observed in the farmers’ fields.
Moreover, in current pest management, decision-making depends upon a large range
of agro-ecological information, besides pest activity. However, a large section of the
farming community is not able to obtain this kind of information rather can observe
qualitative information consisting of uncertainty. Neither there are the methods of
selecting appropriate pest management option based on tentative pest relevant infor-
mation, provided by farmers that can be used in expert systems of decision support
in pest management. Bayesian Network (BN), the most suitable probabilistic rea-
soning approach for dealing with uncertain information has been effectively used in
ecological decision-making nevertheless, very little efforts have been for its usage
in pest management in the country. Application of Bayesian network may greatly
assist in development of innovative methods of selecting appropriate management
option based on uncertain or tentative information of identified agro-ecological fac-
tors besides crop condition that farmers provide. Application of the methods thus
developed in information and communication technology (ICT) based expert sys-
tems (ESs) of agriculture would automate and speed up the process of providing
real-time decision support on insect-pest management to the farmers.

N. Singh · N. Gupta (B)

FCA, MRIIRS, Faridabad, India
e-mail: [email protected]
N. Singh
e-mail: [email protected]

© Springer Nature Singapore Pte Ltd. 2020 45

P. K. Pattnaik et al. (eds.), Internet of Things and Analytics
for Agriculture, Volume 2, Studies in Big Data 67,
https://doi.org/10.1007/978-981-15-0663-5_3
46 N. Singh and N. Gupta

1 Introduction

In the absence of knowledge and expertise, farmers are over-dependant on pesticide

dealers for pest management decision support, which results in excessive, injudi-
cious and irrational use of chemicals for the pest control. This not only degrades
the environment but also affects human health due to retention of pesticide residues
in vegetables [1]. The major decision concern of farmers for pest management is
‘whether and what pest management actions are required’. Over the years, pest ETL
has been the basis of pest management decision-making for selection and timing of
pest management option if necessary, but this process requires quantitative informa-
tion about pest activity, which needs to be scientifically observed from the farmers’
fields. However, a large section of the farming community is not able to obtain this
kind of information scientifically. Moreover, in current pest management, decision-
making depends upon a large range of agro-ecological or pest relevant information
[2], beside pest activity. Timely availability of correct advice on pest management
decision-making to the farmers can result either in saving crop worth crores of rupees
or in non-application of intervention saving the cost of intervention involved and thus
saving the environment [1].
There have been various kinds of tools of decision support such as warning ser-
vices, onsite devices and expert systems, used to provide support on decision-making
in crop pest management. However, expert systems are the most significant and
are widely used. Innovative methods or algorithms are important components of
any dynamic expert system in pest management. There are no methods available
to deal with pest relevant agro-ecological information consisting uncertainty that
farmers provide for decision-making in management option. Bayesian network, a
key computer technology dealing with probabilities in Artificial Intelligence, is one
of the most effective and popular methods to deal with uncertain information and
has been widely and effectively used in ecological decision-making. However, there
have been little efforts to apply Bayesian network in development of innovative
methods/techniques to deal with qualitative agro-ecological information consisting
uncertainty for decision-making in crop pest management. Application of innovative
algorithms/methods in expert systems of agriculture providing decision support sys-
tems on selection of appropriate pest management option, based on the tentative pest
relevant information obtained by the farmers can facilitate and speed up the process
of providing pest management decision support to the farmers.

2 Decision-Making in Pest Management

Decision-making is a mental process resulting in the selection of an action among

several alternative solutions. Every decision-making process produces a final choice
[3]. Decision-making starts with the identification of a problem, which requires the
collection of all relevant information for critical analysis of the problem. This analysis
Bayesian Network for Development of Expert System … 47

leads to development of a set of available alternative courses of actions to solve the

problem; only realistic solutions should be selected considering multiple criteria,
e.g. effectiveness, benefits, costs and the constraints. Based on this analysis, the best
solution is selected, and the decision is converted into an action [4]. Compared to
other economic activities, decision-making in crop protection has received very little
attention [5]. A better understanding of the decision-making process is necessary
given the increased complexity of the decisions required in pest management.
The core of pest management framework is the decision-making process.
Decision-making in pest management is a dynamic and complex process (Fig. 1)
which requires much more knowledge and support than the conventional agricul-
ture. The first decision concerns pest identification and second decision concerns
‘whether and what pest management option is required’. Whenever the decision-
maker notices any pest activity while scouting the field, he needs to identify the
pest and scientifically observe the pest and other relevant information necessary for
pest management decision-making. After the analysing the information observed,
first decision to be made is whether pest management action is required and if so,
what is the most appropriate option; chemical, biological or physical; their doses,
application methods, frequency, etc. In effective pest management, sustainable bio-
logical, physical and other non-chemical methods, which are safer to environment
and human health, are preferred to chemical methods [4].

Fig. 1 Decision-making process in pest management

48 N. Singh and N. Gupta

2.1 Basis of Decision Support in Pest Management

For several decades, pest economic threshold level (ETL) has been the basis for
decision support on pest management to the farmers, which requires scientifically
observed quantitative information about pest activity from farmers’ fields. ETL is
an important pest management concept which integrates biology and economics and
uses pesticides, or other management actions, only when economic loss in anticipated
[6]. ETL is defined as the lowest population density that will cause economic damage
at which control measures should be initiated against an increasing pest population
to prevent economic damage. It is expressed quantitatively and is calculated based
on management cost per hectare, price of the farm produce/kilogram and expected
damage (kilogram/pest). ETL is the best-known and widely used concept for pest
management decision support [7], but there are many reasons for not to use this
concept. Major problem of using ETL as the basis for pest management decision
support is that it requires quantitative pest information scientifically observed from
farmer’s fields which large farming community is not able to provide. Moreover, it
is based on parameters that are changing all the time, and that are often not known
[8].
However, present-day pest management emphasises on agro-ecological situation
of farmer field wherein decisions are based on large range of other pest relevant
information [9] such as crop health, natural enemies and weather. Agro-ecological
information-based pest management results in more reduction in chemical pesticide
usage than ETL based and thus conserves the agroecosystems. Pest management
decisions cannot be based on only a simple pest count but have to consider many other
aspects such as crop health, natural enemies and weather condition. Therefore, agro-
ecological information-based pest management decision support is more holistic
approach than ETL. Even agro-ecological situation-based pest management decision
support need quantitative information on the said aspects, but farmers can provide
the qualitative information on these aspects, which consists of uncertainties.

2.2 Pest Management Expert Systems and Their Limitations

An agricultural expert system is a decision support system for agricultural extension

agents who have to decide what advice to be offered to farmers and in turn who have
to decide what action to take based on it. An important issue for pest management is
to develop dynamic and interactive expert tools of decision support that can facilitate
the process of providing timely and correct information on timing and selection of
appropriate pest management intervention [10]. There are many pest management
options available, and however, most appropriate one is selected based on more
than one criterion. The timing of its application is imperative for the success of the
Bayesian Network for Development of Expert System … 49

intervention. Therefore, it becomes necessary to use knowledge-intensive decision-

making for designing and developing expert systems to enhance the pest management
programmes.
A properly designed expert system for decision support assists the managers or
decision-makers in obtaining helpful information from raw data, documents, personal
knowledge, and methods for identification and solution of the problem and making
appropriate decision. In any process of decision-making, identification of decision-
maker is the first step and very important, since all other process relevant elements
depend upon it. Alternative decisions and outcomes are other essential elements of
a decision-making process. Not all decision outcomes are appropriate; therefore,
decision-maker has to assess them based on their advisability. A pest management
expert system of decision support gives all the required information about decision-
making criteria and management options to the user [11].
Generally, a pest management expert systems of agriculture provide all requisite
information or expertise about pest identification; pest and pathogen life cycle; pest
sampling methods and threshold calculators; decision-making criteria, available bio-
control and chemical control methods; and their safety and environmental impact
issues to the domain users. Many initiatives have been taken for the development of
expert systems in the country by various public and private organisations. However,
most of them are simple web-based static information systems; only few are the
dynamic, providing real-time expert support for decision-making to the farmers based
on pest situation in their fields. Major initiatives taken in the field of crop pest
management in the country as well as world over have been summarised in Tables 1
and 2, respectively.
One of the main reasons for the poor application of expert systems of decision
support in pest management in the country has been the failure of developers to
target farmers as user, not pay attention to their level of knowledge, their need and
requirements and furthermore to the environment in which they work. Most of the
EXs developed in the country are either web-based databases or information systems
targeted to researchers, KVKs, extension workers or State Department of Agriculture
(SDA) personnel. Another lacuna of these systems is that they push information to
the users rather than collecting real-time pest information from the farmer’s field,
process the same and then advice the appropriate pest management solution. For
effective and successful pest management, farmer needs easy and timely decision
support on selection of appropriate pest management option [34] based on the pest
situation in their fields.
Most of the available pest management EXs in the country are either web-based
information systems or web-based databases that address only specific aspect of
pest management, i.e. pest identification and their management strategies, whereas
the most important pest management decision concern of the farmers is to have
decision support on ‘whether and what pest management actions are required’.
Few expert systems provide ETL-based real-time decision support to the farmers
by capturing quantitative pest information scientifically obtained by trained pest
scouts from farmer fields. These systems produce ETL-based pest reports showing
whether pest incidence has reached near ETL or crossed ETL’ to the experts to select
50 N. Singh and N. Gupta

Table 1 Brief description of major expert systems of pest management developed in India
Reference Name of the system Area of application Decision support on
Sharma et al. [12] OPMAS Pest management ETL-based real-time
decision support for
cotton pest
management
Cpsskerala.in [13] CPSS Pest management Pest identification
and field pest
condition-based crop
pest management
Iffcokisan.com [14] IFFCO Kisan Agriculture Market prices,
weather forecast,
agro-advisory,
animal husbandry,
horticulture
Padma et al. [15] IDSS for Apple Pest management Integrated pest
management of apple
orchard
Dath and Expert System on Pest management Pest management in
Balakrishnan [16] Coconut Pest coconut crop
Management
Prabhuraj [17] Electronic solutions Pest management ETL-based real-time
for agri. pests decision support for
(e-SAP) pest management in
selected crops
Hasan et al. [18] CaneDES Nutritional disorders Diagnosis of various
and pest management kinds of disorders in
sugarcane
Kumbhar and Singh e-Sagu Crop production Advice on crop pest
[19] management
Katkamwar et al. [20] AGREX Crop production and Fertilizer application
protection and disease diagnosis
in rice
Sharma et al. [21] E-Pest Reporting and Pest management ETL-based real-time
Alert system decision support for
pest management in
pulses
Marwaha [22] AgriDaksh Crop production Pest identification
and variety selection
Islam et al. [23] EXOWHEM Crop production Wheat agrimony,
varieties and pest
management
Bayesian Network for Development of Expert System … 51

Table 2 Brief description of major expert systems of pest management developed world over
Reference DSS name Area of application Decision support on
Martin and Goebel DECIPESTS Pest management Pest identification and
[24] their level based pest
management in
sugarcane
Asi.ucdavic.edu [25] UC IPM Pest management Pest identification and
location specific pest
management
decision-making
Pest.cabweb.org [26] Crop protection Pest management Pest identification and
compendium management
Carhuapoma et al. Insect life cycle Pest management Temporal temperature
[27] modelling (ILCYM) scenario-based pest
distribution and risk
maps.
Damos and Real agro-net Crop health Real-time moth
Karabatakis [28] topology and Insect phenology forecasts
smart model
application
(RANTISMA)
Orellana et al. [29] SAIFA Pest management Integrated production
for the olive crop
Vincent et al. [30] HDSS Pest management Weather-based pest
forecasting and their
management
Samietz et al. [31] SOPRA Pest management Insect-pest
monitoring and their
management in fruit
orchards
Bourgeois et al. [32] CIPRA (Computer Pest management Weather-based
centre for agricultural prediction of pest,
pest forecasting) crop phonological and
postharvest disorders
affecting storage of
apple in Canada
Magarey et al. [33] NAPPFAST North Pest management Plant pest forecasting
Carolina State system
University (NCSU)
52 N. Singh and N. Gupta

appropriate pest management option to be advised to the farmers. However, scientific

observation of such quantitative pest information from farmers’ fields requires large
number of qualified and trained workforce, which is a difficult and costly proposition
for state agencies. Therefore, can the farmers themselves do the field pest scouting?
Yes, they can. However, large farming community cannot record the quantitative pest
observations as per pest-specific scientific methods rather can observe the qualitative
information consisting of uncertainty. Moreover, in the present-day pest management
decision-making, other relevant factors such as the presence of natural enemies,
weather and crop condition affecting the level of pest activity are to be considered
necessarily for determining the appropriate pest management option. However, there
are not methods available to deal with this kind of qualitative pest relevant information
consisting uncertainty for selection of pest management options, which can be used
in pest management expert systems.

3 Bayesian Network and Its Application

in Decision-Making

Bayesian network, a key computer technology dealing with probabilities in Artificial

Intelligence, is one of the most effective and popular methods of modelling uncertain
knowledge expression and reasoning such as environmental management [35, 36].
BNs emerged from artificial intelligence research wherein originally they emerged
as formal modes to analyse decision approaches under uncertain conditions [37].
Bayesian network can be used to estimate the effectiveness of alternate management
decisions or policies because of being causal in nature. BNs have the ability to sustain
clarity through explicit causal assumptions and so are generally used to model when
the relationships to be described are not easily expressed using mathematical notation.
New computational methods and techniques keep increasing BNs’ abilities and range
of practical applications [38].
In contrast to the other approaches or techniques used in environment studies,
Bayesian networks use probabilistic expressions rather than deterministic to express
the relationships among variables of the system. Bayesian network accounts the lack
of knowledge in the network by the usage of Bayesian probability theory, which
allows the subjective estimation of the probability of occurrence of a particular out-
come that is to be combined with more objective data quantifying the frequency of
occurrence in finding the conditional probabilistic relationships. Bayesian networks
are very appropriate technique to deal with systems where uncertainty is inherent
due to it being accounted in the model, as this is an important issue in ecological
systems [39].
Bayesian networks have many other attractive properties, which make them par-
ticularly applicable in data analysis and decision-making. BNs have simple causal
and graphical structure and can be easily extended and altered. BNs can integrate
missing data readily by using Bayes’ theorem. They can easily be understood without
Bayesian Network for Development of Expert System … 53

much mathematical background and knowledge. They have been found to possess
good accuracy of prediction with small sample size. BNs can also be applied for
predicting the probable values of states of a system for different future scenarios.
Bayesian networks are also useful for participatory processes. They can assist in
evaluating the alternative decisions to optimise the desired outcome. BNs can also
help in processes of social learnings. BNs express the system as network of recip-
rocal actions among system variables from main cause to outcome; subject to all
assumptions of cause-effect are clearly made [39]. Reasoning approaches are very
useful in dealing with uncertain information [40].
A Bayesian network is a combination (G; P) [41, 42] where
• G = (V; E) is a DAG where set of nodes V = {X1, X2, …, Xn} represents the vari-
ables of the system and E, a set of arcs represents direct conditional dependencies
between the variables (nodes);
• P represents the set of conditional probability distributions comparing conditional
probability distribution P (Xi/pa (Xi)) for each variable X given the set of parents
pa (Xi) in the graph.
The joint probability distribution over V can be recovered from the set P of
conditional probability distributions by the application of following chain rule:


n
P(X 1, X 2, . . . , X n) = P (Xi/pa (Xi))
i

Conditional probabilities of BNs can be computed based on the experimental or

trial data, results produced by the models and domain expert’s knowledge elicitation
[43]. Uncertainty is clearly represented by way of presentation of the probabilities
[44].

3.1 Bayesian Network Theory

BNs are probabilistic graphical model consisting of variables of interest of a system

and their conditional relationships [45]. A BN consists of ‘nodes’ that represent
the set of variables of interest and ‘links’ represent conditional interdependencies
(cause-effect relationship) among nodes via a directed acyclic graph (DAG) [46].
Conditional probability distribution captures the quantum of dependence between
system variables and thus explains the relationships between the nodes. For instance,
if node A of a system affects another node B, then nodes A and B be linked to each
other by an arc having direction from node A to node B. Herein, node A is called as
parent of node B, and node B is referred to as child of node A. If node B further affects
another node C, then node B be the parent of node C. Figure 2 shows the conditional
or causal relationship among nodes A, B and C of a basic Bayesian network.
54 N. Singh and N. Gupta

Fig. 2 Conditional or causal

relationship among BN
nodes

A BN can comprise different kinds of nodes, i.e. nature nodes, decision nodes and
utility nodes. Nature nodes are the variables, which can be controlled by actions of
decision-maker and represent the empirical or calculated parameters and the proba-
bilities of occurrence of various states. Input nodes are the nodes without parents and
can be expressed in as either constant or categorical states with associated marginal
probability distributions. The variables or events, which can be directly implemented
by the decision-maker, are represented by a decision node. Decision node represents
two or more alternative decision options that a manager can choose from. A decision
node does not have probabilities associated with it. Utility nodes [47] should always
accompany decision nodes. A utility node represents the value of an outcome or a
decision. Utility node can be directly linked to the decision node [48]. BNs repre-
senting and solving decision problems under uncertainty are also known as Bayesian
decision networks [49].
Bayesian network uses Bayes’ probability theorem to propagate informa-
tion/evidences among the nodes. Bayes’ theorem defines how observed evidence
updates the prior information/knowledge about a hypothesis. In Bayes’ theorem,
prior probability represents the possibility of an input parameter being in a specific
state whereas conditional probability estimates the possibility of a parameter state
given the states of input parameters affecting it. Estimate of likelihood of a parameter
to be in particular state is the posterior probability, given the input parameters, con-
ditional probabilities and the rules controlling how the probabilities combine. The
network is solved using Bayes’ rule:

P (A|B) = P (B|A) P (A)/ P (B)

Where P (A) is the prior probability of parameter A; P (A|B) is the posterior

probability, the probability of A given new data B; and P (B|A) the conditional
probability of parameter B given existing data A [39].

3.2 Major Instances of Bayesian Network in Agriculture

Bayesian networks (BNs) are widely being used for environmental modelling to inte-
grate multiple issues and components of the system; for utilising information from
various sources; and to handle the missing and uncertain data [47]. BNs have been
adequately used in ecological decision-making [50]. They have also been used in
agriculture nonetheless; there are very few instances of their usage in pest manage-
ment [51].
Bayesian Network for Development of Expert System … 55

Tari [52] illustrated the possible applications of BNs for assessing the yield
response of winter wheat to fungicide programmes. Therein, experimental data from
ADAS trials has been used with some assumptions in order to complete the data.
Gu et al. [53] applied Bayesian network to handle the uncertainty in climate
prediction information for studying risk and benefits to crop production in Scotland.
The BN establishes relationship between model of fababean growth and a model of
weather generator and thus is able to respond to the queries on fababean production
under different climate predictions.
Borsuk et al. [54] developed a Bayesian network to know about possible reasons
for reduction in the status of health of brown trout population in Switzerland. Further
investigation was done by Borsuk et al. [55], wherein BN was applied to determine
the importance of historical causes of anthropogenic changes and also for predicting
the effects of management actions proposed.
Bromley et al. [44] developed Bayesian network for integrated management of
four basins from Europe while focussing on management of water. The project aim
was to devise an active and functional method to help the managers in several types of
objective decision-making and simultaneously making sure that various stakeholders
actively participate in the process of decision-making.
Smith et al. [56] also devised a Bayesian network, which interacted with expert
knowledge and geospatial data about the favoured habitats to map the suitability of
Julia Creek dunnart (Sminthopsis douglasi) habitat in north-west Queensland. These
were earlier thought to be extinct [39].
Chen and Yu [57] presented application of Bayesian network for development
of a disease identification system for maize crop. Noisy-or model and conversion
from certainty to probability were applied in the development of Bayesian network.
Thus, a BN-based disease identification system for maize crop was developed. The
practice proves Bayesian network as an effective tool for maize disease diagnosis.
Steventon and Daust [58] developed a BN for modelling the epidemic of moun-
tain pine beetle in north-central British Columbia, Canada. Subsequently, this was
applied for examining the response of habitat and population size of pine beetle to
management scenarios with built in parameter uncertainty.
Meysam and Mahdi [59] studied the BN usage for locust pest management. How-
ever, it was used to find the effect of various trap plants on population. This was an
instance of using Bayesian network for examining the effects of trap plants on locust
population.
Douglas and Newton [60] presented about usage of BNs for supporting protected
area management by way of developing habitat suitability models to address the
conservation of eight species. The BNs were developed using scientific literature and
expert knowledge and were evaluated with the field survey results. Subsequently, they
were applied to see the effect of various management practices such as tree cutting,
grazing and burning on suitability of habitat of species.
Abbal et al. [61] developed a DSS tool for vine-growers to examine quality of
vineyard to be planted. Herein, vineyard quality is determined by the probability
of wine sale profitability. Bayesian network being widely used to deal with uncer-
tainty in management of natural resource was constructed using the relationship
56 N. Singh and N. Gupta

between environment and vineyard status parameters. Information for Bayesian net-
work construction was obtained from vine growing experts. The system facilitates
the assessment of possible impact of vine grower’s decisions on the quality of new
vineyards.

4 Conceptual Bayesian Network for Decision-Making

in Pest Management

Construction of conceptual Bayesian network for pest management system comprises

the following components:
• Framework of pest management system
• Structure of Bayesian network for decision-making in pest management
• Parameterization of the Bayesian network
Construction of BN structure for pest management decision support required iden-
tifying the important variables, establishment of links between variables and com-
putation of conditional probabilities. Crop protection system includes field, plants,
pests and their natural enemies.

4.1 Framework of Pest Management System

In crop protection, all the decisions related to pest management are made by the
farmers at field level. Regular crop monitoring is the cornerstone of pest management.
Figure 3 shows the components of field monitoring in crop pest management.
During field monitoring, farmer has to observe all the pest relevant agro-ecological
information and crop condition to estimate the pest risk to the crop. Crop condition is
measured in terms of crop phenology. Weather and the presence of natural fauna are
the most important factors affecting pest activity. These observations are recorded
weekly. Based on the estimate of overall field agro-ecological situation, he has to
timely decide ‘whether and what pest management action is required’ for effective
and successful pest management.

4.2 Structure of Conceptual Bayesian Network

Bayesian networks are strictly one-way because they represent a conditional prob-
ability distribution that describes the relative likelihood of each value of the ‘child’
node (end of the arrow) conditional on every possible combination of values of the
‘parent’ nodes (start of each arrow). For instance, if a link is directed from node E
towards node H as in Fig. 4, then each value of node E conditionally depends upon
Bayesian Network for Development of Expert System … 57

Fig. 3 Components of field monitoring for crop pest management

Fig. 4 Basic structure of a

Bayesian network

the possible combination of values of node H since E and F parent nodes are having
causal relationship with child node H.
BN structure is usually made of ‘box and arrow’ diagram. A BN can be constructed
using two approaches, using expert knowledge elicitation and based on data learnings.
Data learning-based approach is useful to construct BN for complex systems that are
difficult to be understood. Pest management decision-making processes are complex
in nature wherein quality and comprehensive historical data is rarely available. Hence,
it becomes difficult to determine accurate structure based on data learnings only. So,
the outcome of the data-based structural learnings can be enhanced by combining
it with expert knowledge elicitation together. While developing the conceptual BN
structure, all variables (nodes) of the system influencing the outcome of decision-
making process are considered.
Identifying all the variables of the system and establishing links between them
results in structure of conceptual Bayesian network representing the system under
consideration. Description of pest management framework and knowledge elicited
from experts and published literature revealed prominent variables influencing
58 N. Singh and N. Gupta

Fig. 5 Structure of conceptual BN for decision-making in pest management

decision-making in pest management system are weather (temperature, rainfall and

humidity), pest activity, the presence of pest natural enemies (parasite and predators)
and crop condition.
Weather comprising temperature, rainfall, humidity; the presence of pest natural
enemies comprising predators and parasites; and pest activity and crop condition are
the broad factors to be considered for determining ‘whether and what pest manage-
ment intervention is required’. Using these variables, structure of conceptual BN for
decision-making in pest management was established as shown in Fig. 5.
Once structure of the conceptual BN established, we can start with it to have
refined version of the network. Identification of significant variables (‘nodes’) of the
system and exclusion of non-significant variables are usually done based on review
of the literature, data analysis, experts’ opinion and discussion with the stakeholders.
Once the nodes have been selected, links are revised based on the relationship among
them. If a variable is not at all related to the decision-making in the system, it may
be excluded because inclusion of insignificant variables can make the BN structure
more complex. Since inclusion of insignificant variables costs extra time and efforts
but it does not add any value to the BN.
Significance of these factors depends upon the location and cropping season.
A factor may be significantly affecting pest activity in a particular location and
cropping season, while other may not be having any significant effect. Thus, using
data records obtained from organisations working in the field of pest management and
the knowledge elicited from domain experts and literature, key variables (nodes) of
the system relevant to the process of decision-making in pest management and their
relationship with each other are identified. The variables, which are not significant,
are excluded, and thus, the structure of conceptual network is refined. Mutually
Bayesian Network for Development of Expert System … 59

exclusive states of each variable (node) are also determined through the analysis of
data records and domain expert inputs.

4.3 Parameterisation of the Bayesian Network

After identification of nodes, they are assigned the states to be taken by each of
them. Each BN node represents either an observable or measurable system variable.
Potential values or conditions that can be assumed by a node are represented by
its states. Type and quality of available data determines the types of the state and
their number. The initial values for each node can be determined using available
data records or based on the knowledge elicited from the literature or through expert
consultation. The states of a node may be categorical, e.g. large, medium, small,
Boolean, e.g. yes or no, discrete such as integers or continuous variables, e.g. ≤10,
>10 and ≤20 and >20. However, the decision about the division of range into sub-
ranges should not be arbitrary rather it should be done based on data analysis such as
descriptive, plotting, classification or percentile analysis. Discretisation can be useful
wherever a variable has specific breakpoint that is significant to management. Domain
expert consultation may also be combined with data analysis. Sometimes, defining
fewer state of a variable result in information loss and assigning too many states to a
variable may complicate the BN. For Bayesian networks of decision-making, states
representing the important thresholds can be used as alternatives.
As stated above, weather conditions, the presence of pest natural enemies, level
of pest activity and crop health are the prominent factors broadly considered for pest
management decision-making. Weather conditions are described by a set of variables,
i.e. maximum temperature, minimum temperature, rainfall, humidity; presence of
natural enemies by population of predators and population of parasites‚ as the vari-
ables, whereas pest activity and crop health are not represented as set of variables. Set
of possible mutually exclusive states of these variables or nodes in conceptual BN
for decision-making in pest management may be categorised as shown in Table 3.
After defining the state type and its numbers, conditional probabilities for the
state values of each child node are estimated based on the state values of its parent

Table 3 Possible state values

Variable (s) Possible state values
of prominent variables of
decision-making in pest Temperature Extremely hot, hot, warm, cold, frigid
management Rainfall High, moderate or low
Humidity High, moderate or low
Presence of predators Abundant, moderate, low, scarce
Presence of parasites Abundant, moderate, low, scarce
Pest activity High, medium, low
Crop health Good, average, poor
60 N. Singh and N. Gupta

nodes. Probability of priory expecting a node to be into a definite state can be deter-
mined from known frequencies or can assume a uniform distribution representing
uncertainty [48].
The relationship between parent and child node is defined by the conditional
probability table (CPT). The CPT presents the probability value of a node being in
a state provided combination of probability values of states of parent nodes. CPT
dimension of each node is the multiple of numbers of child node states and its
parent nodes states. If a node has no parents, it can be described probabilistically
by a marginal probability distribution. Figure 6 shows the CPT of a basic Bayesian
network consisting of three nodes. Node H will be in its high, medium and low states,
given the states of node E and F in the network. The nodes E and F are root nodes and
so are defined by marginal probabilities. H is the child node of E and F nodes, and
thus, probabilities of H node states are conditional given the combination of states
of E and F nodes. Parent nodes E and F are the causing factors of child node.
The CPT values associated with each state of a node can be estimated using
expert knowledge elicitation, observed data, probabilistic or empirical equations
and results of simulation models or their combination [62]. Probability distribution
quantifies the uncertainties associated with each relationship. It is very important to
record probability computation method including any assumptions and limitations.
Construction of Bayesian network is complete once the probabilities of all nodes are
computed. Once completed, BN is used for scenario analysis. Different scenarios be
it set of management actions or system observations, can easily be analysed using the
BN. Bayesian networks are simple means of scenario analysis that allows the users
to propagate evidences as node input using well-defined node distribution. The effect
of a scenario is tested through its effect on other nodes of the network by propagating
the probabilities. Fast propagation of evidence through the network is one of the main

Fig. 6 CPT of node in basic structure of a BN

Bayesian Network for Development of Expert System … 61

advantages of BNs. Therefore, BNs are used to instantly find the effect of observed
conditions or decisions at one node in the whole system. Subsequently, the network
is evaluated or validated.

5 Evaluation and Validation of the Bayesian Network

Evaluation and validation of a developed Bayesian network assist in determining

whether its results are feasible and satisfactory. The criteria to evaluate or validate
BN’s performance depend upon its objective. Initially, the interactions of different
components of the network can be examined by application of various input scenar-
ios and determine whether the results are logical and reasonable. Expected results
can include the network showing realistic performance. Preferably accuracy of the
network performance should be examined with empirical data; however, many a
times data is either not available or is partially available. Data other than that used
to develop the Bayesian network should be used for evaluation and validation of
the model. Many BN softwares have the functionalities to allow data to be checked
against network predictions. These softwares revise the probabilities of all the sam-
ples within the case, except the unobserved nodes and then produce the beliefs of
each unobserved node. These values are compared with actual node value. Another
method of BN evaluation is sensitivity analysis, which is done to find sensitive param-
eters. This analysis ranks the variables according to the importance relative to the
variables of the interest. This can be used to confirm whether response of network is
correct and matches to the expectations. Sensitivity analysis also assists in identifi-
cation of the variables having significant influence on the results, and subsequently
[62]. There are other forms of evaluation such as critique of assumptions, ability
of the model to perform under different scenarios [63], whereas validation of the
Bayesian decision network can be done by cross-examining the outcome of the BN
against the results of existing methods of decision-making.
In case of limited availability or complete unavailability, qualitative forms of
evaluation and validation for instance peer review are valuable. By application of
different input values, outcome of the network in the form of resulted probabilities
can be examined by the reviewers whether it is consistent and logical. Aguilera et al.
[64] did a review of over hundred BN applications in environmental management and
found that 38% did not do any validation. This may be because Bayesian network is
frequently applied in cases of scarce data availability. Network evaluation should be
done even with limited data. In case, evaluation or validation results are not logical
and reasonable then network structure and its assumptions need to be reassessed that
includes readjustment of network structure, refinement of conditional probabilities
or variable states.
62 N. Singh and N. Gupta

Development of BN is an ongoing process. Major advantage of BN is that is can

be easily updated. The BNs are very useful in cases of limited data and knowledge
availability and thus can be built with scare or incomplete information.
Environmental and crop pest management are the systems complex in nature,
where data availability is ether scarce or limited. Therefore, evaluation and validation
of BNs developed for these systems can be done using combination of data and exert
knowledge elicitation.

6 Conclusion

Many expert systems have been developed to provide expert decision support in pest
management by private and public organizations world over. However, most of them
are web-based database or information, which provides information on pest identifi-
cation and management. There have been very little efforts to develop interactive and
dynamic expert systems providing real-time decision support to the farmer based on
the uncertain agro-ecological situation in their fields. There are no methods available
to deal with uncertain agro-ecological information provided by farmers for selec-
tion of appropriate pest management option to be advised to them. Application of
Bayesian network, a probabilistic reasoning approach, can help in development of
innovative method or technique for selecting appropriate pest management option
based on uncertain agro-ecological information that farmers provide. Application of
method or technique thus developed speed up and automate the process of providing
real-time decision support to the farmers based on agro-ecological situation in the
their fields.

References

1. Singh, N., Sardana, H.R., Kumar, S.S.: Pest management information system (PMIS) for
tomato: a database tool for IPM. Indian J. Entomol. 74(2), 202–204 (2012)
2. Satyagopal, K., Sushil, S.N., Jeyakumar, P., Shankar, G., Sharma, O.P., Sain, S.K., Saha, S.:
AESA based IPM package for tomato. National Institute of Plant Health Management, Hyder-
abad, AP (2014)
3. March, J.G.: Primer on decision-making: how decisions happen, p. 289. The Free Press, New
York, NY, USA (1994)
4. Rossi, V., Caffi, T., Salinari, F.: Helping farmers face the increasing complexity of decision-
making for crop protection. Phytopathologia Mediterranea 51(3), 457–479 (2012)
5. Eom, S., Kim, E.: A survey of decision support system applications (1995–2001). J. Oper. Res.
Soc. 57(11), 1264–1278 (2006)
6. Damos, P.: Modular structure of web-based decision support systems for integrated pest man-
agement. A review. Agron. Sustain. Dev. 35(1), 1347–1372 (2015)
7. Uniyal, M.: Integrated Pest Management: SATNET Asia Training Manual. Retrieved from
http://www.satnetasia.org (2015)
8. Jeykumar, P.: AESA Based IPM. Retrieved form http://vikaspedia.in/agriculture/crop-
production/integrated-pest-managment/aesa-based-ipmdn (2016)
Bayesian Network for Development of Expert System … 63

9. Prakash, A., Bantur, J.S., Prasad, M.S., Tanwar, R.K., Sharma, O.P., Bhagat, S., … Jeykumar,
P.: Integrated Pest Management for Rice. National Centre for Integrated Pest Management,
New Delhi (2014)
10. Damos, P., Savopoulou-Soultani, M.: Temperature-driven models for insect development and
vital thermal requirements. Psyche 2012(1), 1–13 (2012)
11. Damos, P., Agron, S.D.: Modular structure of web-based decision support systems for integrated
pest management. Agron. Sustain. Dev. 35(4), 1347–1372 (2015)
12. Sharma, R., Kumar, A., Tanwar, R.K., Singh, N.: Web-enabled data extraction system for cotton
pests. Indian J. Agric. Sci. 88(6), 841–844 (2018)
13. Cpsskerala.in.: Crop-Pest Surveillance System. Retrieved from http://www.cpsskerala.in
(2017)
14. Iffcokisan.com: AgroICT Services. Retrieved from http://www.iffcokisan.com (2016)
15. Padma, T., Mir, S.A., Shantharajah, S.P.: Intelligent decision support system for an integrated
pest management in apple orchard. In: Sangaiah, A., Abraham, A., Siarry, P., Sheng, M. (eds.)
Intelligent Decision Support Systems for Sustainable Computing, pp. 225–245. Springer, NY,
USA (2017)
16. Dath, A., Balakrishnan, M.: Expert system on coconut pest management. Int. J. Adv. Res.
Comput. Sci. Softw. Eng. 6(3), 442–446 (2016)
17. Prabhuraj, A.: Electronic Solutions Against Agricultural Pests (e-SAP). Retrieved from
http://www.mmp.cips.org.in/documents/Workshops/2015/30jul-1aug/eSAP_A_Prabhuraj.pdf
(2015)
18. Hasan, S.S., Solomon, Baitha A.S., Singh, M.R., Sah, A.K., Kumar, R., Sudhir, S.: CaneDES,
a web-based expert system for disorder diagnosis in sugarcane. Sugar Tech 17(4), 419–427
(2015)
19. Kumbhar, V., Singh, T.P.: A comprehensive study of application of decision support system in
agriculture in Indian context. Int. J. Comput. Appl. 63(14), 6–11 (2013)
20. Katkamwar, N., Kumar, B., Kalaskar, K.: An expert system approach for improvement of
agriculture decision. Singaporean J. Sci. Res. 5(1), 8–12 (2013)
21. Sharma, O.P., Singh, N., Bhardwaj, A., Vennila, S., Bhagat, S., Mehta, N., Kumar, K.: ICT
based pest management system for sustainable pulse production, a case study. In: Maumbe,
B., Patrikakis, C. (eds.) E-Agriculture and Rural Development, Global Innovations and Future
Prospects, pp. 169–183. IGI Global, Hershey, PA, USA (2013)
22. Marwaha, S.: AgriDaksh—a tool for developing online expert system. 3rd National Conference
on Agro-Informatics & Precision Agriculture (AIPA) proceedings, New Delhi, India, 1–3
August 2012, pp. 18–23 (2012)
23. Islam, S.N., Kundu, S., Sabir, N., Sharma, K., Farooqi, M., Poswal, R., Agarwal, H.O.,
Chaturvedi, K., Sharma, R.K., Sharma, A.: Selection of wheat Triticum aestivum variety
through expert system. Indian J. Agric. Sci. 82(1), 43–47 (2012)
24. Martin, P., Goebel, F.R.: DECIPESTS, a decision-support system for pest management in
sugarcan. Int. Sugar J. 11(9), 978–982 (2017)
25. Asi.ucdavic.edu: UC IPM portal. Retrieved from http://asi.ucdavis.edu/programs/sarep/what-
is-sustainable-griculture/practices/ipm (2017)
26. Pest.cabweb.org: CABI-Crop Protection Compendium. Retrieved form http://pest.cabweb.org/
cpc/cpchp.htm (2017)
27. Carhuapoma, P., Sporleder, M., Mujica, N., Kreuze, J., Simon, R., Kroschel, J.: Insect life
cycle modelling ILCYM—an open source tool for entomologists. In: Presented in Interna-
tional Conference on Statistics and Big Data Bioinformatics in Agricultural Research, CRISAT,
Hyderabad, 21 November 2016 (2016)
28. Damos, P., Karabatakis, S.: Real-time pest modelling through the World Wide Web, decision-
making from theory to praxis. Integr. Protect. Fruit Crops 91(1), 253–258 (2013)
29. Orellana, F.J., del Sagrado, J., del Águila, I.M.: SAIFA, A web-based system for integrated
production of olive cultivation. Comput. Electron. Agric. 78(2), 231–237 (2011)
30. Vincent, P.J., Jay, F.B., Gary, G.G., Brad, P., Gerald, V.T., Wendy, E.J.: A web-based decision
support system to enhance IPM programs in Washington tree fruit. Pest Manag. Sci. 66(6),
587–595 (2010)
64 N. Singh and N. Gupta

31. Samietz, J., Graf, B., Höhn, H., Höpli, H.U., Schaub, L.: SOPRA, phenology modelling of
major orchard pests—from biological basics to decision support. Acta Hortic. 803(1), 35–42
(2008)
32. Bourgeois, G., Plouffe, D., Chouinard, G., De Ell, J.: The apple CIPRA network in Canada,
using real-time weather information to forecast apple phenology, insects, diseases and physi-
ological disorders. Acta Hort. 803, 29–34 (2008)
33. Magarey, D.R., Fowler, G., Borchert, D.M., Sutton, T.B., Colunga-Garcia, M., Simpson, J.A.:
NAPPFAST, An internet system for the weather-based mapping of plant pathogens. Plant Dis.
91(4), 336–345 (2007)
34. Bajwa, W.I., Leonard, Coop, Kogan, M.: Integrated pest management (IPM) and internet-based
information delivery systems. Neotropical Entomol. 32(3), 373–383 (2003)
35. Bi, C., Chen, G.: Bayesian networks modeling for crop diseases. In: Li, D., Liu, Y., Chen, Y.
(eds.) Computer and Computing Technologies in Agriculture IV, CCTA 2010: 4th IFIP TC 12
Conference proceedings, Nanchang, China, 22–25 October 2010, pp. 312–320 (2011)
36. Uusitalo, L.: Advantages and challenges of Bayesian networks in environmental modelling.
Ecol. Model. 203(3–4), 312–318 (2007)
37. Varis, O.: Bayesian decision analysis for environmental and resource management. Environ.
Model Softw. 12(2–3), 177–185 (1997)
38. Mead, R., Paxton, J., Sojda, R.: Applications of Bayesian networks in ecological modelling. In:
The Second IASTED International Conference on Environmental Modelling and Simulation.
St. Thomas, Virgin Islands (2006)
39. Pollino, C.A., Henderson, C.: Bayesian Networks, A Guide for Their Application in Natural
Resource Management & Policy. Retrieved from http://www.utas.edu.au/data/assets/pdf_file/
588474/TR_14_BNs_a_resource_guide.pdf (2010)
40. Pan, H., McMichael, D.: Fuzzy Causal Probabilistic Networks—A New Ideal and Practical
Inference Engine. Retrieved from http://citeseerx.ist.psu.edu/viewdoc/download? (1998)
41. Jensen, F.V., Nielsen, T.D.: Bayesian Networks and Decision Graphs, 2nd edn. Springer-Verlag,
NY, USA (2007)
42. Kjrulff, U.B., Madsen, A.L.: Bayesian Networks and Influence Diagrams. Springer-Verlag,
NY, USA (2008)
43. Borsuk, M.E., Reichert, P., Peter, A., Schager, E., Burkhardt-Holm, P.: Assessing the decline of
brown trout (Salmotrutta) in Swiss rivers using a Bayesian probability network. Ecol. Model.
191(1–2), 224–244 (2006)
44. Bromley, J., Jackson, N.A., Clymer, O.J., Giacomello, A.M., Jensen, F.V.: The use of Hugin
to develop Bayesian networks as an aid to integrated water resource planning. Environ. Model
Softw. 20(2), 231–242 (2005)
45. del Águila Isabel, M., del Sagrado, J.: Metamodeling of Bayesian networks for decision-support
systems development. In: 8th Workshop on Knowledge Engineering and Software Engineering
(KESE), Montpelier, USA, 949, 12–19 (2012)
46. Pearl, J.: Probabilistic Reasoning in Intelligent Systems: Networks of Plausible Inference.
Morgan Kaufmann Publishers Inc., San Francisco, USA (1988)
47. Singh, N., Gupta, N.: Bayesian network for decision support on pest management of tomato
fruit borer, H. armigera. Int. J. Eng. Technol. 6(4), 168–170 (2017b)
48. Kragt, M.E.: A Beginners Guide to Bayesian Network Modelling for Integrated Catchment
Management (Technical Report No. 9, Landscape Logic). Department of Environment, Water
Heritage and the Arts, Australia (2009)
49. Singh, N., Gupta, N.: Decision-making in integrated pest management & Bayesian network.
Int. J. Comput. Sci. Inf. Technol. 9(2), 31–36 (2017a)
50. Rivot, E., Prevost, E., Parent, E., Bagliniere, J.L.: A Bayesian state-space modelling framework
for fitting a salmon stage-structured population dynamic model to multiple time series of field
data. Ecol. Model. 179(4), 463–485 (2004)
51. Fabre, F., Pierre, J.S., Dedryver, C.A., Plantegenest, M.: Barley yellow low dwarf disease
risk assessment based on Bayesian modelling of aphid population dynamics. Ecol. Model.
193(3–4), 457–466 (2006)
Bayesian Network for Development of Expert System … 65

52. Tari, F.: A Bayesian network for predicting yield response of winter wheat to fungicide pro-
grammes. Comput. Electron. Agric. 15(2), 111–121 (1996)
53. Gu, Y., Crawford, J.W., Peiris, D.R., Grashoff, C., McNicol, J.W., Marshall, B.: Modelling
fababean production in an uncertain future climate. Agric. For. Meteorol. 79(4), 289–300
(1996)
54. Borsuk, M.E., Stow, A.C., Reckhow, K.H.: Ecological prediction using causal Bayesian net-
works: a case study of eutrophication management in the Neuse River estuary. Duke University,
North Carolina, USA (2002)
55. Borsuk, M.E., Stow, A.C., Reckhow, K.H.: A Bayesian network of eutrophication models for
synthesis, prediction and uncertainty analysis. Ecol. Model. 173(2–3), 219–239 (2004)
56. Smith, C.S., Howes, A.L., Price, B., McAlpine, C.A.: Using a Bayesian belief network to predict
suitable habitat of an endangered mammal—The Julia Creek dunnart Sminthopsis douglasi.
Biol. Cons. 139(3–4), 333–347 (2007)
57. Chen, G., Yu, H.: Bayesian network and its application in maize diseases diagnosis. In: Li,
D. (eds.) Computer and Computing Technologies in Agriculture-II proceedings. Wuyishan,
China, 18–20 August 2007, pp. 917–992 (2008)
58. Steventon, J., Daust, D.K.: Management strategies for a large-scale mountain pine beetle out-
break, Modelling impacts on American martens. For. Ecol. Manage. 25(7), 1976–1985 (2009)
59. Meysam, M., Mahdi, S.: Bayesian network and pest management: a case study of trap plants
on locust population: Adv. Environ. Biol. 4(2), 147–151 (2010)
60. Douglas, S.J., Newton, A.C.: Evaluation of Bayesian networks for modelling habitat suitability
and management of a protected area. J. Nat. Conserv. 22(3), 1–12 (2014)
61. Abbal, P., Sablayrolles, J., Matzner-Lober, E., Boursiquot, J., Baudrit, C., Carbonneau, A.: A
decision support system for vine growers based on a Bayesian network. J. Agric. Biol. Environ.
Stat. 21(1), 131–151 (2015)
62. Pollino, C.A., White, K.A., Hart, B.T.: Examination of conflicts and improved strategies for
the management of an endangered Eucalypt species using Bayesian networks. Ecol. Model.
201(1), 37–59 (2007)
63. Jakeman, A.J., Letcher, R.A., Norton, J.P.: Ten iterative steps in development and evaluation
of environmental models. Environ. Model Softw. 21(5), 602–614 (2006)
64. Aguilera, P.A., Fernández, R., Fernández, A., Rumi, R., Salmerón, A.: Bayesian networks in
environmental modelling. Environ. Model Softw. 26(12), 1376–1388 (2011)
An IoT and Wireless Sensor
Network-Based Technology
for a Low-Cost Precision Apiculture

D. D. Dasig Jr. and J. M. Mendez

Abstract The current environmental ramifications due to industrial agriculture,

climate change, and pathogens have destructed the habitat and loss of biodiversity.
These ecological modifications have threatened the apiculture including the bees
and wild pollinators which plays a vital role and are indispensable in the terrestrial
ecosystems. This chapter presents the foundational space of apiculture, challenges,
and emerging techniques of the apiarists in beekeeping including precision apicul-
ture, and a low-cost IoT and remote sensing technology deployed for data gathering
and monitoring the vigor and productivity of beehive colonies. The system has redun-
dancy components and utilized the three-level hierarchical model including wireless
node, local information server, and the cloud data server. The system also consists
of sensors deployed across distinct locations to capture the beehives’ temperature,
relative humidity, and loads of the hives and honeycombs, and stored to the local and
cloud server. This system will complement the urgent need to countervail the honey-
bee colonies’ collapse by helping the apiarists, apiculture subsector, and agriculture
sector to embark emerging precision apiculture technologies.

Keywords Precision apiculture · Internet of things · Wireless network sensors ·

Beekeeping · Bees

D. D. Dasig Jr. (B)

Graduate Studies, College of Science and Computer Studies, De La Salle University-Dasmariñas,
DBB-B, West Ave, 4115 Dasmariñas, Cavite, Philippines
e-mail: [email protected]
J. M. Mendez
College of Computer Studies and Engineering, Lorma Colleges,
San Fernando La Union 2500, Philippines
e-mail: [email protected]

© Springer Nature Singapore Pte Ltd. 2020 67

P. K. Pattnaik et al. (eds.), Internet of Things and Analytics
for Agriculture, Volume 2, Studies in Big Data 67,
https://doi.org/10.1007/978-981-15-0663-5_4
68 D. D. Dasig Jr. and J. M. Mendez

1 Introduction

Internet of things (IoT) has allowed countless sectors and companies to boost effec-
tiveness through the implementation of centralized surveillance solutions for the most
significant procedures over the previous few years. In apiculture or beekeeping, one
such instance can be discovered. Honeybees are creepy crawlies or the insects of sig-
nificant of critical monetary worth and by pollinating plants provide such a crucial
provider to agriculture, to the apiculture particularly. Honeybees face an astonish-
ing population decrease without a clear comprehension of the cause. This situation
brought and causes serious and genuine natural, money, or economical related and
environmental mishaps, also known as the disorder from the colony. Inadequate
information collection is one shortcoming in investigating and handling colonies of
honeybees. Sensor systems that track in-hive features continually could enable fresh
methods for research and apiary leadership.
The pollinators of insects are vital to forest ecosystems and many farmer and agri-
cultural enterprises. They guarantee that certain environmental procedures, including
plant reproduction, are maintained. Bees are the most advanced insect pollinators
because they can effectively carry and store pollen [1]. Honeybees are essential not
only for their manufacturing of honey but also for the equilibrium of the ecosystem
since they are crucial for the pollination of plant flowers [2].
Honeybees are environmental insects and live in big, well-structured family com-
munities. Honeybees are commonly known as social insects. Superorganism as
referred to by entomologists or the social insects are heavily engaged and devel-
oped insects that perform many complex tasks that are not performed by many
isolation insects. Interaction, complicated nest design, environmental monitoring,
labor defense, and partition are just a few of the behaviors established in the social
colonies. These astonishing behaviors are one of the most interesting beings in the
world, social insects generally and honeybees.
Typically, three types of adult bees form a honeybee colony. A colony is composed
of a queen, the drones, and the workers. In nest construction, several thousand work-
ing bees collaboratively work together to build their nest including those activities of
gathering their foods and brood rearing. Each worker has a specific task to perform
in connection with their adult age. But survival and reproduction take the colony as
a whole combined effort [3].
There is only one queen in each colony, except during and after swarming or
substitute activities. The fact is that the queen is the only sexually developed bee
from the women; her primary role in the colony is reproduction. The queen spawns
pollinated and unfertilized eggs. In the springtime and early summertime, the queens
in the beehives lay the highest number of eggs including fertilized and the unfertilized
ones. The queens can lay eggs of approximately up to 1500 eggs daily.
In early October, they gradually stop laying eggs and generate limited number of
eggs or no eggs at all by early following spring (January). A queen lays up to 250000
eggs per year, and over a million lives are being produced by queen.
An IoT and Wireless Sensor Network-Based Technology … 69

The most important bees in the colony are drones (male bees). They are usually
only present in late summer and spring. The head of the drone is far bigger than the
queen and the worker’s compound eyes tally at the uppermost of the dome. Drones
do not have a stinger, wax glands, or pollen baskets. The drones mainly fertilize the
virgin king when she sees her, but this function is performed only by a tiny number of
drones. About a week after emergence, drones mature sexually and die immediately
after maturation. While drones do not perform such good job for the colony, its
presence is considered essential for the ordinary functioning of the colony.
Worker bees are the lowest corporate adults, and most bees are present in the group
or colony. They are women who are sexually developed and do not lay spawns under
ordinary hives. Workers are structured with speciality facilities like brood glands for
food, perfume glands, waxing glands, and pollen baskets that enable them to carry out
the entire hive’s work. They are purifying and polishing cells, feeding larvae, taking
care of the queen, collecting waste, handling incoming nectar or a sap, constructing
peas, watching the entrance and air condition, and ventilating the peat in the original
weeks as adolescents. They drill for the water, sap, pollens, and propolis (plant sap)
later as held bees.
At the point when a colony winds up to a point that it becomes queenless, the
ovaries of a few specialists or worker bees create and laborers start to lay unfer-
tilized eggs. Similarly, the development of the laborers’ ovaries is accepted to be
hindered by the nearness of brood and the queen and her synthetic substances or
chemicals. The manifestation of laying laborers or worker bees in a colony describes
and implies that the colony has been queenless for at least one to several weeks
already. Notwithstanding, laying laborers or worker bees may additionally be found
in ordinary “queenright” colonies throughout the swarming period and as soon as
the colony is being ruled and controlled by a pitiable queen.
Every one of the three sorts of grown-up bees goes through three formative stages
before developing as grown-ups: egg, hatchling or larvae, and pupa. The three phases
are all things considered brood. While the formative stages are comparable, they do
vary in terms. Unfertilized eggs become rambles or the drones, while those fertil-
ized ones become laborers or worker bees, and or the queens in the colonies. The
colonies’ health and nutrition are primarily important, and significantly, it impacts
in development of female honeybees; hatchlings or larvae bound to end up laborers
or worker bees, and get less illustrious jelly and more a blend of nectar and pollens
contrasted with the bounteous measures of imperial or royal jelly that the colony
ruler or queen larvae would receive.
Physical variables, including weight, temperature, humidity, air gases, vibrating,
sound, and forager traffic, are continuously monitored in most scientists as the price
and size of electronic detectors reduce as their accuracy and capability rise. To obtain
data about the behavior and phenology of colonies, researchers adopted various
methods to collate the results.
Temperature affects the behavior of bees. Rarely do they operate when it is below
57 °F or above 100 °F. If the temperature is less than 55 °F, they cannot fly. Bees gather
outside shaded hives during very warm days and do not operate. But in the summer,
too much shade is irritating for bees. Honeybees keep the brood nest temperature
70 D. D. Dasig Jr. and J. M. Mendez

between 32 and 35 °C dynamically so that the brood usually delves. Once the nest
temperature is too high, the bees ventilate the warm water out of the nest by fanning
or use processes for evaporation.
Using accuracy apiculture, it enables apiarists to control beehives for many poten-
tial purposes as mentioned in several studies, data analytics and best practices on bee-
keepers’ real-time monitoring of bees [4, 5], and learning how to reduce the resources
and time allocated in beekeeping activities without compromising the quality and
productivity of honeys.
This paper’s general objective is twofold: to develop a wireless sensor system
for Internet surveillance of beehive microclimate variables, including temperature,
relative humidity, and weight that could enhance techniques of apicultural research
and development. The second objective of the study is to utilize IoT architecture and
wireless sensor network in the design and development of low-cost in-hive models
for precision apiculture.
The users of the study project are La Union’s beekeepers in the Philippines. The
project aimed to assist them to take care of the bees and verify their returns regularly.
Beekeepers of La Union often have the ability and skills to be a beekeeper of the world
class. Not only did this project create things simpler for beekeepers, but it also made
beekeeping a company or career more in demand. The unit has been implemented on
one of La Union’s beekeepers, making the bee farm more efficient and exceptional.
Table 1 exemplifies the honey-gathering calendar and the flow of honey. The
season of honey supply is the busiest moment in the year of a beekeeper. The best
way for the beekeeper to assess the arrival and termination of the honey flow is to
control adjustments in the hives’ weight. The honey flow month is between February
and March and finishes in May. The beekeeper also extracts surplus honey from the
hive during the honey flow season. June to October month is the season in which bee
has low honey owing to the season’s shortage of food. Therefore, beekeepers feed
them sugar for them to survive in the rainy season.
For the beekeepers to monitor the beehive, a mobile application was developed for
the project. The mobile application displays the current temperature and humidity,
the current weight of the hive, and the bee counter which indicates the bees that
go inside and outside the hive [6, 7]. The mobile application would require Internet

Table 1 Honey calendar

Honey flow activities Month
1 2 3 4 5 6 7 8 9 10 11 12
Pre-honey flow season x
Honey flow season x x x
Harvesting time x x x
Post-honey flow x
Critical period for feeding bees x x x x x
1 January, 2 February, 3 March, 4 April, 5 May, 6 June, 7 July, 8 August, 9 September, 10 October,
11 November, 12 December
An IoT and Wireless Sensor Network-Based Technology … 71

connectivity for the user to monitor the beehive wherever they are. The application
also sends weekly reports of the status of the hive; this serves as an overview of the
beehive in the past days, and it would make the beekeeper to maintain the beehive.
A database would also be developed for storage of data that comes from the hive,
and this would make data easier to access. The beehive must also have connectivity
to the Internet so data that will be gathered inside would be sent to the database.
This project explored the possibilities of Internet of things in beekeeping for
enhancing or substituting manual inspection of hives from a user-centered perspective
and at the same time provides an automated cooling mechanism to assist the colony.
Using IoT-enabled devices for monitoring beehives made the work of collecting data
more efficient. Since the observed beehives might be in distant places, it is preferred
to use technology to record and to transfer data [4, 6].
The research project has the following constraints: The project is purely for mon-
itoring and data-gathering purposes only; it cannot act if the honey is ready to be
harvested. The beekeeper is still the one responsible for maintaining the good status
of the honeybees and their beehives. The weight of the bees and the honeycomb
will also be included in the measurement of the weight of the beehive; hence, the
accuracy of measuring the weight is reduced.

2 Related Literature

Though the honeybee is frequently associated with honey manufacturing, honeybees’

function in agriculture is predominantly for pollination, allowing for fruit develop-
ment. Pollination is the fertilization phase of the spermatophyte or fruit-bearing
plant development process. Male gametophyte cells are contained in pollen. The
male pollen grains germinate when transmitted to the stigma of a flower and reach
the female gamete cells, and this method allows for the growth of reproduction and
fruit. The wind acts as pollen vectors along with birds, insects, and other livestock.
In the first years of the twentieth century, interest in constant monitoring of hon-
eybee colonies started. In 1914, a manually gathered data was released related to
beehive temperature in 1907 for several days [8]. In 1926, thermocouples were used
to control the temperature in a beehive [9]. The sensor technology development and
electronic data acquisition systems have enabled the measurement procedures to be
improved. Different kinds of tracking techniques were used in this manner, from
easy observation of information in the hive [10] to systems that can analyze those
data [11], while a summary of some of the continuous monitoring methods was
concurred by specifying the parameters (temperature, O2 , CO2 , relative moisture,
weight, vibration, etc.) and the duration of each technique.
Apidictor [12] was among the first electronic bee monitoring devices. This scheme
comprised of a low-pass frequency filter to detect sound changes within the hive
up to two to three weeks before swarming. An analysis of sound captured on a
computer with a sound card and controlled temperature and relative humidity through
a pyrometer was also conducted [13]. Three omnidirectional microphones captured
72 D. D. Dasig Jr. and J. M. Mendez

the noise. Researchers also defined an imaging system at the entrance to the in-hive
[14] to find out how many times a bee entered and left the hive. To recognize the bee
under research, they used a tag connected to the body of the bee.
In another study, supervised bees’ wintering and measuring the beehive tem-
perature were investigated. The information was sent to a processing computer to
determine the hive status by applying an algorithm [15], while RFID scheme was
utilized to determine the frequency and duration of honeybee queens’ nuptial jour-
neys [16]. A 12-bit resolution data logger was used to evaluate the evolution of a
beehive’s weight [16] by looking at the hourly weight of two hives in a solar panel
operated system [17].
The improvement of the performance and size of the microcontrollers has allowed
the development of low-cost beehive monitoring systems based on Arduino® , Make® ,
Sparkfun® , and Adafruit® [18]. An example of an Arduino-based monitoring system
is proposed by a group of researchers [19]. The system used an Excel database to
store the temperature and relative humidity information in a microSD memory card.
To be able to process this information later, the beekeeper had to go to the beehive
colony and download the SD memory card content for each hive to his/her computer.
Honeybee hive surveillance systems based on WSNs have been used in the past
few years. These devices have the benefit of being able to monitor the environ-
mental circumstances of the hive remotely in a non-intrusive manner. Technological
advancements paved the way toward better apiculture with another study using a pri-
mary unit scheme that communicated wirelessly with each node in the hive, acting
as an Internet gateway at the same time [20]. Using an embedded SHT15 sensor,
the system tracked the temperature and relative humidity. The information gathered
was sent to a remote database server by the primary unit. Another development on
apiculture was the Raspberry Pi-based temperature monitoring scheme as an Internet
gateway [21]. The temperature sensors situated through a one-wire network in the
hives linked to the Raspberry Pi. If only one Raspberry Pi was used for all beehives,
the writers noted the low price of the solution.
In the advent of wireless sensor networks in precision agriculture, a wireless sensor
network (WSN) was used to monitor a hive colony and gather the most important
information about the beehive activity/environment and its surroundings [22]. Each
beehive has been monitored by two nodes, and each node has been implemented by
Libelium (Zaragoza, Spain) through a Waspmote® . Several heterogeneous sensors
(temperature, CO2 , pollutants, NO2 , etc.) were used to monitor the honeybees in
the hive. Two commercial extension boards have been used, one for each Waspmote
because of the large number of sensors. At a frequency of six samples/day, each
sensor was read. The gathered specimens were stored for backup on an SD card and
transmitted to the base station once every 24 h on a Zigbee network. The base station
was a bridge, sending the information over a 3G/GSM link to the remote server. Also,
a surveillance scheme was used to suggest multiple algorithms in order to detect
changes in the hive and warn the apiculturist of prospective threats to the colony
[23]. A weather algorithm has been suggested based on the CO2 concentrations of
the hive.
An IoT and Wireless Sensor Network-Based Technology … 73

Similarly, researchers have designed and developed a scheme for measuring the
temperature of the beehive using an LM35 sensor using an Arduino platform [24].
The information has been transmitted wirelessly to an Internet-connected desktop
base station via an XBee-PRO module. The records were processed to discover
patterns to detect future absconding circumstances in the heat stress in the shed. The
information was gathered.
In another study, it described an environmental control system to monitor bee
activity frequency [25]. The system monitors the frequency of bee entering and
exiting from the hive, as well as indoors and outdoors, the temperature and moisture.
Accordingly, the precision of the frequency was based on hive activity, a 10% average
mistake and 20% maximum.
The beam also adheres to Poisson’s ratio, where a beam, when in tensile straining,
will extend in length in the direction of the force, but its cross-sectional area (CSA)
will contract. When a beam experiences bending, its top surface experiences a tensile
strain and its bottom surface experiences a compressive strain (depending on the
direction of bending). If the beam is behaving elastically, then the strain is directly
proportional to stress, and hence to the applied force. These changes in the beam can
be used to measure the force applied precisely, by means of strain gauges [26].
There have been recent advances in the development of load cells, pressure sen-
sors, and strain gauges which made these technologies more attractive for use in
weight measuring applications. Load cells have become a standard in beehive weigh-
ing systems. In the application of a single, robust load cell used to weigh an entire
beehive [14], the maximum weight the load cell can accommodate is 200 kg, which
satisfies the requirements of an average-sized beehive. The jig used to translate the
weight of the beehive to the load cell allows for uniform weight distribution and can
cope should there exist an imbalance of weight occur in the beehive.
A prototype was developed to collect images and audio within a beehive [27]. The
platform utilizes a Libelium Waspmote, and Raspberry Pi in order to process and store
the data. Additionally, microphones, accelerometers, thermal and infrared cameras
were used, along with emergency notifications to the user in case an undesirable
event occurs to the beehive. The goal was to utilize the said equipment’s to collect
data in a discreet manner.
A proposed framework to classify system-planning methodologies was also made
available for selecting a proper methodology to develop an information system that
is a key factor for its success [28]. The framework aimed at identifying appropriate
methodology utilized to implement myBee and was concurred successful based on
the obtained observations. In addition, the authors mentioned that several methodolo-
gies for implementing and planning information systems are available; thus, myBee’s
architecture is flexible to be utilized in various fields, which simplifies and stream-
lines the development process.
74 D. D. Dasig Jr. and J. M. Mendez

3 Methodology

The researchers employed descriptive-developmental research design. A descriptive

secondary data analysis was used by looking into the manual records, active observa-
tion, documentation, databases of the apiculturists as inputs to the project planning.
During the development phase, prototyping was employed to allow the researchers
to improve the system quality based on the requirements’ specifications. It also facil-
itated user involvement in the development making way for reduced time and cost.
The researchers utilized the IoT-architectured embedded system and wireless sensor
networks in designing the low-cost precision apiculture. The system was tested and
piloted in the province of La Union, Philippines, where apiculture is one of the viable
economic means.

3.1 The IoT Architecture of Beehive

In the design framework, the IoT architecture with three domains or layers was
considered in the system including the IoT device, IoT gateway, and IoT platform
layer as illustrated in Fig. 1. The IoT device layer is the IoT architectural component
responsible for the observation and data gathering. The input domain includes the
sensing instrumentation tools, counter, temperature sensor, moisture sensor, and the
weight sensor of the hives.
In this architecture, the IoT gateway layer is responsible for the connectivity, the
protocol, data aggregation, filtering and processing, data security, data management.

Fig. 1 IoT architecture of beehive

An IoT and Wireless Sensor Network-Based Technology … 75

A wireless sensor network (WSN) was developed using a low-energy adaptive clus-
tering hierarchy (LEACH) while Zigbee, GSM/GPS, and Wi-Fi protocols were used
in the system.
The third layer of the architecture is the IoT platform. This layer affords the
cloud computing services in the system to provide public access and data sharing
capabilities for the end-users. This layer allows system provisioning through the
“Beehive” online apiculture management system with a mobile application. The IoT
platform layer allows management as well as the automation of connected devices
in this IoT-architectured low-cost precision apiculture.

3.2 Beehive Hardware Components

The study envisioned to improve productivity in beekeeping of the apiculturists with-

out modifying the internal components and configurations of beehives commonly
known as a man-made nest of any bee colony. In these modern days, Langstroth
hives are commonly used by apiculturists which is a modular beehive that is verti-
cally oriented with entrance for the bees. The hives have also boxes and frames for
brood and honey. In this study, the hives developed by the researchers are illustrated
in Fig. 2 with the hardware requirements for bees counting and hive status. The
embedded system of this project is enclosed in this hive.
Beehive embedded system includes the IoT devices such as sensors and actua-
tors, microcontroller, the Arduino compatible modules for IoT gateways, and was
programmed and configured according to the desired system functions.

Fig. 2 Langstroth beehive type

76 D. D. Dasig Jr. and J. M. Mendez

Fig. 3 Arduino ATmega328P

3.2.1 Beehive Microcontroller (Arduino ATmega328P)

A microcontroller used in the system is the Arduino ATmega328 as illustrated in

Fig. 3. It consists of both a physical programmable circuit board with and IDE.
ATmega328 is an input and output processing low-voltage microcontroller. The pro-
gram that comes from the Arduino IDE also was stored. ATmega328 is used as an
Arduino off-board and will work the way it works. The latter is one of the most
important parts of this de-appliance because it is necessary for all the device’s major
parts DHT22, HX711, and ESP8266.

3.2.2 Temperature and Humidity Sensor

In this IoT-architectured precision apiculture, one of the most important data-

gathering sensors in the IoT device layer is the temperature and moisture sensor.
In precision apiculture, the efficiency and effectivity of measuring the conditions of
the hive can be improved even without human intervention, and it is very crucial
that these ambiental conditions such temperature and moisture can be determined.
The sensor DTH22 is compatible with the development board Arduino ATmega328P
that was used in the system. The sensor utilizes thermistor and a moisture resistor to
evaluate temperature. The element is linked to the digital pin ATmega328 and placed
inside the machine to monitor the beehive’s temperature and humidity (Fig. 4).

3.2.3 Load Cells (HX711) Information

Apiculturists need to remove the hexagonal prismatic wax cells or the honeycomb
from the hives and make observations to know the amount of honey in the hives. These
removals of honeycomb activities interfere and lower the production in the apiculture
An IoT and Wireless Sensor Network-Based Technology … 77

Fig. 4 Temperature and humidity sensor

industry due to interfering factors to bee colony activities of honey production. In

this system, apiculturists need not remove the honeycomb from the hives.
The design utilized a semi-bridged weight sensor and an HX711 Load Cell-
Amplifier as illustrated in Fig. 5 to measure the study quantity of honey in the
hives. HX711 is a precision 24-bit analog-to-digital converter (ADC) developed for
weighing scales purposes that can be used in small scale to industrial application by
having it interfaced with a bridge sensor. In this design, a 200 kg maximum weight
can be measured in the scheme.
Further, to improve the system’s ability to measure the honey in the hives, multiple
sensors were used simultaneously. In this context, modifications on the configuration
were performed in HX711 Load Cell-Amplifier at the level of the millivolt; it is used
to make these modifications to the level that can be read by the Arduino development
board. The load cells are connected to the HX711 sensor, and the HX711 sensor
was connected to the Arduino development module so that Arduino can evaluate
the incoming information and transmit to the server. In this research, there is a

Fig. 5 Load sensor (HX711)

78 D. D. Dasig Jr. and J. M. Mendez

Fig. 6 Bee counter (QRE1113)

proportional placement of four load cells on the plywood tray and a plywood shelf
and to be connected to a mobile module.

3.2.4 Bee Counter (QRE1113) Information

Apiculturists need to learn and monitor the honeybee colony’s reproduction or

swarming situation. This can be done by monitoring how many bees left the original
hive location with the old queen. A prime swarming happens when the queen bee
left the colony with 60% or a large group of workers, and these activities should
be tracked by the system using counter. The QRE1113 was used for this purpose
to afford important advantages for the apiculturists such that swarming is a good
indication of healthy hives (Fig. 6).

3.2.5 Wi-Fi Module

In this IoT-based precision apiculture architecture, the ESP8266 Wi-Fi module was
utilized as an IoT gateway layer. In this layer, the Wi-Fi gateway serves the connec-
tivity services between the IoT devices and the IoT platform. ESP8266 is a low-cost
SOC with TCP/IP protocol stack which allowed the Arduino module to access the
Wi-Fi network and provide services and connection points between the temperature
and humidity sensors, and load sensors to the Beehive online and mobile application.
An IoT and Wireless Sensor Network-Based Technology … 79

Fig. 7 Wi-Fi module (ESP8266)

This module also supports additional security in the IoT network as well as in data
security and transports (Fig. 7).

3.2.6 Solar Charger

A solar charger as illustrated in Fig. 8 is a voltage or current regulator that is connected

to the solar panel. It prevents the battery from overcharging. The solar panel absorbs
sunlight and will generate direct current DC energy from it. It will then be converted
into an alternating current AC to make it usable. A battery bank will be used to store
energy for long-lasting and non-stop consumption. The battery bank will give life to
the whole device and will serve as the main source of power.

Fig. 8 Battery charger

80 D. D. Dasig Jr. and J. M. Mendez

3.3 Beehive System Architecture

In this developed precision apiculture technology, beehive IoT devices were expected
to fulfill functional criteria: (i) Use a minimally invasive technique, (ii) operate in dis-
tant fields for lengthy periods, and (iii) enable real-time surveillance and reporting. It
was highlighted and defined that various automatic surveillance system architectures
for real-time beehive surveillance system afforded numerous advantages [26].
This system employed the IoT architecture with three domains or layers includ-
ing the IoT device layer, IoT gateway layer, and IoT platform layer. The system
utilized five stages of IoT architecture including the sensors and actuators; IoT gate-
way, protocols, and data acquisition; edge computing, datacenter management, and
cloud computing; and reporting. Temperature and humidity sensors were utilized to
gather and communicate Ambiental parameters of the hives including temperature
and humidity sensors, load sensor through wireless links and communicate to the
cloud. For the system data aggregation and Internet gateways, a wireless sensor net-
work (WSN) was developed with a planned spatially deployed sensor nodes in La
Union, while GSM/GPS and Wi-Fi TCP/IP protocols were used in the IoT gateway.
Cloud computing services were embedded to provide public access and data shar-
ing of Beehive application with a mobile application for the end-users (beekeepers
and professional apiculturists). On the last stage of the architecture, a system dash-
board is available for system administrators, and vectoral or numerical forms sent to
the users (Fig. 9).

Fig. 9 System architecture

An IoT and Wireless Sensor Network-Based Technology … 81

4 A Detailed Look of the System

A schematic diagram illustrated in Fig. 10 has been specifically designed to allow

configurations on specific functionalities (components) in developing flexible and
agile maintenance system architecture. In this diagram, all sensors are linked to the
ATmega328 together with the Wi-Fi module. The ESP8266 works as it connects
from ATmega and transmits data over the Internet.
The DHT22 receives its data from the ATmega328 and measures the temperature
and humidity inside the device. The HX711 module has four load cells linked to it
and mounted under the beehive. The load cells put a balanced measurement on the
four corners of the under the beehive.
The schematic diagram illustrated in Fig. 11 is the connections of the beehive
solar panel, battery, and CPU fan components. The solar panel absorbs energy from
the sun and transforms it into electricity. The electricity from the solar panel will
flow straight into the solar charge controller and its rechargeable batteries. The CPU
fan is connected to the ATmega328 and was installed on the top of the hive to prevent
too much heat, and help the water flow inside the beehive.
The schematic diagram illustrated in Fig. 12 is about the bee counter components,
devices, and their connectedness. To achieve the counting function of bees leaving
and coming into the hives, QRE1113 was utilized with two reflectance detectors per

Fig. 10 Schematic diagram of the ATmega328 sensors and ESP8266

82 D. D. Dasig Jr. and J. M. Mendez

Fig. 11 Schematic diagram of solar panel, battery, CPU fan

Fig. 12 Schematic diagram of the QRE1113

gateway; one sensor was used to detect the outgoing bees and the other one to detect
the incoming bees.
A beekeeper or apiarist performs numerous hive management activities to make
sure that apiculture production is high. Before the application of technologies in api-
culture, apiarists were doing manual beekeeping. Figure 13 illustrates the flowchart
of manual beekeeping. The apiculturist monitors the hive’s ambiental conditions.
Hives’ ideal temperatures should be between 60 and 100 °F (15–38 °C) if its too hot
bees do not work because they cluster in unshaded hives. In traditional beekeeping,
an apiarist checks the weather. IF it is too hot, THEN, go to the beehive and cool
it down, check the frames, IF there is enough honey, THEN harvest. Now, IF the
weather is not hot or between 15 and 38 °C, THEN check the frames, IF there is
enough honey, THEN harvest.
In this developed precision apiculture technology, the apiarists can either use
the online application using a web browser or launch a mobile application in a
An IoT and Wireless Sensor Network-Based Technology … 83

Fig. 13 Flowchart of
manual beekeeping

smartphone where the mobile app was installed. Apiarists will be able to select
which hives conditions they would like to check and monitor. Apiarists are also
informed via notification or dashboard as to whether it is good to harvest already
based on the load or weight of the honey from beehive’s frame. The mobile app is
designed with icons representing (temperature, humidity, fan, and weight) that a user
can select from to improve user experience. Figure 14 illustrates the flowchart of the
developed precision apiculture technology.
84 D. D. Dasig Jr. and J. M. Mendez

Fig. 14 Flowchart of the developed precision apiculture technology

4.1 Device Prototype

The two sensor nodes have been incorporated into the roof of a hive where the samples
can effectively gather the hive conditions. This also resulted in effective operation
without interfering with the beekeeper’s activity. The data outlined in the next sub-
sections of this paper was gathered using this prototype to conduct the data gathering
and system evaluation. The data collected was used to evaluate the effectiveness of
classification algorithms and recommended features of the hardware’s future design
(Fig. 15).
Figure 16 illustrates the actual beehives in the bee farm of Bacnotan, La Union.
The actual hive is a modular beehive oriented vertically with frames inside. It has
an entrance for the bees and several boxes inside the enclosure intended for the bee
brood and honey. It is also top capped to protect the beehives from weather and water.
There are three nodes deployed in this location.
Bacnotan is a first-class municipality in the province of La Union where apicul-
ture is one of the livelihoods and viable means for economic development. It has the
La Union Honeybee Center where visitors can have the first-hand experience on the
lifecycle of bees and honeys from bees are manufactured through a live demonstra-
tion.
The power consumption measurements of the nodes were carried out on the wire-
less sensor networks installed for testing the precision apiculture system. There were
three nodes and a local database server in the network. The measurements were car-
ried out with the active node which continually acquires information and sends it to
An IoT and Wireless Sensor Network-Based Technology … 85

Fig. 15 Device prototype

Fig. 16 Actual deployed devices (sensor nodes)

the local server. In the first two cases, a Wi-Fi module was set in receiving mode,
and the acquired data was not transmitted via the node. In the latter case, the nodes
continually transferred the obtained information.
The battery-compensation capacitors work when an electricity supply failure
occurs in one of the components of the system and sensor networks installed in
beehives. The primary energy of the nodes installed within the ring was intended for
75 h to assess the independence of the nodes and the local database server was shut
down. During that time, the sequence of operations is like the one described above,
except that the Wi-Fi module does not transmit data and the samples are stored on the
86 D. D. Dasig Jr. and J. M. Mendez

Fig. 17 Sensor nodes’ power consumption

microSD card. With a rechargeable lithium-ion battery of 3.7 V/2300 mAh, the bat-
tery level of the sensor nodes that were analyzed fell an average of 70%, as shown in
Fig. 17. This time of autonomy is enough to solve the problem since the global server
of the developed system generates an alarm when losing communication. Therefore,
it was not necessary to install a battery with higher capacity.
In this IoT-architectured precision apiculture, a mobile application was developed
to support the users in the IoT platform layer. The mobile application allows the end-
users (beekeeper, apiculture professionals, and researchers) to check and monitor the
measurements of different sensors in the beehive such as temperature, humidity, fan,
and weight of the beehives. The app is also used for hive inspection, management, and
maintenance by using smartphones and tablets PCs. The “Bee Watch” as illustrated
in Fig. 18 was developed using Android Studio and APIs.
Figure 18 shows the user interface (UI) of the Bee Watch mobile application. The
app allows the users to check and monitor the measurements transmitted by the IoT
devices to the IoT platforms including the temperature, humidity, and weight of the
beehive. The fan is also included in the design to allow users to activate fan remotely
using the mobile app when the weather or temperature reading indicates it is higher
than the heat threshold of the application. Similarly, the mobile app will also inform
the users if the fan is working or not with an indicator (Fig. 19).
The illustration presented in Fig. 20 is the user interface for the Bee Watch weight
feature. This feature provides the weight of the bees and the weight of the honey.
This feature informs the user on the rate of change in weight as well as the health of
the colony. It will let the beekeeper know if swarming conditions are likely to happen
An IoT and Wireless Sensor Network-Based Technology … 87

Fig. 18 Beehive mobile

application “Bee Watch”

Fig. 19 Bee Watch user

interface for fan status
88 D. D. Dasig Jr. and J. M. Mendez

Fig. 20 Bee Watch user interface for weight

or have happened already as shown in the current weight of the beehive base on the
time stamp.
In Fig. 21, the average weight of the hives is shown over 32 days as an average
development. It describes that from day 26, the weight of the beehives can be observed
as stable. The changes in weight were associated with honey production depending
on the quality of incoming pollen resources and swarming conditions. Based on the
illustration in Fig. 21, by the end of the period of bloom, the weight of beehive was

Fig. 21 Average weight of the hives

An IoT and Wireless Sensor Network-Based Technology … 89

stabilized which implied that the beekeeper can now remove the honey from the
hives.
When the weather is warmer than the device threshold, the outfitted fan activates
help serve as a cooling system and maintain an appropriate temperature in the hives.
Figure 22 illustrates the average temperature measurements in Celsius recorded dur-
ing the experimentation for 32 days. The information indicates the ambient tempera-
ture outside the beehives and the temperatures in various fields in the beehives: mid-
brood; region in the periphery of the same brood comb with honey/pollen reserves,
and honeycombs separated from brood combs. The figure corresponds to the four
hives’ average daily temperature.
Figure 23 illustrates the environmental moisture and humidity of the beehives in
various regions: mid of brood; the area in the periphery of the same brood comb
with the honey/pollen reserves; and honeycombs separated from the brood combs.

Fig. 22 Average temperature of the hives

Fig. 23 Average relative humidity

90 D. D. Dasig Jr. and J. M. Mendez

The information matched the average moisture content of four hives per day. The
humidity inside of the beehive stayed more stable than beehives’ outside humidity.

5 Conclusion and Recommendation

In this IoT-architectured precision apiculture technology, the researchers aimed at

assisting the beekeepers on apiary inspection, monitoring, management, and main-
tenance. The descriptive-developmental methodology was explored using the proto-
typing techniques to facilitate user involvement in the project development stages.
The IoT architecture and wireless sensor networks (WSNs) were used in the design
and implementation of the system. The three layers of the IoT architecture were
configured based on the results of system analysis and design. IoT architecture lay-
ers were considered in the system including the IoT device, IoT gateway, and IoT
platform layers.
Based on the system deployed in Bacnotan, La Union, Philippines, and the data
gathered during the 32-day experimentation, the researchers concur that IoT archi-
tecture can be complemented with a wireless sensor network approach in the design
of IoT gateway layer.
Also, the system provides real-time surveillance of the defined ambiental param-
eters of a bee colony that has become a popular instrument for both research and
practical beekeeping, as the tiny, precise, and robust sensors make it easy for the
electronic parts to deploy and retain. Hive weight monitoring allows beekeepers to
monitor hive health and can record quality control during pollination. Precision bee-
keeping is still evolving, but multiple weight surveillance systems for bee colonies
are already accessible. The developed and deployed project acquired samples and
dataset from all hives of apiary synchronously, with a real-time hive weight mea-
surement that is an innovation in the monitoring of wireless beehive systems.
Similarly, wireless beehive system allows apiarists to monitor the temperature,
relative humidity, and weight monitoring non-intrusively, hence, it promotes beehives
to become healthy and increase honey’s productivity, and effectively perform the
overall aspects of apiary management using the Bee Watch mobile application.
The embedded system of this project was powered by a solar panel and compen-
sation capacitors in the rechargeable battery. However, redundancy components are
made available should power interruption will deter the transmission of ambiental
data to the servers. Aside from the analytical tools and reports the system provides,
it is also a reliable low-cost IoT-architectured beehive monitoring system based on a
WSN.
The researchers recommend that empirical investigation should be conducted
based on the data gathered compared with other “top of the line apiary technolo-
gies” available in other countries. Future researchers may consider on swarm intelli-
gence to address the decreasing populations of bees and applying machine learning
techniques to improve beehives’ conditions and increase honeybee production and
manufacturing.
An IoT and Wireless Sensor Network-Based Technology … 91

References

1. Patrício-Roberto, G.B., Campos, M.J.O.: Aspects of landscape and pollinators—what is impor-

tant to bee conservation? Diversity 6, 158–175 (2014)
2. Goulson, D., Nicholls, E., Botías, C., Rotheray, E.L.: Bee declines driven by combined stress
from parasites, pesticides, and lack of flowers. Science 347, 1255957 (2015)
3. Matthias, A.B., Hanno, H., Charlotte, K.H., Robin, F.A.M.: Brood temperature, task division
and colony survival in honeybees: a model. Ecol. Model. 221(5), 769–776 (2010)
4. Evans, H.: Remote bee hive monitoring. Arnia [Online]. Available: http://www.arnia.co.uk/.
Accessed on 22 Sept 2014 (2014)
5. Dongyun W., Kai W., Mingcong D.: The application study of intelligent PID algorithm for the
internal combustion engine control system. In: International Conference on Mechatronics and
Automation, China, pp. 923–927 (2010)
6. Arnia Remote Hive Monitoring. Arnia [Online]. Available: http://www.arnia.co.uk/. Accessed
on 22 Sept 2014 (2014)
7. Southwick, E.E.: Metabolic energy of intact honeybee colonies. Comparative. Biochem Physiol
71, 277–281 (1982)
8. Gates, B.N.: The Temperature of the Bee Colony; Bulletin of the U.S. Department of Agricul-
ture; U.S. Department of Agriculture: Washington, DC, USA (1914)
9. Dunham, W.: Hive temperatures for each hour of a day. Ohio J. Sci. 31, 181–188 (1931)
10. Odoux, J.F., Aupinel, P., Gateff, S., Requier, F., Henry, M., Bretagnolle, V.: Ecobee: a tool for
long-term honey bee colony monitoring at the landscape in West European intensive agroe-
cosystems. J. Apic. Res. 53, 57–66 (2014)
11. Kviesis, A., Zacepins, A., Riders, G.: Honey bee colony monitoring with implemented decision
support system. In: Proceedings of the 14th International Scientific Conference Engineering
for Rural Development (ERDev), Jelgava, Latvia, 20–22 May 2015, pp. 446–451 (2015)
12. Woods, E.F.: Means for detecting and indicating the activities of bees and conditions in beehives.
US Patent 2,806,082, 10 Sept 1957
13. Ferrari, S., Silva, M., Guarino, M., Berckmans, D.: Monitoring of swarming sounds in bee
hives for early detection of the swarming period. Comput. Electron. Agric. 64, 72–77 (2008)
14. Chen, C., Yang, E., Jiang, J., Lin, T.: An imaging system for monitoring the in-and-out activity
of honey bees. Comput. Electron. Agric. 89, 100–109 (2012)
15. Zacepins, A., Stalidzans, E.: Architecture of automatized control system for honey bee indoor
wintering process monitoring and control. In: Proceedings of the 13th International Carpathian
Control Conference (ICCC 2012), Podbanske, Slovakia, 28–31 May 2012, pp. 772–775 (2012)
16. Heidinger, I.M.M., Meixner, M.D., Berg, S., Büchler, R.: Observation of the mating behavior
of honey bee (Apis mellifera L.) queens using radio-frequency identification (RFID): factors
influencing the duration and frequency of nuptial flights. Insects 5, 513–527 (2014)
17. Meikle, W.G., Holst, N.: Application of continuous monitoring of honeybee colonies. Api-
dologie 46, 10–22 (2015)
18. Meikle, W., Holst, N., Mercadier, G., Derouané, F., James, R.: Using balances linked to linked
to dataloggers to monitor honey bee colonies. J. Apic. Res. Bee World 45, 39–41 (2006)
19. Bromenshenk, J.J., Henderson, C.B., Seccomb, R.A., Welch, P.M., Debnam, S.E., Firth, D.R.:
Bees as biosensors: chemosensory ability, honey bee monitoring systems, and emergent sensor
technologies derived from the pollinator syndrome. Biosensors 5, 678–711 (2015)
20. Sánchez, V., Gil, S., Flores, J.M., Quiles, F.J., Ortiz, M.A., Luna, J.: Implementation of an
electronic system to monitor the thermoregulatory capacity of honeybee colonies in hives with
open-screened bottom boards. Comput. Electron. Agric. 119, 209–216 (2015)
21. Kviesis, A., Zacepins, A., Durgun, M., Tekin, S.: Application of wireless sensor networks in
precision apiculture. In: Proceedings of the 14th International Scientific Conference Engineer-
ing for Rural Development (ERDev), Jelgava, Latvia, 20–22 May 2015, pp. 440–445 (2015)
22. Zacepins, A., Kviesis, A., Ahrendt, P., Richter, U., Tekin, S., Durgun, M.: Beekeeping in
the future—smart apiary management. In: Proceedings of the 17th International Carpathian
92 D. D. Dasig Jr. and J. M. Mendez

Control Conference (ICCC), TatranskaLomnica, Slovakia, 29 May–1 June 2016, pp. 808–812
(2016)
23. Murphy, E., Popovici, E., Whelan, P., Magno, M.: Development of a heterogeneous wireless
sensor network for instrumentation and analysis of beehives. In: Proceedings of the IEEE
International Instrumentation and Measurement Technology Conference, Pisa, Italy, 11–14
May 2015 (2015)
24. Murphy, E., Magno, M., Whelan, P., Popovici, E.: b+WSN: Smart beehive for agriculture,
environmental, and honey bee health monitoring—preliminary results and analysis. In: Pro-
ceedings of the 10th IEEE Sensors Applications Symposium (SAS), Zadar, Croatia, 13–15
April 2015, pp. 1–6 (2015)
25. Kridi, D.S., Carvalho, C.G.N., Gomes, D.G.: Application of wireless sensor networks for
beehive monitoring and in-hive thermal patterns detection. Comput. Electron. Agric. 127,
221–235 (2016)
26. Jiang, J.-A., Wang, C.-H., Liao, M.-S., Su, Y.-L., Chen, W.-S., Huang, C.-P., Chen, C.-H., Yang,
E.-C., Chuang, C.-L.: A WSN-based automatic monitoring system for the foraging behaviorof
honey bees and environmental factors of beehives. Comput. Electron. Agric. 123, 304–318
(2016)
27. Kviesis, A., Zacepins, A.: System architectures for real-time bee colony temperature monitor-
ing. Procedia Comput. Sci. 43, 86–94 (2015)
28. Libelium Communications. Waspmote Catalogue. Available online: http://www.libelium.com/
products/waspmote. Accessed on 10 Oct 2016
29. Benchoff, B.: Building a digital scale from scratch. Hackaday, 12 June 2013. [Online]. Avail-
able: http://hackaday.com/2013/06/12/building-a-digital-scale-from-scratch/. Accessed on 6
Jan 2015
Adaptive Encryption Technique
for Collaborative Cloud Environments

Y. Harold Robinson, E. Golden Julie and A. Vegi Fernando

Abstract In recent times, the data storage in collaborative cloud environment has
distorted a lot when compared with prior times. Security flattering is one of the
major imperative challenges in database storage and classification. Data stored in
databases are susceptible and thus focusing the safety only on the susceptible data
that minimizes the delays or troubles in the classification. This paper describes an
extremely innovative and novel methodology for securing numeric data in databases.
It presents a realistic solution to the problem where numeric data are transformed into
alphanumeric types and hence there is minimal probability of storing encrypted data
in the existing numeric field. The proposed algorithm allows translucent verification
intensity encryption that does not modify the data field category or the fixed time
span.

Keywords Encryption · Decryption · Database · Security · Authentication

1 Introduction

The most excellent technique of securing the data is to limit the access rights to the
information which can be achieved by the procedure of authentication and autho-
rization [1] in collaborative cloud environment. A consumer should be checked for
authentication before being allowed to access the information and must only be
allowed to attain the operations for which access rights are obtainable [2, 3]. If the

Y. Harold Robinson (B) · A. Vegi Fernando

Department of Computer Science and Engineering, SCAD College of Engineering
and Technology, Cheranmadevi, India
e-mail: [email protected]
A. Vegi Fernando
e-mail: [email protected]
E. Golden Julie
Department of Computer Science and Engineering, Anna University Regional Campus,
Tirunelveli, India
e-mail: [email protected]

© Springer Nature Singapore Pte Ltd. 2020 93

P. K. Pattnaik et al. (eds.), Internet of Things and Analytics
for Agriculture, Volume 2, Studies in Big Data 67,
https://doi.org/10.1007/978-981-15-0663-5_5
94 Y. Harold Robinson et al.

data to be accessed is on a local machine, applying access control is easy, but if data
is accessible from a remote client, user credentials and data needs to be secured over
the network [4]. In such situations, security protocols should be used. In case that
a malicious user somehow breaches the above security provisions and gets access
to data; the only solution is scrambling the data [5]. Security also requires that data
should be kept safe from unauthorized access [6, 7]. And the best line of defence is
physical security (placing the machine to be protected behind physical walls) [8, 9].
Some of the algorithms used for achieving security are Diffie-Hellman [10], DES
[11], TDES [12], IDEA [13] and AES [14]. Usage of these existing cryptographic
techniques is meant to secure the data, which is not suitable for protecting the sen-
sitive data of database [15]. Balaji et al. [16], the result is that most conventional
database systems have leaks that attacker can use to penetrate the database [17]. No
matter what degree of security is put in place, sensitive data in database are still
vulnerable to attack [18]. A remedy, therefore, is to turn to cryptographic means of
storing data [19]. Encrypted data stored in a database can prevent their disclosure to
attackers even if they manage to circumvent the access control mechanism [20, 21].
Thus, cryptographic techniques can ensure excellent security for databases, by reduc-
ing the whole security process down to the protection of only few cryptographic keys
[22]. Nonetheless, the time cost involved in encrypting and decrypting data items can
greatly degrade the performance of the database system [23]. A compromise solu-
tion must be found between performance and security, by encrypting only sensitive
data in the database [24]. The database is usually assumed to be trustworthy [25].
Under this assumption, the goal is to achieve security against external attacks and
probably moreover alongside users trying to acquire data afar their constitutional
rights, for instance, by various forms of statistical conjecture [26]. Encryption is a
healthy established knowledge for defensive responsive information. Nevertheless,
formerly encrypted, information can no longer be easily queried sideways from accu-
rate matches [27]. Encryption can be used to provide an extra level of security [28].
Unfortunately, the use of standard encryption techniques for this purpose results in
degradation in the performance of the database system [29, 30]. The main source of
the problem is that the standard techniques do not preserve order and therefore the
database indices such as B-tree can no longer be used for answering range queries
[31–33].
The database is shared by multiple parties such as internal users, partners, contrac-
tors and other sensitive data stored in database could be a target to attackers [34]. The
attackers are not only external but also from within the organization [35]. Adding the
database encryption, valuable information in database becomes more secure since
the encrypted data ensure the confidentiality of the data [36] in collaborative cloud
environment.
Adaptive Encryption Technique for Collaborative Cloud … 95

2 Proposed Scheme

The proposed research is to secure the numeric data in the databases. It presents a
practical solution to the problem that exists in the standard algorithms where numeric
data is converted to binary or alphanumeric type and hence encrypted data is not
possible to be stored in the existing numeric column. The 3KDEC algorithm allows
record-level encryption that encrypts numeric data and gives numeric encrypted
value; hence eliminating the need to change the data type of column. 3KDEC is
a symmetric key block encipher algorithm. 3KDEC aims to provide a simple and
efficient method for database encryption, overcoming the shortcomings of the prior
database encryption methods with an objective to suggest how to encrypt the entire
contents of the database without changing its structure. It provides a method of
decrypting only the data of interest, wherein queries are not changed because of
the encryption, ensuring that existing applications can use the encrypted database
without the need for any changes in the application software.

2.1 3KDEC Algorithm

3KDEC algorithm is a symmetric key block cipher algorithm. Symmetric key algo-
rithm is one which uses single secret keys for both encryption and decryption. Encryp-
tion/decryption can be considered as an electronic lock where sender puts the message
in a box and locks the box using the shared secret key; receiver unlocks the box with
the same key and takes out the message. The original message is called the plaintext
and the message sent through the channel after encryption is called the cipher text.
To create cipher text from the plaintext, sender uses an encryption algorithm and a
shared secret key. To create plaintext from the cipher text, receiver uses a decryption
algorithm and the same secret key.
The key can be visualized as a set of values/numbers that the cipher as an algorithm
operates. In symmetric key encipherment, the encryption and decryption algorithms
are inverses of each other. If P is the plaintext, C is the cipher text and K is the key, the
encryption algorithm EK(x) creates the cipher text from; the decryption algorithm
DK(x) creates the plaintext from the cipher text. Encryption algorithm EK(x) and
the decryption algorithm DK(x) are inverses of each other and they cancel the effect
of each other when applied one after the other on the same input.
Encryption: C = EK (P)
Decryption: P = DK (C)
That is,

P = DK (C) = DK (EK (x)) = EK (DK (x)) = x

Block ciphers mean a group of plaintext symbols of size m (where m >1) are
encrypted together creating a group of cipher text of the same size.
96 Y. Harold Robinson et al.

2.2 Key Components of 3KDEC Algorithm

3KDEC algorithm uses:

• Numeric data to be encrypted.
• Three keys (stored as three 3 × 3 matrix)
• Substitution Box (commonly referred as S-Box) and an inverse S-box
• Variable number of Rounds (which can be 3, 6 or 9).
Algorithm operates on following steps in sequence for set number of rounds.
• Key Expansion
• Find and Substitute
• Row Shift
• Add set Key.

2.2.1 Key Expansion

In this step, the single key of the three keys is expanded to as many as three keys
summing up the total of nine keys to be used in the maximum nine rounds. Keyl is
expanded as Key10 , Key1l and Key12 and similarly Key20 , Key21 , Key22 , Key30 , Key31
and Key32 . The complexity of key expansion is directly depending on the number of
rounds. In case of three rounds, there will be algorithm complexity of one; while in
case of six rounds, there will be a complexity of two and similarly three in case of
nine rounds. The process of expansion of Key1 into its constituent Key10, Key11 and
Key12 is:
• Key10 is same as Key1
• Key11 is shifting the row 1 zero time, row 2 elements one time and row 3 elements
two times with respect to original keyl
• Key12 is shifting the row 1 one time, row 2 elements two times and row 3 elements
zero time with respect to the original.
With variable number of rounds and varying key expansions complexity of the
algorithm increased.

2.2.2 Find and Substitute

In this step, the byte to be encrypted is found and substituted independently to provide
the confusion effect. There is no fixed mechanism or any mathematical correlation in
the formation of S-box. The entries of S-box can be different in different encryption
processes.
Adaptive Encryption Technique for Collaborative Cloud … 97

So, the simple structure and variable entries of substitution box make the algorithm
eligible to be used as a personal encryption algorithm where different S-Box variants
can be used in encryption process depending on the party we are dealing with and
the varying complexity levels can be set according to our needs.

2.2.3 Rowshift

This transformation step shifts towards the left. The number of shifts depends on
the row number of the matrix. This means the first row of matrix is shifted zero
time, second row of the matrix is shifted one time and the third row is shifted two
times towards the left. During the decryption process, the inverse row shift process
is carried out and the shifting is done towards the right. The number of shifts is same
as the row number.
where,
Row 0: no shift
Row 1: one shift
Row 2: two shifts

2.2.4 Add Setkey

With each round, the matrix is added using XOR operation with the above-expanded
keys. This means during first round, encryption Key10 is used. In the next round
Key11 and then Key12 and so on.

2.2.5 Steps of 3KDEC Algorithm

Step 1: Encryption Step

Consider the sample input as follows (Figs. 1, 2, 3, 4 and 5):
Input M = 1234
Key 1 = 520255270
Key 2 = 104425204
Key 3 = 459105254
98 Y. Harold Robinson et al.

Fig. 1 Architecture of 3KDEC algorithm

Fig. 2 Block diagram for encryption

Fig. 3 Block diagram for decryption

Adaptive Encryption Technique for Collaborative Cloud … 99

Fig. 4 Flow diagram for

encryption

Fig. 5 Flow diagram for

decryption
100 Y. Harold Robinson et al.

Step 2: Key Generation Steps

Input Key 1

5 2 0
2 5 5
2 7 0

The Generated Key 10

5 2 0
2 5 5
2 7 0

The Generated Key 11

5 2 0
5 5 2
0 2 7

The Generated Key 12

2 0 5
5 2 5
2 7 0

Input Key 2

1 0 4
4 2 5
2 0 4

The Generated Key 20

1 0 4
4 2 5
2 0 4
Adaptive Encryption Technique for Collaborative Cloud … 101

The Generated Key 21

1 0 4
2 5 4
4 2 0

The Generated Key 22

0 4 1
5 4 2
2 0 4

Input Key 3

4 5 9
1 0 5
2 5 4

The Generated Key 30

4 5 9
1 0 5
2 5 4

The Generated Key 31

4 5 9
0 5 1
4 2 5
The Generated Key 32
5 9 4
5 1 0
2 5 4
102 Y. Harold Robinson et al.

Step 3: SBOX

0 1 3 3 4 5 6 7 8 9
7 3 6 1 0 9 2 5 4 8

Step 4: Round Operations

The Encryption Steps
The Input data given:

0 0 0
0 0 1
2 3 4

The Round Steps

The Round 1
Substitution Step

7 7 7
7 7 3
6 1 0

Row shift Step

7 7 7
7 3 7
0 6 1

Add set Key Step

The key used is:

7 7 7 5 2 0 2 9 7
7 2 7 + 2 5 5 = 9 8 2
0 6 1 2 7 0 3 3 1

Repeating the same steps up to 9 rounds

Adaptive Encryption Technique for Collaborative Cloud … 103

The Round 9
S-Box Step

0 4 9
1 7 3
0 0 0

Row shift Step

0 4 9
7 3 1
0 0 0

Add Set Key Step:

0 4 9 5 9 4 5 3 3
7 3 1 + 5 1 0 = 2 4 1
0 0 0 2 5 4 2 5 4

The Final Output

5 3 3
2 4 1
2 5 4

For the Input Data M = 1234, the cipher data obtained is: 533241254. The time
taken to perform these nine rounds is (in Milliseconds): 15

Step 5: Decryption Step

Consider the Input cipher text given is
Input CP = 533241254
Key 1 = 520255270
Key 2 = 104425204
Key 3 = 459105254
104 Y. Harold Robinson et al.

Step 6: Key Generation Steps

The Key generation steps in decryption are same as that of encryption because both
the encryption and decryption of this algorithm take only same secrete key.

Step 7: Inverse SBOX

0 1 2 3 4 5 6 7 8 9
4 3 6 1 8 7 2 0 9 5

Step 8: Round Operations

In decryption, the round operation is performed by taking last key of encryption
process as the first key, i.e., in the reverse order. Here, the cipher text is taken as the
input and the original output is obtained by performing the round operations in the
reverse order, i.e., Round 9 of the encryption process is the Round 1 of decryption
process.

Step 9: The Decryption Steps

The Cipher Text given:

5 3 3
2 4 1
2 5 4

This is the input to the next round.

Step 10: The Round Steps

The Round 1
Inverse add set key Step:

5 3 3 5 9 4 0 4 9
2 4 1 + 5 1 0 = 7 3 1
2 5 4 2 5 4 0 0 0

Row shift Step

0 4 9
1 7 3
0 0 0
Adaptive Encryption Technique for Collaborative Cloud … 105

S-Box Step

4 8 5
3 0 1
4 4 4

The above operation is repeated

The Round 9
Inverse add Set Key Step:

7 7 7 5 2 0 7 7 7
7 3 7 - 2 5 5 = 7 3 7
0 6 1 2 7 0 0 6 1

Row shift Step

7 7 7
7 7 3
6 1 0

S-Box Step

0 0 0
0 0 1
2 3 4

At final the Original data is extracted 1234 (Figs. 6, 7, 8 and 9)

106 Y. Harold Robinson et al.

Fig. 6 Unit testing for encryption process

3 Performance Evaluation

As the encryption is from numeric to numeric, one cannot know that the information
is encrypted. It appears content as if the encrypted data is itself the original content.
As the number of keys is more and the key combination increases to 1027 which
makes guessing of keys are very harder. As S-Box has simple structure and variable
entries so the 3KDEC algorithm can be used as Personal Encryption Algorithm where
different encryption processes can be carried out with varying degree complexity
depending on the user requirements. Since 3KDEC algorithm encrypts numeric to
numeric data, encrypted data is possible to be stored in the existing numeric field,
thereby algorithm does not change the data field type and set fixed data length.
A. Analysis of 3KDEC Algorithm
To analyse the computational complexity of a cryptographic algorithm, we need an
encryption algorithm to have low level of complexity and we need an algorithm used
by the cryptanalyst who is trying to break the code to have high level of complexity.
This means we need to perform encryption and decryption in short span of time,
but at the same time, we want the intruder to run her computer infinitely if trying to
break the code.
Adaptive Encryption Technique for Collaborative Cloud … 107

Fig. 7 Unit testing for decryption process

Fig. 8 Integration testing for encryption process

Fig. 9 Integration testing for decryption process

108 Y. Harold Robinson et al.

Complexity of program is based on two types of resources: space complexity

of algorithm which refers to amount of memory needed to store the algorithm and
the data and the time complexity which refers to amount of time needed to run the
algorithm and get the results.
The time complexity of 3KDEC algorithm can be measured (independent of the
computer on which it is running) by defining the bit-operation complexity which
counts the number of bit operations the computer needs to perform to create the
output from the input.
Bit operation can be defined as the time required for the computer to add, subtract,
multiply or divide two single bits or to perform one single bit shift. For example:
bit-operation complexity of function that adds two integers each having, d decimal
digits can be calculated as: d log2 10.
B. Security of 3KDEC Algorithm
The main strength of 3KDEC algorithm is that it converts the numeric data input to
numeric data output which makes it harder to see and guess whether the input is even
encrypted or not.
(1) Brute-force attack
Also known as exhaustive key search method that tries to use all the possible keys. In
3KDEC algorithm, the three keys are expanded based on the number of rounds; there
will be 1027 different key combinations. Lack of weak keys is another advantage of
3KDEC algorithm.
(2) Statistical attacks
The strong confusion and diffusion provided by the substitution box lookup and row
shift transformations remove any frequency pattern in the plaintext.
(3) Differential and linear attack
The algorithm was designed to be resilient to linear and differential attacks. Differen-
tial cryptanalysis/chosen plaintext attack needs the analysis of encryption algorithm
to collect information about plaintext–cipher text relationships; with the goal of
finding the cipher key (Figs. 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21 and 22).
The probabilistic relationship can be created from the information about the S-box
input/output table. In case of 3KDEC algorithm, the S-box values are not fixed and
do not depend on any algebraic structure. The sender may use different S-box values
each time attempting to encrypt the input. So, the attacker has to even guess what
were the values used for the S-Box in encrypting that data.
Linear Cryptanalysis/Known-Plaintext Attack where the attacker needs to have
information about plaintext/ciphertext pairs in addition to the intercepted cipher
text that needs to be broken along with the assumption that key has not changed.
Relationship between previous pair is used to analyse the current cipher text. But the
strong point in the algorithm is that every time the three keys as well as S-box entries
can be changed which makes it difficult to find the appropriate relationships among
the plaintext and the cipher text pair.
Adaptive Encryption Technique for Collaborative Cloud … 109

Fig. 10 Encryption

Fig. 11 Encrypted data

110 Y. Harold Robinson et al.

Fig. 12 Key generation

Fig. 13 Rounds
Adaptive Encryption Technique for Collaborative Cloud … 111

Fig. 14 S-Box

Fig. 15 Decryption
112 Y. Harold Robinson et al.

Fig. 16 Decrypted data

Fig. 17 Rounds
Adaptive Encryption Technique for Collaborative Cloud … 113

Fig. 18 Inverse S-Box

Fig. 19 Original amount before encryption

114 Y. Harold Robinson et al.

Fig. 20 Original amount after encryption

Fig. 21 Original amount before decryption

Adaptive Encryption Technique for Collaborative Cloud … 115

Fig. 22 Original amount after decryption

4 Conclusion

Understanding the need to secure your data is the first step towards securing it. In
today’s age, every detail—personal to corporate secrets—is present in form of data.
For computers and networks which store and transfer this data, it is just numbers. It
is for us to realize that the damage caused to this data may fall into the hands of an
unscrupulous person. Whether the data is on your laptop, desktop or on an organi-
zation’s storage network, it must be secured and should not come in the hand of an
unauthorized entity. Proper access control mechanism should be enforced for secur-
ing the data. While in motion, data should be well protected. It is advisable to encrypt
the data before putting it on a network even if it passes through a secure channel.
The algorithm can be implemented for securing any corporate-related accounting
information to data of personals use. This algorithm does not take into consideration
decimal point numeric data which opens up another area of research and improve-
ments. In 3KDEC algorithm, the numeric data is converted to numeric so that the
data field length and the data field type will not change. In future, this work can be
modified and used in mail services and wireless. The S-box can be made dynamic
so that it will be more secure. It is focused only on numeric values, but we can use
this with slight modification for floating point numbers.
116 Y. Harold Robinson et al.

References

1. Maurer, U.: The role of cryptography in database security. In: ACM SIGMOD (2004)
2. Agrawal, R., Kiernan, J., Srikanth, R., Xu, Y.: Order preserving encryption for numeric data.
IBM Almandan Research Center
3. Harold Robinson, Y., Rajaram, M.: Trustworthy link failure recovery algorithm for highly
dynamic mobile adhoc networks. World Acad. Sci. Eng. Technol. Int. J. Electr. Comput. Energ.
Electron. Commun. Eng. 9(2), 233–236 (2015)
4. Arshad, N.H.: The design and implementation of database encryption (2007)
5. Yang, Z., Sesay, S., Chen, J., Xu, D.: A secure database encryption scheme (2004)
6. Robinson, Y.H., Rajaram, M.: A memory aided broadcast mechanism with fuzzy classification
on a device-to-device mobile Ad Hoc network. Wirel. Pers. Commun. 1–23 (2016). https://doi.
org/10.1007/s11277-016-3213-0
7. Perkins, C.E., Royer, E.M.: Ad-Hoc On-Demand distance vector routing. In: Proceed-
ings WMCSA’99 Second IEEE Workshop Mobile Computing Systems and Applications,
pp. 90–100 (1999)
8. Setton, E., Yoo, T., Zhu, X., Goldsmith, A., Girod, B.: Crosslayer design of ad hoc networks
for real-time video streaming. IEEE Wirel. Comm. 12(4), 59–65 (2005)
9. Robinson, Y.H., Rajaram, M.: Energy-aware multipath routing scheme based on particle swarm
optimization in mobile ad hoc networks. Sci. World J. 1–9 (2015). https://doi.org/10.1155/2015/
284276
10. Stallings, W.: Cryptography and network security: principles and practice, 2nd edn
11. Lin, C.R., Gerla, M.: Asynchronous multimedia multihop wireless networks. Proc. IEEE Info-
com 1, 118–125 (1997)
12. Biswas, S., Morris, R.: Opportunistic routing in multi-hop wireless networks. ACM SIGCOMM
Comput. Commun. Rev. 34(1), 69–74 (2004)
13. Rozner, E., Seshadri, J., Mehta, Y., Qiu, L.: Soar: simple opportunistic adaptive routing protocol
for wireless mesh networks. MobileComput. IEEE Trans. on 8(12), 1622–1635 (2009)
14. Kaur, K., Dhindsa, K.S., Singh, G.: Numeric to numeric encryption of databases: using 3Kdec
algorithm. In: IEEE International Conference IACC 2009, pp. 1501–1505 (2009)
15. Harold Robinson, Y., Rajaram, M.: Establishing pairwise keys using key predistribution
schemes for sensor networks. World Acad. Sci. Eng. Technol. Int. J. Comput. Electr. Autom.
Control Info. Eng. 9(2), 608–612 (2015)
16. Balaji, S., Robinson, Y.H., Rajaram, M.: SCSBE: secured cluster and sleep based energy-
efficient sensory data collection with mobile sinks. Circ. Syst. 7, 1992–2001 (2016). https://
doi.org/10.4236/cs.2016.78173
17. Chang, C., Chan, C.W.: A database record encryption scheme using RSA public key cryptosys-
tem and its master keys. In: The International Conference on Computer networks and Mobile
Computing (ICCNMC), pp. 312–315 (2013)
18. Robinson, Y.H., Balaji, S., Rajaram, M.: ECBK: enhanced cluster based key management
scheme for achieving quality of service. Circ. Syst. 7, 2014–2024 (2016). https://doi.org/10.
4236/cs.2016.78175
19. Li, Y., Chen, W., Zhang, Z.L.: Optimal forwarder list selection in opportunistic routing. In:
2009 IEEE 6th International Conference on Mobile Adhoc and Sensor Systems MASS’09,
pp. 670–675. IEEE (2009)
20. Füßler, H., Widmer, J., Käsemann, M., Mauve, M., Hartenstein, H.: Contention-based forward-
ing for mobile ad hoc networks. Ad Hoc Netw. 1(4), 351–369 (2003)
21. Yang, S., Zhong, F., Yeo, C.K., Lee, B.S., Boleng, J.: Position based opportunistic routing for
robust data delivery in MANETs. In: GLOBECOM 2009-2009 IEEE Global Telecommunica-
tions Conference, pp 1–6. IEEE (2009)
22. Harold Robinson, Y., Rajaram, M., Golden Julie, E., Balaji, S.: Dominating set algorithm and
trust evaluation scheme for secured cluster formation and data transferring. World Acad. Sci.
Eng. Technol. Int. J. Comput. Electr. Autom. Control Info. Eng. 10, 388–393 (2016)
Adaptive Encryption Technique for Collaborative Cloud … 117

23. Darehshoorzadeh, A., Cerda-Alabern, L. Distance progress based opportunistic routing for
wireless mesh networks. In: 2012 8th International Wireless Communications and Mobile
Computing Conference (IWCMC), pp. 179–184. IEEE (2012)
24. Gupta, S., Dhurandher, S.K., Woungang, I., Kumar, A., Obaidat, M.S.: Trust-based security
protocol against blackhole attacks in opportunistic networks. In: 2013 IEEE 9th International
Conference on Wireless and Mobile Computing, Networking and Communications (WiMob),
pp. 724–729. IEEE (2013)
25. Harold Robinson, Y., Rajaram, M., Golden Julie, E., Balaji, S.: Tree based data fusion clustering
routing algorithm for illimitable network administration in wireless sensor network. World
Acad. Sci. Eng. Technol. Int. J. Comput. Electr. Autom. Control Inf. Eng. 10(6), 1123–1130
(2016)
26. Mahmoud, M.M., Lin, X., Shen, X., et al.: Secure and reliable routing protocols for hetero-
geneous multihop wireless networks. IEEE Trans. Parallel Distrib. Syst. 26(4), 1140–1153
(2015)
27. Robinson, Y.H., Rajaram, M., Julie, E.G., Balaji, S.: TBOR: tree based opportunistic routing
for mobile Ad Hoc networks. world Acad. Sci. Eng. Technol. Int. J. Comput. Electr. Autom.
Control Inf. Eng. 10(6), 1115–1122 (2016)
28. Cho, J.-H., Swami, A., Chen, R.: A survey on trust management for mobile ad hoc networks.
Commun. Surv. Tutorials IEEE 13(4), 562–583 (2011)
29. Yu, H., Shen, Z., Miao, C., Leung, C., Niyato, D.: A survey of trust and reputation management
systems in wireless communications. Proc. IEEE 98(10), 1755–1772 (2010)
30. Michiardi, P., Molva, R.: Core: a collaborative reputation mechanism to enforce node coop-
eration in mobile ad hoc networks. In: Advanced Communications and Multimedia Security,
pp. 107–121. Springer (2002)
31. Ramalakshmi, S., Robinson, Y.H.: ATMPH: Attaining optimal throughput capacity of MANET
with power control in heterogeneous network. Program. Device Circ. Syst. 6(4), 111–115
(2014)
32. Ayyasamy, A., Venkatachalapathy, K.: Context aware adaptive service based dynamic channel
allocation approach for providing an optimal QoS over MANET. Int. J. Eng. Technol. (IJET).
6(3), 1465–1479 (2014). ISSN: 0975-4024
33. Harold Robinson, Y., Rajaram, M.: A novel approach to allocate channels dynamically in
wireless mesh networks. World Acad. Sci. Eng. Technol. Int. J. Comput. Electr. Autom. Control
Inf. Eng. 8(10), 1865–1868 (2014)
34. Han, G., Jiang, J., Shu, L., Niu, J., Chao, H.-C.: Management and applications of trust in
wireless sensor networks: a survey. J. Comput. Syst. Sci. 80(3), 602–617 (2014)
35. Robinson, Y.H., Rajeswari, S.R.: Energy-based dynamic encryption for wireless sensor net-
works. Wirel. Commun. 3(9), 661–663 (2011)
36. Buchegger, S., Le Boudec, J.Y.: Performance analysis of the confidant protocol. In: Proceed-
ings of the 3rd ACM international symposium on Mobile ad hoc networking & computing,
pp. 226–236. ACM (2002)
Role of Internet of Things and IT
in Health Care

Rohit Sharma and Rajendra Prasad Mahapatra

Abstract The aim of this chapter is to introduce; however, IoT is remodeling health
care and also the role of Information Technology in health care (Lee et al. in 2015 EEE
International Conference on Industrial Engineering and Engineering Management
(IEEM). IEEE, pp. 1855–1859, 2015 [1]). The IoT applications are obscurity essential
in remodeling lives of individuals than in health care. Internet of Things indicates to
physical devices, like a measuring system, weight scale and patients very important
watching devices (glucose, force per unit area, vital sign, and activity watching, etc.)
connect with the web and transforms data from the physical to the digital world. In
step with Gartner, there will be nearly twenty-six billion devices on the web of things
by 2020. This data provides insight into the supplements and health actions to boost
the health, while not the hindrance of the daily routine. During this chapter, we will
explore in depth the importance of the Internet of Things devices in health care and
also the role of Information Technology in managing the massive volume of high-
security patient’s information of medical. Expected key exclude from this chapter is
current challenges, trends, case study, and a project management expertise for a true
world. This chapter proposes few IoT applications in rural health care and ways to
boost primary health desires of the developing nations (Rizwan et al. in International
Conference on IEEE Emerging Technological Trends (ICETT). pp. 1–7, 2016 [2].

Keywords Health care · Cloud · Internet of Things · Data analytics

1 Introduction

Internet of Things (IoT) might be a promising model to incorporate numerous corre-

spondence arrangements and innovations. The Internet of Things characterizes wher-
ever each physical articles to be associated with any place and whenever abuse net

R. Sharma (B) · R. P. Mahapatra

SRM Institute of Science and Technology, Ghaziabad, India
e-mail: [email protected]
R. P. Mahapatra
e-mail: [email protected]
© Springer Nature Singapore Pte Ltd. 2020 119
P. K. Pattnaik et al. (eds.), Internet of Things and Analytics
for Agriculture, Volume 2, Studies in Big Data 67,
https://doi.org/10.1007/978-981-15-0663-5_6
120 R. Sharma and R. P. Mahapatra

and be prepared to build up themselves to various gadgets. Lately, there are essential
advances inside the field of IoT. It will deliver modern upheaval in an exception-
ally sizable measure of utilizations like tending frameworks, sensible living, setting
recognition, and sensible homes. These square measure a few IoT applications, and
among those tending frameworks thought of one among the chief crucial difficulties
inside the blessing scene. Well-being and eudemonia are one among the preeminent
promising application territories of IoT innovation [3].
Distinctive utilize cases encapsulate rising a patient’s consistence to treatment and
prescription in healing facilities, centers, and diverse consideration offices. Medicinal
gadgets like individual home-utilize analytic gadgets or low-end symptomatic and
imaging gadgets that square measure used by portable specialists square measure
one among the key innovation parts. A modernized tending framework should give
higher tending administrations to people whenever and from wherever in amicable
way. Well-being is one among the world difficulties for humankind. Solid individuals
will downsize weight on the as of now overwhelmed healing facilities and centers.
Hence to remain individuals sound, a great in vogue tending framework is pivotal.
IoT has a few advantages in tending, through the usage of sensors, locators and
instrumentation, and so forth. These empower the ID and patient after online, the
areas of the specialist, and monitor the report of the patient [4].
The term IoT was first authored by Kevin Ashton. The RFID group characterizes
web of things in light of the fact that the overall system of interconnected ques-
tions unambiguously accessible upheld ordinary correspondence conventions [5]. It
consolidates antiquated fields like administration systems, wireless gadget networks,
embedded systems, and automation to encourage device-to-device (D2D) correspon-
dence through the net. The musing was first utilized at the Auto-Id focus at MIT.
Radio-frequency identification (RFID) was viewed as a pre-imperative for executing
frameworks that were delegated IoTs.
The tending advancement is predicated on rising the human life quality and remote
sensors square measure put on the body of the patients to watch their well-being
state. Continuous patient recognition frameworks square measure upheld for the
point of early identification of clinical crisis. They are acclimated endlessly track
the imperative indications of the patient, for instance: pulse oximetry, respiration
rate, temperature, heart rate, rate inconstancy, circulatory strain, skin temperature
and electrical wonder, blood liquor focus, and so forth. The sensors square measure
feasible to catch changed data with respect to the patients, in regards to their physical,
mental, and action state. Distinguished data is delivered to a restorative focal unit
conveyed by any authorized individual [3, 6].
The restorative focal unit investigates the gotten data and demonstrates the out-
comes to medicinal experts to get a handle on patient’s well-being state. Remote
body sensors can give the probability of nonstop recognition, early identification,
and quick association in crisis conditions. It causes tending experts to stay away
from move from one place to an alternate, and conjointly, it serves to patients to
maintain a strategic distance from substitute lines. Sensors can encourage recog-
nition for the senior citizens and individuals all square measure having perpetual
Role of Internet of Things and IT in Health Care 121

illnesses like polygenic issue, malignant growth, asthma, memory decay, and man-
ifestation heart condition, and so forth, they will maintain a strategic distance from
movement to clinics from one place to various and time to see their well-being state.
At the moment, the term IoT is frequently utilized as an offering move. This move
is named “vaporware”; promoting an item that is not offered in any case or does not
exist. Firms are utilizing this move to establish the connection. IoT innovation are
now offered to have an expedient appropriation once they have completed their
advancement.
As every one of them square measure in unsteady condition which will create to
crisis at any minute, for this condition, the sensors can caution the medicinal laborers
remotely in order to intervene them. The patients can even utilize the remote sensors
themselves for following their well-being every day and even remotely through the
net [7].

2 About (IoT) Internet of Things

The term IoT was first authored by Kevin Ashton. The RFID group characterizes
web of things in light of the fact that the overall system of interconnected questions
unambiguously accessible upheld ordinary correspondence conventions [5]. It con-
solidates antiquated fields like administration systems, wireless gadget networks,
embedded systems, and automation to encourage D2D correspondence through the
net. The musing was first utilized at the Auto-Id focus at MIT. RFID was viewed
as a pre-imperative for executing frameworks that were delegated IoTs. Today, it’s
applications for every close to home also as business clients.
From the edge of individual clients, social insurance, e-learning, domotics are the
key fields while from business clients’ point of view, computerization, providing,
and mechanical creating are the vital spaces [8].
A. Inescapable Computing
The headways and union of miniaturized scale electromechanical frameworks
(MEMS) innovation, remote interchanges, and computerized physical science have
LED to the occasion of smaller than usual gadgets. These gadgets have the adapt-
ability to detect, process, and impart remotely [9].
These smaller than expected gadgets alluded as hubs interconnect to make a
remote gadget systems (WSN).
B. Applications
In the Internet of Things (IoT), its capability to assemble, record and break down new
data streams faster and a ton of precisely by making gadgets accumulate and share
information specifically with each other and furthermore the cloud. The Internet of
Things can affect numerous application areas. The applications might be grouped
upheld the sort of system openness, inclusion, scale, heterogeneousness, repeatabil-
ity, client association, and effect [10].
122 R. Sharma and R. P. Mahapatra

Fig. 1 Figure showing the Internet of Things application areas

These applications might be ordered into four areas: Personal and help, venture,
utilities, and versatile. This can be portrayed in Fig. 1, that speaks to personal and
help IoT at the measure of an individual or home, endeavor IoT at the extent of
a network, utility IoT at a national or local scale and portable IoT that is some of
the time unfurl crosswise over elective spaces chiefly because of the character of
property and scale [11].

3 The IoT in Health Care

The point of this section is to a great degree general and does not state an incred-
ible arrangement concerning its motivation. In this manner first, some foundation
information is imperative to offer a motivation why this subject brought about an
extremely postulation [12].
The Internet of Things might be an exceedingly respected point nowadays and is
generally seen as each sort of articles that impart misuse the net. For some clients,
this can be detectable once looking for transmission gadgets. Its intense at present
to search out a television or radio that is not associated with the net. This pattern can
proceed inside the years to return by interfacing extra questions. This may prompt
huge systems and increase the monstrous data slant that began inside a similar way
on the grounds that the snare of Things [13].
To comprehend this thought, the essential advance is to appear at anyway it works
while not delving as well such a great amount into the specialized points of interest.
Next, a definition is required to have a common comprehension. This can be tended
to inside the first a piece of the writing survey.
A fascinating point is to imagine anyway firms can utilize this mechanical idea
to make worth for his or her clients. An accommodating method for this reason
for existing is gazing at the plans of action of those organizations. This can be the
Role of Internet of Things and IT in Health Care 123

second a piece of the writing survey, plans of action for firms that utilization web of
things applications in their valuable suggestion. The instructional exercise writing
regarding this matter dressed to be rare since it is such an ongoing idea [14].
Seeing plans of action of firms that utilization the net of things might be an
expansive subject for a section of a year. Constraining this subject to a specific
division would slash this down and gives the opportunity to appear at it in extra
detail. Inside the writing, there is typically an outline of different segments with the
most essential prospects for the net of things. Human services are that the segment that
is never-endingly made reference to and have some horribly encouraging chances.
An outline of this division and trademark of these open doors is that the last piece of
the writing audit.
The scholastic writing on plans of action for the net of things is not different, the
writing on web of things plans of action in medicinal services is even however that.
This makes a propelling investigation chance that might be tended to amid this postu-
lation. Similarly, important is perceiving the troubles, difficulties, and confinements
for this thought amid this setting [15].

4 IoT Devices Enabling to Work Together

Guidelines speak to relate degree inalienable test for any surroundings that during
which an outsized scope of confounded gadgets got the chance to speak with each
other which is unequivocally the situation for the IoT in tending. One expert has
depicted the “more noteworthy institutionalization of interchanges convention” as
essential to propelling the selection of the IoT [16].
Luckily, models associations are working at present to make pointers for remote
interchanges between recognition gadgets and with consideration providers. The
Continua Health Alliance (CHA), of that freescale could be a part, could be an alliance
of tending and innovation firms that were bolstered in 2006 to find out pointers for
down to earth individual well-being arrangements. The association has just printed
a gathering of determinations to help ensure capacity. Inside the future, associations
that get a Continua-guaranteed gadget can have the genuine feelings of serenity that
it will interface with various confirmed gadgets in IoT-driven applications.
Continua’s gadget norms are a piece of a greater principles surroundings that
highlights data innovation models setup by the arrangement for standardization (ISO)
and building gauges set by the Institute of Electrical and physical science Engineers
(IEEE® ) [17].
In remote innovation, IEEE principles for LANs plot ZigBee and Wi-Fi systems.
Models for PANs typify Bluetooth and BLE, which are the IEEE norms identified
with the body space organize (BAN). Gauges for cell systems typify CDMA and
UMTS/GSM. Restrictive remote systems still play one thing of an assignment in
tending conditions regularly and IoT applications particularly, anyway that job gives
off an impression of being contracting in light of the fact that the business keeps on
moving toward models-based structures.
124 R. Sharma and R. P. Mahapatra

5 The Time Is Now: IoT in Health Care

The since quite a while ago anticipated Internet of Things upheaval in social insur-
ance is as of now in the air, in light of the fact that the models amid this part give
subtle elements. Also, those territory units just the tip of the famous chunk of ice,
as new utilize cases still develop to deal with the squeezing need for sensible, open
consideration.
Freescale is worked up to be a zone of this upheaval by giving end-to-end pro-
cess and property answers for IoT-driven medicinal services arrangements, working
toward setting up guidelines for these arrangements and quick development for asso-
ciations expecting to comprehend the upsides of the IoT in human services.

5.1 Focused Applications

At the moment, the new businesses region unit concentrating their valuable sugges-
tion on a chosen downside in medicinal services:
• Up the lifetime of patients with a perpetual condition.
• Simplify/computerize body assignments of the consideration representatives.
Johnson et al. proclaimed accuracy is that the most critical component of a valu-
able suggestion and along these lines the organization in medicinal services was
demonstrated commonly of the three issues, wherever IoT will have the most vital
effect inside the writing survey. Regardless these two classes of uses region unit
sketchy low-hanging natural products, an administration or item which will be sold-
out essentially. The new businesses region unit mindful their present technique has
Associate in Nursing termination date and is endeavoring to grow advance PC code
administrations to upgrade their contributions. The issue is that these further admin-
istrations need self-learning calculations [18].
A possible future situation is that partnerships can at present develop their admin-
istrations and grow their specific value recommendation to various issues in human
services. This can set aside your opportunity to attempt and do this with progress
because of an enough client base will be important to get enough information and
extra enhance the administration. Assuming, in any case, partnerships will achieve
this progression, the $64,000 possibilities of IoT can surface. What is more, inside
the future, the blending of those applications or the persistent expansion of their
administration may prompt Associate in Nursing real IoT application that coordi-
nates the definition from the writing survey. As clarified inside the audit, these extra
propelled value recommendations would require rectify plans of action in help.
Role of Internet of Things and IT in Health Care 125

5.2 Health Care and Government Regulations in Belgium

The announcements from (2008) Johnson and van Limburg concerning the social
insurance area were to a great extent affirmed for the European nation. It is an
elegant market and innovation is implemented gradually. The government laws zone
unit is a crucial issue here. The pay frameworks and venture determinations region
unit associated with the government and backup plans.
In the writing survey, security was referred to as a noteworthy issue since all the
associated articles may as of now be a conceivable danger for unapproved get to. Be
that as it may, at the moment this is not the condition for new businesses. The new
companies accept comparative measures as various firms in Associate in Nursing
IT-condition. The well-being of data in all probability is not a tangle because of
there is not huge amounts of self-governing connection between articles inside their
application and without a doubt not on a scale stunning in the writing [19].
The equivalent is valid for the client’s security. Inside the expressions of nobility
Van Pee, one in every one of the interviewees: “Protection might be a test anyway
not a genuine issue.” In might 2018, the last data Protection Regulation can wind
up viable. This can be a fresh out of the box new European direction concerning
the use of private data. These laws obviously state what firms will and can not do
with the data from their clients. None of the new companies supply data to various
associations. The data is not important for various gatherings after they should change
in accordance with these protection laws. So now, data evidence is anything but an
accommodating asset of income for them [18].

6 The Internet of Things in Health Care (Flanders)

The talked with new companies do not have to relate degree application that coor-
dinates my meaning of IoT. When exploitation the meanings of associated thoughts
from the writing survey, the arrangement of Partheas is frequently sorted as close
insight. Exploitation RFID labels and screens, the doctor’s facility turns into a detect-
ing setting with partner degree algorithmic guideline that figures the normal holding
uptime. This technique part computerizes patients stream inside the healing facility
with some predefined abilities (shut condition). The machine of Ectosense might be
a demonstrative decent gadget associated with a cell phone that sends data to a cloud
stage wherever it is broke down. This fulfills a tremendous piece of the IoT defini-
tion. Be that as it may, the data originates from one great question, there is no mix
{of information of knowledge of information} and in this manner, the information
streams primarily one way [18].
In view of the interviewee’s remarks, the impression is giving partner degree IoT
application that coordinates my definition is not offered at the moment in human
services in the European nation. The new businesses that were fixed for meetings
had the preeminent likenesses with the anticipated definition from the writing survey.
126 R. Sharma and R. P. Mahapatra

All the equivalent, it is difficult to demonstrate this supposition since this decision
was bolstered the information of the new companies’ sites. The investigate for new
companies was bolstered well-being school and not the only one on IoT since it is
feasible that organizations do not utilize this term to blessing their application, even
once it is partner degree IoT application steady with the anticipated definition. Be
that as it may, the example of meetings is confined and in this manner, the pursuit
was escalated anyway not careful since a few new companies would not out in the
open demonstrate their application if it is not completed in any case [17].
Interviewees affirmed that square measure a few times to use IoT applications in
medicinal services anyway there square measure still some mechanical confinements.
In any case, they trust extra propelled applications can progress toward becoming
offered inside the future, anyway, it will take longer than what is commonly expected.

7 Roadmap for IoT

SRI Consulting Business Intelligence sees the occasion of IoT in waves according
to Fig. 2. The first wave began with the work of RFID labels to encourage directing,
order and misfortune obstacle, as offer chain assistants. Inside the second wave,
the vertical-advertise applications e.g. police examination, security, human services,
transport, sustenance offer, and archive administration are region unit created. The
third wave, which we tend to territory unit by and by encountering, is concerning
ubiquitous situating, e.g., finding people and ordinary items. Ensuing wave, that is
anticipated to develop in a couple of decade, is the production of a physical-world net,
e.g., tele-task and tele-nearness, capacity to watch, and administration inaccessible

Fig. 2 Roadmap for IoT (SRI Consulting Business Intelligence)

Role of Internet of Things and IT in Health Care 127

items. Amid this piece of IoT advancement, it is normal that everyone the physical
items are consistently incorporated on the whole three dreams [18].
These square measure some of the long run spaces in consideration zone recorded
underneath, that is fundamentally determined by the adaptability to watch and admin-
istration far off items when.
a. Ingestible identifier
The ingestible identifier is that the finder based for the most part innovation wrapped
as a pill. It is made totally of fixings found in nourishment and initiated upon uti-
lization. This is regularly taken on board the recommended drugs, catching the exact
time of utilization for interest the consistence to medicine. There are no any battery
and no reception apparatus. When the pill with the ingestible identifier is wrapped,
a mid-region liquid compound process gives required power supply and enacts the
finder. This finder stays inside the patient’s mid-region and gives timeframe infor-
mation with respect to anyway the patient reacts to drugs. The patients do not need to
be constrained to persist x-beams, blood tests, or biopsies for his or her specialists to
see regardless of whether a medication is working. This might be especially helpful
for patients WHO regularly take drugs for incessant gadgets [19].
After initiation, this locator produced and transmits the particular assortment.
The fix, expendable and body-worn, catches and transfers your body’s physiological
reactions and practices. It also gets the particular information from the ingestible
locator, identifies fundamental sign, action, and rest, and sends information to your
cell phone. At that point, these data sources square measure passed on to the focal
entranceway to the anchored data server. This information is any dispersed to required
consideration providers as appeared in Fig. 3 [15].
b. Digital drugs
This is relating degree development of the ingestible gadget; here, the medicine itself
goes advanced, instead of an additional pill for trailing consistence. Computerized

Fig. 3 Ingestible sensor network

128 R. Sharma and R. P. Mahapatra

medicines are a proportionate physician endorsed drugs devoured nowadays, with

one little change: each pill will contain a little gadget that may convey, by means of the
advanced well-being criticism framework, imperative information in regards to the
prescription taking practices and thus the responsiveness of the body. These ingestible
sensors among meds are initiated exclusively on contact with belly corrosive. This
aides in giving higher truth and higher harshness of trailing the strength of patients.
This is succeeding phase of constant, wherever divisions can a little while later
endorse ingestible sensors, which can remotely report back on the body’s critical signs
on an every minute of everyday premise. Computerized medicines are underneath
improvement and do not appear to be regardless government organization cleared.
Directly clinical preliminaries are led in these remedial territories [5]:

• Heart Failure
• Nervous System
• Transplant

c. Customized drugs

As of now, the supply is predicated on populace insights. Patients are isolated into
groups sketched out from various perspectives that anyway normally by comparable
side effects or by the consequences of fundamental work environment tests (like
cholesterin levels).
d. Prophetical investigation
The utilization of prophetical investigation in medicinal services can have the advan-
tage of the converging of different data archives that has the patterns of sustenance
propensities. A ton of we as a whole know in regards to an individual or populace,
that is, the bigger the picture, a considerable measure of exact the forecasts will
be. These models are regularly bespoken bolstered the data focuses, to a specific
patient or bunch of patients that at last outcomes in a considerable measure of exact
and compelling medicines that are ensured to enhance the general effectuality of the
human services framework while at a proportionate time diminishing costs [15].
In the near future, the well-being buddy versatile application is additionally ready
to caution you days preceding that you essentially are taking care of show at least
a bit of kindness disappointment by detecting beyond any doubt genomic signals
current in your circulatory system and making you your heart specialist or to the ER.

8 Conclusion

As made reference to amid this part, all the physical articles can work flawlessly with
machine-to-machine and human-to-machine interfaces. This level of interconnection
might be an aid for the social insurance. These variables along the edge of the genomic
inputs will fabricate it possible to anticipate the well-being patterns and sensitivities
of the individual; in this manner, the innovation will give bespoken proposals on
Role of Internet of Things and IT in Health Care 129

fitting physical exercises, eats less carbs, and so forth. These portable specialist
buddy applications do not appear to be intended to be the swap for aptitude of the
specialists. Proposition of IoT medicinal services gadgets for the creating scene
are remote counseling, hand-held symptomatic gadgets for identification pestilence
perishes like protozoal contamination and Indian cholera. These gadgets will have
the manner in which more extensive achieve contrasted with the standard medicinal
guide human services. It is apparent that IoT can encourage new plans of action
and new supply models inside the future for each creating and created universes,
regardless of the difficulties really young looking right now.

References

1. Lee, C., Yeung, C., Cheng, M.: Research on iot based cyber physical system for industrial big
data analytics. In: 2015 EEE International Conference on Industrial Engineering and Engineer-
ing Management (IEEM), pp. 1855–1859. IEEE (2015)
2. Rizwan, P., Suresh, K., Babu, M.R.: Real-time smart traffic management system for smart
cities by using internet of things and big data. In: International Conference on IEEE Emerging
Technological Trends (ICETT), pp. 1–7 (2016)
3. Zhang, Q., Zhang, X., Zhang, Q., Shi, W., Zhong, H.: Firework: Big data sharing and processing
in collaborative edge environment. In: 2016 Fourth IEEE Workshop on IEEE Hot Topics in
Web Systems and Technologies (HotWeb), pp. 20–25 (2016)
4. Rathore, M.M., Ahmad, A., Paul, A.: Iot-based smart city development using big data analytical
approach. In: IEEE International Conference on Automatica (ICA-ACCA), pp. 1–8. IEEE
(2016)
5. Minch, R.P.: Location privacy in the era of the internet of things and big data analytics. In:
2015 48th Hawaii International Conference on System Sciences (HICSS), pp. 1521–1530.
IEEE (2015)
6. Arora, D., Li, K.F., Loffler, A.: Big data analytics for classification of network enabled devices.
In: 2016 30th International Conference on Advanced Information Networking and Applications
Workshops (WAINA), pp. 708–713. IEEE (2016)
7. Yen, I.-L., Zhou, G., Zhu, W., Bastani, F., Hwang, S.-Y.: A smart physical world based on
service technologies, big data, and game-based crowd sourcing. In: 2015 IEEE International
Conference on Web Services (ICWS), pp. 765–772. IEEE (2015)
8. Mishra, N., Lin, C.C., Chang, H.T.: A cognitive adopted framework for IoT big-data manage-
ment and knowledge discovery prospective. Int. J. Distrib. Sens. Netw. 2015, 6 (2015)
9. Mukherjee, A., Paul, H.S., Dey, S., Banerjee, A.: Angels for distributed analytics in iot. In:
2014 IEEE World Forum On Internet of Things (WF-IoT), pp. 565–570. IEEE (2014)
10. Bonomi, F., Milito, R., Zhu, J., Addepalli, S.: Fog computing and its role in the internet of
things. In: Proceedings of the First Edition of the MCC Workshop on Mobile Cloud Computing,
pp. 13–16. ACM (2012)
11. Nandimath, J., Banerjee, E., Patil, A., Kakade, P., Vaidya, S., Chaturvedi, D.: Big data analysis
using apache hadoop. In: 2013 IEEE 14th International Conference on Information Reuse and
Integration (IRI), pp. 700–703. IEEE (2013)
12. Hashem, I.A.T., Anuar, N.B., Gani, A., Yaqoob, I., Xia, F., Khan, S.U.: Mapreduce: Review
and open challenges. In: Scientometrics, pp. 1–34 (2016)
13. Morabito, V.: Managing change for big data driven innovation. In: Big Data and Analytics,
pp. 125–153. Springer (2015)
14. Bhardwaj, A., Bhattacherjee, S., Chavan, A., Deshpande, A., Elmore, A.J., Madden, S.,
Parameswaran, A.G.: Datahub: Collaborative Data Science & Dataset Version Management at
Scale. arXiv preprint arXiv:1409.0798 (2014)
130 R. Sharma and R. P. Mahapatra

15. Hortonworks: Accessed on 3rd June 2017. [Online]. Available: https://hortonworks.com/

16. Sharma, R.: Evolution in smart city infrastructure with IOT potential applications. In: Internet
of Things and Big Data Analytics for Smart Generation, Vol. 154. Springer, ISRL (2018)
17. Sharma, R.: Security threats in E-Commerce. In: Handbook of E-Business Security. CRC Press,
Taylor & Francis Group (2018). ISBN 9781138571303
18. Sharma, R.: Steps for implementing big data and its security challenges. In: Data Intensive
Computing Application for Big Data, Vol. 29. Advances in Parallel Computing series of IOS
PRESS (Scopus Indexed) (2018). ISSN 1879-808X
19. Ding, Z., Yang, Q., Wu, H.: Massive heterogeneous sensor data management in the Internet
Of Things. In: IEEE International Conferences on Internet of Things, and Cyber, Physical and
Social Computing, pp. 100–108 (2011)
Adopting Big Data Analysis
in the Agricultural Sector: Financial
and Societal Impacts

Rasha Kashef

Abstract Big data analytics (BDA) is constantly formulating decisions in

agriculture and transforming the processes by which agriculture operates and con-
trols. The agriculture process can be composed of a sequential flow of stages, starting
from planting, spraying, fertilization, collection, and distribution to the end consumer.
The variety within agriculture data is distinctly heterogeneous. This big farming data
can come in all shapes and sizes from a variety of sources, which composes a hard
analytical process for decision-makers. Big data analytical techniques have been
applied to each stage in the agricultural operational process. These techniques have
been proven to establish both economic gain for farmers and environmental and
safety benefits for society at large. This chapter summarizes the state-of-art analyti-
cal methods that have been recently used in the agriculture industry, along with their
financial and societal impacts.

Keywords Big data · Machine learning · Data analytics · Image processing ·

Agriculture

1 Introduction

A crucial factor in the profitability of a crop-based agricultural operation is the

unpredictable nature of weather conditions, which can render the harvest completely
unusable. In these cases, destroyed crops are not only an economic loss for the
producer but also a loss for the society in overall sustainability. One solution to
the growing demand for food is the use of big data through precision agriculture.
Using big data can create easy-to-use and accurate forecasting models to ensure that
crops are produced at their full potential, and this is when the idea of predictive
analytics comes in. In general, predictive analytics is used to extract trends and
predictive insights from data. Traditional weather forecasting models have been built
based on deterministic modeling starting with initial conditions, and then entered a

R. Kashef (B)
Electrical, Computer, and Biomedical Engineering, Ryerson University, Toronto, ON, Canada
e-mail: [email protected]

© Springer Nature Singapore Pte Ltd. 2020 131

P. K. Pattnaik et al. (eds.), Internet of Things and Analytics
for Agriculture, Volume 2, Studies in Big Data 67,
https://doi.org/10.1007/978-981-15-0663-5_7
132 R. Kashef

supercomputer model to produce a prediction; and later in the 1990s, an ensemble

forecasting approach was introduced where multiple systems are combined to process
and handle data [1]. The current challenge is to enhance the modeling, the statistical
analysis, and visualization technologies, and adapt them to the agricultural industry.
The latest advances in big data analytics around the economic gain of all types
of production; yet, societal concerns, such as sustainability and food safety, tend to
dictate the trend of big data analytics which is especially pertinent in the agricul-
tural industry. In 2015, venture capital funds invested a total of $64.5 billion across
thousands of technology startups, ranging from autonomous car systems to social
media companies [2]. During the same period, approximately 7% of this amount was
invested in agriculture-related startups. While still a significant amount, it is ironic
that an industry critical to sustaining life is not the recipient of more innovation.
In part, this is likely due to the traditional nature of the agriculture industry; how-
ever, the innovation that is occurring within the industry has the potential to bring
about significant environmental, social, and economic social change. The following
chapter outlines big data technologies that are impacting the agriculture industry in
the aforementioned areas and also analyzes the industry trends that are limiting its
innovation. In this chapter, major characteristics of machine learning, big data, and
the analytics technologies that are impacting the agriculture industry in the afore-
mentioned areas and also analyzes the industry trends that are limiting its innovation.
To pave the way of discussion in this chapter, the list of abbreviation that are used
across the entire chapter is summarized in Table 1.
This book chapter is organized as follows: in Sect. 2, machine learning and its
categorizations are presented. The notion of big data is illustrated in Sect. 3. An
introduction to agriculture data is presented in Sect. 4. Current big data analytics
techniques in the agriculture field are presented in Sect. 5. Section 6 provides a
discussion of analytics in food safety. Section 7 introduces some analytics software
in the agriculture industry. Section 8 discusses the future of analytics in the field.
Finally, conclusions are presented in Sect. 9.

2 Machine Learning: Definition, Categorization,

and Validation

Machine learning is known as the automated mechanism of discovering information

and knowledge from data using intelligent machine that uses either the unsupervised
learning, as opposed to supervised learning when it comes to learning from data.
Supervised learning is when data has to be manually put in by a user to help cat-
egorized and organized clustered data while unsupervised learning allows for the
categorization and organization of data without being programmed by a user [3].
Advances in digitization and computing power allowed data scientists to train com-
puters to model predictive events instead of constructing finished models. It includes a
collection of algorithms, methods, and tools able to embody intelligence to machines.
Adopting Big Data Analysis in the Agricultural Sector … 133

Table 1 List of symbols and abbreviations

Abbreviation Description
ANNs Artificial neural networks
DNNs Deep neural networks
CNNs Convolutional neural networks
NMI Normalized mutual information
VI Variation of information
NVI Normalized variation of information
NLP Natural language processing
LTI Labeled parse tree induction
AMI Adjusted mutual information
NVD Normalized Van Dongen
CSI Centroid similarity index
HSV Hue-saturation-value
RI Rand index
FGRI Fuzzy generalized Rand index
DI Davies-bouldin index
EED Expected edge density
SI Silhouette index
DAG Directed acyclic graph
SOM Self-organizing maps
GIS Geographic information systems
RFID Radio frequency identification
SEO Search engine optimization
BoW Bag of words
SVM Support vector machines

Machine learning tools have been used extensively in several applications across dif-
ferent industries. Machine learning is also used in other industrial applications like
pattern recognition, video surveillance, text mining, and even for cancer prognosis
and prediction. Given the advancement projected in the next few years, machine
learning is expected to be incorporated in more applications. The true potential of
machine learning is realized by its modeling tools, which, when trained via a learning
procedure with a dataset, uses algorithmic learning to respond to similar data trends.
However, it still falls quite a way short of replicating human learning techniques.
Where machine learning excels, however, is while crunching large data sets while
accounting for several permutations of variables to model an event. Overall, machine
learning is characterized as a type of the artificial intelligence field and it increases
the overall efficiency of data analysis, as processes that were once done by humans,
can now be automated by machines.
134 R. Kashef

Machine learning in big data is preferable to traditional methods, such as Excel.

Programs such as excel have a hard time handling big data due to the sheer size
of the information, which causes the program to crash. However, machine learning
works with big data because it feeds off of it and if you provide machine learning
algorithms with more information, they become more accurate [4]. There are a variety
of different techniques that are used for big data analysis and machine learning. One
method of machine learning is decision tree learning. With decision tree learning, the
algorithms use a set of defining characteristics, where they assign data points under
these characteristics and produce a set of predictive probabilities [5]. Another method
of classification is clustering. Clustering is useful for analysis, as most modern-day
analysis only requires a few attributes because having too many attributes will prevent
any usable trends to be found. It is useful when the big data points are fairly close
together, and you want to discover discernible trends [6].
Finally, another method of machine learning is the artificial neural network
(ANN), which is mainly mimicking the human brain and tries to mimic human neural
patterns. Information is clustered together in “artificial neurons” and linked together
axons, relationships between the neurons. ANN is useful when there are nonlinear
relationships for the information and allows multiple relationships between clusters
of information [6].
For supervised learning, the commonly used method is k-nearest neighbors. The
k-nearest neighbor’s method does not split the dataset to the training set and testing
set. It simply uses the minimum-distance classification method to classify the result
into the nearest user-specified k value. For unsupervised learning, the k-mean method
is used to classify the unlabeled data. Compared to the k-nearest neighbor, k-mean
method has the same goal which is to separate data into groups. It does not have a
set neighbor cluster centers but calculates a center point of the clusters and groups
similar data into clusters by minimum-distance classification based on the similarity
of the features. The group label will be added after the classification procedure [7].

2.1 Performance Validation

Results validation is the process of approximating how the output fits the structure
underlying the input data [8]. Evaluating the clustering results is more challenging
than evaluating classifiers, as in supervised learning labels are available, and we can
compute performance statistics such as accuracy [9]. Validation metrics are classi-
fied into two categories—external validation and internal validation. The external
ones evaluate the result based on some supervised information available while inter-
nal ones evaluate the result using the information intrinsic to the input data [10].
F-measure, NMI measure, entropy, and purity are popular external measures. Rand
index, fuzzy Rand index, expected edge density, and Davies-Bouldin index are inter-
nal validation measures. Some of these measures are discussed next.
Adopting Big Data Analysis in the Agricultural Sector … 135

2.1.1 External Validation Measures

F-measure is mainly defined on the idea of pairing similar classes (or clusters) in
two partitions [11]. It depends on a pre-processing stage; in this stage, each cluster
is mapped to a class. F-measure is widely used in document clustering [12]. The F-
measure cannot applicable to nested clustering [13]. Moreover, it cannot deal properly
with the issue of class size imbalance [14]. hF-measure is one of the most promising
indices of nested clustering quality based on the experiments on 12 datasets of Polish
Wikipedia articles [13].
L-measure is derived from F-measure. It is a useful evaluation metric that can be
used to measure the quality of a computational lexicon based on clustering concepts.
Compared with the F-measure, it performs well on a small data sample [15].
Entropy is the weighted average of the sizes of each cluster. It provides a well-
designed solution to the matching problem. However, it only measures the homo-
geneity of a solution [12]. V is like other entropy-based criterion that evaluates the
level at which both the homogeneity and completeness criteria have been met. It is
dependent on how many clusters, the clustering methodology, and the data size. This
metric supports solutions having many partitions (clusters) [11]. It has value in the
interval [0, 1], and it strongly favors a solution of many small-sized clusters.
VI stands for the variation of information. It is known of two main characteristics;
first, it satisfies the metric axioms. Second, it is convexly additive [16]. Nevertheless,
the range of scores provided by the VI measure is mainly related to the data size
[11]. (The values of VI lie in the interval [0, 2 log N], where N is the data size).
Normalized Variation of Information (NVI), the value of the NVI good cluster-
ings considered lie in the interval [0, 1]. It maintains the convex-additivity property
of the previous VI but not its metric axioms [17]. For the perfect solution, NVI = 0.
NVI performs well on large real datasets [16].
Purity only measures the homogeneity of a solution [12]. Clustering solutions
with purity values close to 0 are considered bad solutions; a purity of 1 means opti-
mum clustering [18]. The singletons clusterings are considered optimal; although, it
has low quality [11]. The matching criteria are the main factor to differentiate between
purity, normalized Van Dongen (NVD), the criterion H (CH), and the centroid sim-
ilarity index (CSI). If the matching solutionis the same, all of the above indexes
should obtain a similar clustering result. Various measures known as classification
error (or accuracy) are also similar to the purity measure [9].

2.1.2 Internal Validation Measures

The Normalized Mutual Information (NMI) measures the shared information that
is adopted in both the clustering and the partition. The value of the NMI is between
the values of [0, 1]. Values close to 1 show better solutions [19]. NMI equals the
adjusted forms of mutual information (AMI) that equals the normalized variation
of information (NVI) which also equals the variation of information (VI) under
hypergeometric distribution factors, as illustrated by results in [9].
136 R. Kashef

NMI_SM performs well on fuzzy partition generated by fuzzy-c-means, using a

collection of 25 artificial and ten real datasets [20]. The artificial datasets contain five
ground-truth groups, collected as a mixture of Gaussian probability distributions in
the two-dimensional space vector.
Rand Index (RI) measures the similarity between crisp clustering. It corresponds
to accuracy even when the class labels are not found. Adjusted Rand index (ARI) is a
variation of the Rand index (RI) that is attuned for the chance grouping of elements.
The values of the RI ranges from 0 to 1, with 0 showing that the two clusters disagree
on any pair of points and the value of 1 denotes that the data clusters have the same
points [21]. Fuzzy generalized Rand index (FGRI) broadens the scope of the RI for
considering the cases of fuzzy partition and fuzzy cluster ensembles [22]. On real-
world document clustering data, the Expected Edge Density (EED) outperforms the
classical cluster validity measures, including Dunn and Davies-Bouldin index [23].
Davies-Bouldin Index (DI) is known as the fraction of the sum of “within-cluster”
distance to “between-cluster” separations [24]. It cannot correctly find the optimal
clustering among clusterings that can be stemmed from a document cluster, for
example. [23]. The experimental results from the comparison of the effectiveness
of a multiple classifier intrusion detection system reveal that Davies-Bouldin index
is faster than Silhouette Index (SI) [25]. Modified Davies-Bouldin index is pro-
posed for the rough fuzzy algorithm that adds relative significance to lower versus
upper approximations. However, it underperforms on multiclass data because it either
increases or decreases monotonically [26].

3 What Is Big Data? And Why?

Big data is defined as large sets of information that exceed current tools to distill this
information and provide analysis [27]. Big data is information that is a combination of
two main characteristics; size and complexity [28]. The size of big data is not defined
by the amount of memory it takes up, rather the number of sources it contains. For
example, planes can have up to a hundred thousand different sensors, measuring
carbon dioxide levels, speed of the plane, and the outside temperature. For an hour-
long plane ride, the total cumulative information from all these sensors only takes up
roughly 3 GB. [29]. As well, technology is improving over time, meaning that the
amount of memory the information takes up will decrease [27]. The complexity of big
(large-scale) data is defined as the relationships and how the data can be arranged with
one another. Facebook has 30 billion pieces of unique information shared each month
[27]; the sheer size of this data can cause an immeasurable amount of combinations
for relational data. Big data analytics includes the methods and devices to synthesize
this information into presentable findings.
Analytics allows for the discovery of trends and relationships, which companies
and organizations can act on to improve efficiencies, create new opportunities, and
remain competitive in their respective industries. There is a multi-step process when
it comes to big data analytics. The first step is the data requirements, which are the
Adopting Big Data Analysis in the Agricultural Sector … 137

purpose of acquiring data–for example, trying to find the relationship between age
and income of an individual. The second step is the acquisition of data, which can be
collected by completing experiments, as well as publicly available information. The
third step is data processing, which includes converting the data acquired to a form
of presentable finding. The fourth step is extrapolative/interpolative trend analysis,
which is the process of using the data acquired to predict trends within and outside
of your data range. The final step is modeling, which is the process of turning your
information and the data you acquired into a general formula/relationship.
The first big data technique was used in 1890 for the Census Bureau. Herman
Hollerith was contacted by the bureau, and he came up with the idea to use punch
cards and tabulation machines to process the massive amounts of data [30]. There
are also a large variety of big data that can be used. There is text analytics, which
is used to analyze social media, emails, news, and documents. The most common
application for text analytics is text summarization, producing a simple summary
from large sources of text. Audio analytics are currently most commonly used in
call centers and health care, mainly in the area of diagnosis, to improve the overall
customer experience. Video analytics mainly focus on the analysis of closed-circuit
television, to detect security breaches. There is also predictive analytics, which is
used to discover trends and extrapolate them. This can be applied to predict the failure
of certain products and predict consumer behaviors before an upcoming shopping
season [31].
For data analytics, NoSQL, Hadoop, HDFS, and R are just a few examples of
programs and tools that allow big data to be analyzed [29]. Hadoop is a popular
tool for analyzing big data because it started off as an open-source project, allowing
many different insights and perspectives to be implemented in the software. As well,
having many people working on the software allowed the program to scale quickly
[32].

4 Agriculture-Based Data

The variety within agriculture data is distinctly heterogeneous [32]; this means that
all the data in farming comes from various sources. For instance, historical crop
patterns typically govern the crops that can be planted in a present year and dictate
the quality of the crop in future seasons. Additionally, climate conditions and weather
conditions can be strictly predicted based on historical trends and forecasting. These
are very important aspects when it comes to pre-planting insights and decision-
making. However, beyond this historical data, there is an increasing percentage of
data observations being delivered and processed in real-time through sensor data. This
is referring to the “Internet of Things”, where sensors are continually reporting on
the status of a physical object within a process. These are high volume, high velocity,
high veracity data that play a major role in agriculture. Sensor data capturing is a
relatively new concept that is now being implemented intensively in agriculture. For
example, GPS receivers and satellite images can continuously monitor a crop and
138 R. Kashef

provide insights into the strength of a field [1]. Moreover, web-based data is merging
into the analytics field within the agricultural process as it can indicate food safety
once a crop or protein has been delivered to the end consumer.

5 Analytical Models in Agriculture

As the world population continues to grow over 7 billion, feeding people will become
more of an issue, and this responsibility falls on the shoulder of the agricultural
industry. As the demand for both food quantity and quality rise, improving pro-
duction efficiency becomes a challenge that must be addressed. A crucial factor in
the profitability of a crop-based agricultural operation is the unpredictable nature
of weather conditions, which can render the harvest completely unusable. In these
cases, destroyed crops are not only an economic loss for the producer but also a loss
for the society in overall sustainability. One solution to the growing demand for food
is the use of big data through precision agriculture. Adopting big data can create
easy-to-use and accurate forecasting models to ensure that crops are produced at
their full potential, and this is when the idea of predictive analytics comes in.

5.1 Weather Forecasting Models: Precision Agriculture

The two techniques widely used to build weather forecasting models currently are
regression (both linear and nonlinear), and time series (moving average, autore-
gression, autoregression moving average (ARIMA). On one hand, one can “easily”
regress a production decision variable (the dependent variable, such as frequency and
level of irrigation required) against weather conditions (the independent variables,
such as the amount of natural precipitation, sunlight and temperature) to discover
inherent relationships, which can be generalized to reliable practices as experiences
accumulate. On the other hand, time series that is primary based on the regression
structure, is a very powerful data representation as well in terms of capturing patterns
in historical data.
Modeling techniques, such as regression and time series, could be very powerful
in terms of making production-related decisions. Challenges still exist in the realm
of implementation. One challenge for big data analytics solutions in agriculture is
aggregating a large volume of data through multiple sources and then able to properly
handle this data. In addition, people (mainly producers) have to be properly trained
to use the models and make reasonable business decisions from them, which is a
long-term, expensive, and time-consuming task [1].
Taking advantage of precision agriculture, producers can estimate business needs
ahead of time and make production plans accordingly. Resource allocation tasks,
such as machinery rental or purchase, and the timing and usage of fertilizers, can
be projected more effectively, which would contribute to the cost-effectiveness and
Adopting Big Data Analysis in the Agricultural Sector … 139

overall sustainability of the business over time. The effect of weather can have a
costly impact on farmers. The failure to predict an upcoming storm can lead to over-
irrigation in the days leading up to it, causing water to be wasted and the potential
for flooding in some areas. If the storm arrives soon after the farmer recently sprayed
crops with pesticide or fertilizer, this can wash away the chemicals, and the farmer
will have to repeat the process once the weather clears.
Traditional weather forecasting methods use numeric weather prediction (NWP)
mathematical modeling based on Bayesian probabilistic arguments, which have
historically led to accurate weather approximations. However, NWP problems are
largely constrained by the amount of computing power available, which has made it
difficult to provide weather predictions at a high enough resolution for it to be useful
for farmers IBM is solving this problem through the use of precision agriculture [33].

5.2 Neural Network Models

Artificial neural network (ANN) provides smart alternative prediction methods that
are widely applied for agricultural forecasting. The commonly used learning pro-
cesses in artificial neural networks include the radial basis function (RBF) [34], the
perceptron learning [35], the backpropagation algorithm [36], and resilient backprop-
agation (RBP) [37]. The feedforward backpropagation neural network (FFBP-NN)
is the widely known neural nets. The architecture uses a combination of neurons and
layers to create a series of connections in a directed acyclic graph (DAG), to provide
a classification output. The architecture has three main layers including the input,
the hidden (of 1 or more neurons), and the final layers as shown in Fig. 1.

Fig. 1 FF-BP neural network

140 R. Kashef

Fig. 2 Success rate using CNN [43]

The ANNs-based learning methods have been used in various fields, such as the
counter propagation algorithms [36], the auto-encoder, and the supervised Kohonen
networks [37]. The Hopfield Networks [38], the self-organizing maps (SOM) [39],
the extreme learning processes [40], the generalized regression-based NN [17], and
the adaptive neuro-fuzzy inference systems [41]. Ensemble Neural Networks, and
Self-Adaptive Evolutionary Learning Systems [42]. A deep neural network (DNN) is
considered as an ANN which is composed of multiple hidden layers and it is classified
as supervised-learning NN, semi-supervised NN, or completely unsupervised NN.
DNNs are known for achieving higher accuracy as compared to the traditional
ANNs; however, it suffers from high time complexity. The convolutional neural
networks (CNNs) are DNNs, where the feature maps can be obtained by applying
convolutions in the image-domain. In [43], a CNN model is applied to digital images
for recognizing the face of pigs. The CNN model aims at identifying animals without
the use of tags of the radio frequency identification (RFID). The model has achieved
an accuracy of 96.7%. Another CNN-based model [44] is designed to provide detec-
tion and diagnosis of a disease based on images of leaves to categorize healthy and
infected leaves in multiple plants. The success rate of the CNN model is presented
in Fig. 2.

5.3 Image Processing Models

Besides the effective usage of geographic information systems (GIS) in weather

forecasting, which shows a significant role in the agricultural analysis. Common
examples of GIS include using satellite images, drone pictures, or thermography to
detect field conditions such as temperature, soil condition, and pest damage. The
condition helps farmers to create efficient techniques to grow crops which increase
food production. On the other hand, the regulator also benefits from the geographic
information system, which helps regulators enact effective ordinance by providing
specific information in real time. The big data analysis has been in applied in agri-
culture using both image processing and classification. On the macrolevel, a shared
Adopting Big Data Analysis in the Agricultural Sector … 141

website for agronomists is a good example of using satellite images to monitor the
land type and current condition. The result can be used to help fertilizer providers
and grain suppliers to have a general idea about the field condition before making
decisions.
On the micro-level, farmers can get the real-time field condition information to
make planting decisions by reviewing analytical results generated from images taken
by drones instead of observing the crops on foot through the whole field.
The common algorithm behind the image processing and classification is that
each pixel is converted from the picture into a data array. The data array includes the
coordinate of the pixel, the color (in an RGB scale) of the pixel, the reflectance, etc.
Each pixel becomes a row in the dataset, and the whole picture will be converted to
a data table. After the conversion, all the visionary information becomes numbers
which can be processed and classified by a computer [33]. In general, a 1920 * 1080
image contains more than two million pixels, which mean each data table contains
a massive amount of data. The next step will be to classify the picture by the data
table information. At this step, different machine learning methods will be used
based on the study need. If the training data was labeled, the supervised learning
will be used; otherwise, the unsupervised learning would be used for classification.
The combination of computer vision and machine learning can be applied to identify
plant species. These two techniques manifest themselves in a detection–classification
two-step process. First, a camera takes a picture of a leaf from the plant under
consideration. In a typical application, it must be verified that the picture that there
actually is a leaf within the frame before proceeding. This is done by dividing the
image into different regions and creating a low dimensional feature vector of each
region.
This collection of low dimensional feature vectors is called a “gist”. The gist
is subsequently fed into a classification algorithm, for example, a support vector
machine (SVM). The classification SVM algorithm is then trained by taking as input
labeled images which either contain a leaf or do not, and then creates an optimal
hyperplane which maximizes the margin between the two sets of images. If the
image contains a leaf, the process continues, and the image is segmented into a
binary image (a black–white image), separating the leaf image from the background.
First, to improve the segmentation accuracy, the data image is converted to a hue-
saturation-value (HSV) space to remove the effects of lighting. In the next step, an
expectation-maximization (EM) algorithm assigns probabilities to individual pixels
and determines if a pixel corresponds to either the leaf space or non-leaf space,
coloring the pixel appropriately. Finally, features from the leaf such as color, shape,
and size are extracted from the binary image to construct a vector of features. This
vector of features is then used as input to the network to predict the plant species
from which the leaf came from.
142 R. Kashef

5.4 Social Network Models

Using social and mobile big data applications could be a way to avoid a situation
where a crisis could occur. By using text mining and social network analytics, orga-
nizations can have faster responsive actions and therefore limit the consequences
associated with food insalubrity. An example of the utilization of those techniques
can be seen when a high-risk contaminating disease is being detected by streams of
data coming from keywords used in research engines and IP locations [45].
For analytics’ purposes, keywords such as indigestion, mold, and insalubrious
could be targeted by keyword trackers. Then, a system could flag the entities that
sold food to this area depending on the search engine optimization (SEO).
Having access to frequent, low-cost, and online data makes social networks an
ideal start to crowd source trends using big data. So, by using the “crowd” as a data
source, it is possible for all participants included in the food production process to
locate areas with a high probability of food-safety issues at a rapid pace.
The velocity of this data is nearly unfathomable as social network analytics deals
with a large volume of data from social media sites every millisecond. The volume
also corresponds to this large amount of data, as there is a constant stream of informa-
tion from sites such as Facebook and Twitter. However, these social media platforms
provide valuable insights into food safety and early detection method. In the spread
of contaminated food, early detection can quite literally be the difference between
life and death. Therefore, having constant surveillance on social media means that
the food industry is able to detect when and where certain buzzwords are being used
that would relate to an outbreak of food contamination. This way, using big data, the
spread of dangerous food can be halted at the source.
In [46], authors have monitored Yelp reviews in the area to try and detect cases of
food poisoning. In the way described above, the department used machine learning
to mine through thousands of Yelp reviews and attempted to classify if a person or a
group was experiencing foodborne illness. On all the reviews sampled, the authors
found 468 yelp reviews that indicated a foodborne illness. Of those 468, only 14 were
reported to food-safety agencies, which leave room for earlier detection and better
response from the industry [47]. To use machine learning, the authors converted each
document (review) in Bag-Of-Words (BOW), used three words per BOW (n-gams
= 3) and then applied logistic regression by classifying if a person was “sick” using
“Yes” or “No” [47]. This program was adopted in 2012 by the NYC Department of
Health and Mental Hygiene and has been able to correctly identify ten outbreaks of
illness due to contaminated food using review sites. The potential for this kind of
application is unending as fundamental text mining process can be applied to Twitter
and other social databases. In this way, the frequency of keywords can be coupled
with the location of the complaints to pinpoint certain areas and food vendors who
may be infected.
Also, cell phones have become prevalent across the world, especially in developing
countries. From a big data perspective, this gives people access to data on orders of
magnitude far greater than we have been exposed to previously. Just as big data has
Adopting Big Data Analysis in the Agricultural Sector … 143

widespread application, so do technologies such as cell phones [33]. Based on the

statistics provided by TechAsia, 29 million people use twitter, accounting for 2.4% of
the global population [48]. Twitter is a major source of data for text mining that can be
used to extract insights about common interests and issues is Twitter. In developing
countries, Twitter is being used to derive insights about social well-being in relation
to the food price index, which measures the monthly change in commodity prices.
Families in countries such as Indonesia are highly exposed to fluctuations in food
prices, where a spike in prices could have a devastating impact on the nation’s food
security.
This means that citizens in these countries are highly aware of events that could
lead to food price increases and are constantly sharing their thoughts on food prices
through social media channels. In assessing such a scenario, social media acts as an
informant to monitor food security.
As a powerful analytical tool, social media analytics can provide insights between
conversations, food price, and other external events. Analyzing social media patterns
can be a formidable challenge given the large-scale nature of the accessible data.
The first step involves creating categories for “tweets” describing food and fuel
prices. A supervised algorithm was trained by creating categories such as “positive”,
“negative”, “confused”, and “neutral” [49]. A score is then given to each category,
and an overall sentiment score is produced. In supervised learning, the algorithm
typically has more context around the words and therefore provides to be a more
accurate method of interpreting tweets in real time. Typically, there is a lag between
the food price index and the actual cost of goods. In analyzing the Twitter data, there
was an association determined between the content and the number of Tweets and
the final impact on the food price index. The number of tweets increased as food
prices increased [2]. This is demonstrated in Fig. 3.
As food prices are also dictated by fuel prices, potential increases in fuel price
were related to people’s perception of overall food security. Moreover, Twitter con-
versations are driven up by an increase in food prices as well as other food-related
topics such as policy and law [33, 49]. For example, the establishment of a food
agency coincided with increased chatter on Twitter.
Furthermore, Twitter can serve as a powerful tool to analyze people’s perception
of food prices in a real-time environment. It can lead to powerful insights that can be
used to extract daily food prices from conversations and essentially create a real-time
food price index.
Since there is a lag between the food price index and the actual cost of goods,
twitter analytics can be used to nowcast the cost of food [50, 51]. Although twitter
analytics serves as a robust tool to establish trends regarding social and economic
factors in a country, efforts need to be made to increase the accuracy of these tech-
niques.
144 R. Kashef

Fig. 3 Public sentiments for food price rise in twitter conversation [49]

6 Food Safety

The typical food chain production starts with a producer and ends with a customer
buying the product. However, the journey in between is incredibly varied and depends
on the product type, the region, and the legislation around food production. Even if
the channels are not standardized, the food must be. Following regulation is what
assures the wholesome quality of the food that is consumed. Unfortunately, in some
cases, it may show that a single event of contaminated products can heavily impact
a well-established business as well as the industry from which it is derived. The
future of food safety as the end process in agriculture operations will be in refining
the classification of the data. With so much data coming in and being processed
in real time as well as the rapid increase in internet circulation, the use of social
network analytics will become more pertinent to the food-safety agencies and for
those involved in the food industry [52].

7 Business Adoption Software

As stated previously, there are several benchmarks by which one could judge the
impact of a big data technique in agriculture. The two main categories include finan-
cial and societal impacts. In this case, estimating the environmental impact of farming
is becoming an increasingly important process by which to assess the societal sustain-
ability of certain processes. Therefore, as more big data techniques are developed, the
natural progression points to enabling agricultural processes to diminish their impact
Adopting Big Data Analysis in the Agricultural Sector … 145

on both water and land. In the case of combining real-time sensor data with historical
data and GIS, the applications can be limitless. For example, being able to track wind
conditions with fertilizer placement can give an indication of the concentration of
fertilizer that is entering nearby water bodies. In doing so, big data applications can
merge a whole variety of data to ensure that agricultural processes become more
sustainable in the future; some of these software applications are discussed next.

7.1 AgriBigCAT [51]

AgriBigCATis an online software platform which can estimate the environmental

impact of agricultural activity by geospatial and big data analysis. This platform not
only assists farmers in their decision-making process but also helps government and
administration to balance the farmland reclamation and the wetland or forest preserve.
The platform allows the acquisition and analysis of the significant parameters of a
large-scale in many agricultural systems in real time. The high volume and high
variety data are stored in Apache Hive and the analyzed by ArcGIS. This variety of
data spans from sensor data to historical data. The architecture of the AgriBigCAT
is displayed in Fig. 4.
In terms of velocity, there is a constant stream of sensor data being produced on
site and this is being processed and recorded by the program. This leads to many
insights during a long time periods. On the data veracity part, the accuracy was guar-
anteed by cooperating with trustworthy organizations for collecting data. There are
several applications of this technology. Furthermore, nitrogen production from farm
animal dejection can be tracked and can be coupled with municipal data to ensure that

Fig. 4 AgriBigCAT architecture [51]

146 R. Kashef

nitrogen is not being produced over the acceptable limit. On a manure management
and planning project, the platform considers all the farms and animals in the sys-
tem, calculates the estimated emissions under the current weather and temperature
condition, and prevents the negative effect created by manure land overdosing.
In general, this system and this brand of big data are useful as it can give guidance
to a producer in where they may dispose of manure, among other things. However,
there is also inherent usefulness for the policymakers and municipal offices as the
data can provide an estimate of the pollution being created, and policies can be set
in place to mitigate environmental damage.

7.2 LettuceBot [53, 54]

One commercial application of the image recognition technology is the LettuceBot,

which has been designed by Blue River Technology as a smart machine to thin lettuce
fields. In order to ensure that a high yield of lettuce is produced, farmers will plant
lettuce tightly together, but this makes it difficult to selectively remove weeds and also
creates the potential for closely packed plants to choke themselves out. LettuceBot
is towed behind a tractor and uses multiple cameras to identify lettuce plants from
weeds and the surrounding soil.
An optimization algorithm then determines the optimal lettuce plants to keep
based on spacing and size measures, killing any weeds or unwanted lettuce plants
by spraying them with herbicide. The LettuceBot significantly reduces the environ-
mental footprint of the farm by reducing the quantity of herbicides that are needed
to protect the crops. This prevents over spilling issues that cause chemicals to drain
into underground water supplies. The additional benefit is the added cost savings by
efficiently managing the use of chemicals. In addition, LettuceBot replaces a process
that would have previously been undertaken by hand. With further refinement of
this technology, other plants and farming operations can be automated to reduce the
impact of overusing herbicides. The application that LettuceBot addresses also leads
to some simplifications of the computer vision problem. First, the soil provides a
large contrast in color, making it easier to segment a plant from its background in an
image. Furthermore, only one type of plant requires identification. This allows both
the simplification of algorithm development and the potential for the algorithms to
demonstrate better real-time performance. At the same time, variations in plant size
and shape can introduce errors. Fortunately, the criticality of a false-positive or false-
negative is small, and as long as the rate at which these errors occur is minimized,
these errors are acceptable. Figure 5 shows the rendering of LettuceBots thinning a
field.
Adopting Big Data Analysis in the Agricultural Sector … 147

Fig. 5 Rendering of LettuceBots thinning a field [53, 54]

7.3 Deep Thunder [55, 56]

Deep Thunder extends NWP mathematical modeling by taking advantage of the deep
neural network (DNN) embedded in the IBM Watson AI [56]. The purpose of this
neural network is not only to provide precise weather predictions but to determine
the business impact of these predictions for farmers. Localized sensory inputs from
the farm are included as inputs to the model to understand the effect of weather
on a particular region. As an example, Georgian farmers are typically dependent
on irrigation even though they get far more rain than other parts of the country.
The porous soil prevents it from retaining the moisture from a heavy rainfall [56].
By using camera inputs on the farmer’s property, a machine learning mechanism is
trained for estimating the amount of rainfall based on computer vision and combine
this information with sensory information from the soil moisture level [57]. The
resulting model provides an output for the effect of rainfall on a particular area of
the farm. The impact of rainfall is affected by the landscape; a hilly landscape will
cause a serious runoff followed by a heavy accumulation, especially in lower-based
valleys. By using hydrology and flood modeling, Deep Thunder is able to estimate
where the runoff will go and where flooding might occur [56].
The final step for Deep Thunder is to combine the weather forecasts with the
sensory data from the farm. The result is a predictive model that is able to pinpoint
exactly where the farmer needs to irrigate today and for how long. An output is
produced every 12 h to give farmers the tools to make daily business decisions. The
predictions provide more efficient use of irrigation and pesticide supplies, ultimately
reducing the cost of materials and maximizing crop yield [58]. In addition to reducing
148 R. Kashef

costs for the farmer, the cost advantages propagate throughout the entire supply chain.
Being able to forecast weather predictions will make agricultural options cheaper to
pursue, leading to cost advantages for the end consumer. In Brunei, Deep Thunder
is being used to better understand regional conditions on the production of rice.
Historically, only 3% of rice was produced internally on the island, with the rest
imported. The objective is to use the analytics provided by Deep Thunder to increase
this number to 60%, which would reduce the cost of food in the country and improve
food security [56]. IBM’s Deep Thunder initiative operates on a parallelized, high-
performance computing system based on the scalable modeling system architecture.
This distributed approach allows the prediction problem to be divided into a hybrid
model involving multiple nested layers of prediction, each with a varying resolution
level, and the results are combined to provide hyper-local precision forecasting. The
initial resolution begins at 16 km and provides weather prediction for a region at
a high level. The output is fed into a 4 km resolution layer that creates a narrower
prediction grid. The final nest is a prediction grid with 1 km resolution, resulting in a
hyper-localized forecast, with 10-min intervals, up to 86 h in advance [59]. Figure 6
shows a sample output of the nested models from Deep Thunder.

Fig. 6 Output of the deep thunder model (nested layers) [59, 56]
Adopting Big Data Analysis in the Agricultural Sector … 149

8 Future of Analytics in Farming

The adoption of big data is not only facilitated by the current structure of the food
industry. While data collected on one farm might give some insights, it is insufficient
to support analytics on their own fully. To gain benefit from analytics using big data,
the producers would have to aggregate their data with other organizations in the
industry along the “food-to-table” chain because the capital cost of getting such data
can be significant.
Some producers are vertically integrated; they either possess the supplier or buyer
section of their business; this means that the corporation or cooperative they repre-
sent can aggregate the data and make a profit by selling it to other consumers. Having
independent producers, suppliers, processors, distributors, and the government col-
laborating by sharing a variety of data (sensor, streamed, historical) is the biggest
challenge the industry has to face right now given its structure [51].
In essence, the problem with implementing some of these big data techniques
within the processes of agriculture is that the farmers’ choice to aggregate their data.
With the scale of singular farm data, there is not enough data available to come up
with entirely formative insights. The additional scale from many farmers working in
collaboration is when real understandings are made, and real big data applications
can be created. Nevertheless, for the farmer, there needs to be an inherent return on
investment. It is likely that farmer will only aggregate data and buy remote sensors
if they are guaranteed a reward back. However, farming is not sufficiently funded
by the government to provide farmers with incentives. Therefore, education must be
done so that farmers can see the financial reward in getting behind big data.
However, protecting the environment and ensuring food safety are increasingly
important elements of importance for the consumers. The industry ultimately works
for the consumer to match demand. The pressure from consumers might help the
industry to go over the individual incentives given by the current structure and allow
for aggregation of data in a competitive industry. In the future, processors might
finance sensors for producers to have access to data related to the quality of their
crops, for example. These kinds of partnerships using big data will undoubtedly
affect the structure and define which player is going to stay competitive within the
industry.
Big data is already having a significant impact on individual farming operations.
LettuceBot and Deep Thunder both represent ways in which farmers can harness
big data and technology to improve their own activities. The competitive advantage
gained by using such techniques is clear within smaller markets and for individual
farms, and the drive to adopt such technologies is quickly being spurred by the
accompanying economic advantages. Farm managers are able to use the plethora of
internet-connected devices and sensors to monitor their operations at increasing levels
of precision. Yield measurement monitors, introduced into combine harvesters in the
early 2000s, have allowed farmers to track field growth statistics down to the square
meter. Soil sensors and moisture monitors can track nutrient levels and moisture
levels down to the specific plant. Bringing all this information together has allowed
150 R. Kashef

farmers to practice what’s known as precision farming, where they break down their
fields into tiny sections and treat them all as individual fields.
Companies such as John Deere, Monsanto, and DuPont Pioneer are leading
the revolution in precision farming techniques. Tractors can now plant seeds
autonomously, and John Deere and Pioneer both offer variable-rate seeding “pre-
scriptions” that take into account the specific production ability of portions of a field
and adjust the density of planting accordingly.
Monsanto offers a program called FieldScripts that includes tools like nitrogen
advisor; these tools track the available nitrogen at the field, and also it offers projec-
tions, recommendations, and alerts [49].
Two big factors stand in the way when it comes to making field data useful, both
for improving the individual operations of a farmer and for improving the agricultural
industry as a whole. These factors are integration and analysis. With all the different
sources of information being made available, integrating these sources in a useful
manner, one that facilitates decision-making, is a serious issue for farmers. The
ability to analyze all the information in a way that provides useful insights is beyond
the practical scope of an average farmer. Additionally, information gathered from a
single farm is not nearly as useful as information aggregated from across multiple
similar operations. This information can be used to analyze the effect of variations
between farming practices and large corporations are better suited when it comes
to collecting and aggregating this sort of data from multiple operations, and benefit
from the scale when it comes to the cost of analytics work [52]. Large companies like
John Deere and Monsanto have all been collecting data and performing the analyses
to aid farmers with applications like variable-rate seeding. Due to the competitive
nature of corporations, however, information is not being shared—to the detriment
of farmers, and the agricultural industry as a whole.
“By 2050 the world population reaches approximately 9.2 billion. Thus, if we
continue producing food at the current speed, the world cannot be fed” [49]. The vast
amount of agricultural data available leaves opens the possibility for research that can
optimize farming practices and revolutionize the industry. Historically, agricultural
research has been led by governments and was done in the public sector [51]. Past
research was done on small plots to test how different practices affected plant yields
and was made public to benefit society as a whole. Over the past couple of decades,
however, the majority of agricultural research has shifted to the private sector [51].
Research is being performed by large corporations whose business interests pre-
vent them from sharing what they discover. With all the data available, there is
potential for large-scale experiments that benefit the industry as a whole; however,
bringing together all of the data in a format that is useful is proving to be a challenge.
There are several initiatives being taken to try to bring order to the chaotic world
of agricultural data. Multinational corporations are driving the initiative with their
projects like Monsanto’s Fieldscripts; however, these proprietary solutions are creat-
ing issues over data ownership. Farmers are hesitant to provide their data to support
the analytics when they are forced to pay a subscription afterward to use the insights
gained from it. A number of startups have entered the game in an attempt to bring
Adopting Big Data Analysis in the Agricultural Sector … 151

together a wider range of data sources while giving more power to the farmers as
defined next.
• Farm Business Network, a Google Ventures startup, aims prioritize Farmers, this
can be established by democratizing information, establishing a non-biased ana-
lytics, and producing competitive advantages for farmers [60].
• Farmobile aims to help farmers own their data outright by providing a platform
for storing, sharing, and selling agronomic and machine data [59].
• Global Open Data in Agriculture and Nutrition (GODAN) is a joint inventiveness
defined by G8 governments to provide a standard for collaboration between partner
companies. Their goal is to endorse partnership to employ the increasing size of
data produced by innovative technologies to resolve long-standing difficulties and
to help farmers and the health of customers [1].
These companies are helping bridge the gaps by integrating data from multiple
systems and multiple vendors to fully make use of the technology available to farmers.

9 Conclusion

Big data analytics is continually informing decisions in agriculture and revolution-

izing the processes by which agriculture operates. Agriculture can be thought of in
several stages, from planting to spraying to the end consumer and supplying food to
mass populations. By applying analytical techniques to each step in the agricultural
operations process, there can be an economic gain for farmers as well as environ-
mental and safety benefits for society at large. This data can come in all shapes
and sizes from a variety of sources. The planting process is defined by forecast-
ing. Big data analytics can inform decisions on prediction and provide insights for
how producers should efficiently allocate their time and resources. Continually, big
data, in collaboration with geographic information systems, can provide a base for
large-scale agricultural mapping. This process can produce several insights on how
to treat different crops and how to efficiently supply fertilization techniques based on
k-means clustering. This can evolve into the identification of environmental impacts
on land and water bodies over time. Overall, the process can be defined by creat-
ing value for producers and sustainability for society. This way, policymakers can
make informed decisions on limiting policies and municipal bylaws for spraying and
planting techniques. Eventually, when the food arrives at a plate of the end consumer,
data analytics has been involved in the whole process and can detect any food-safety
concerns which assure the customer that they are getting the safest product available.
Agriculture is an industry that shoulders the responsibility of sustaining billions of
lives every day. Increasing investment in this space will further give rise to inno-
vations that have both an environmental, economic, and social impact. However, it
is critical that these innovations work together to help improve farmers’ operations
rather than make their work more complicated. By breaking down the barriers that
prevent the sharing of data across individuals’ farms, big data analytics can have a
152 R. Kashef

significantly greater impact with the goal of enhancing farm yields at both a local and
global scale as discussed in this chapter. This benefits not only farmers but humanity
as a whole.

References

1. Bendre, M.R., et al.: Big data in precision agriculture: weather forecasting for future farming.
Next Generation Computing Technologies (2015)
2. Jones, S., Rapp, N.: Where $64.5B in VC funding went last year. Fortune, 2 July 2015. Web.
12 Apr. 2017
3. Kotsiantis, S.B.: Supervised machine learning: a review of classification techniques (2007)
4. Dietterich, T.G.: Ensemble methods in machine learning (2000)
5. Kouru, K., Exarchos, T.P., Exarchos, K.P., Karmaouzis, M.V., Fotiadis, D.I.: Machine learning
applications in cancer prognosis and prediction (2015)
6. Magidson, J., Vermunt, J.: Latent class models for clustering. Can. J. Market. Res. (2002)
7. Halkidi, M., Batistakis, Y., Vazirgiannis, M.: On clustering validation techniques. J. Intell. Inf.
Syst. 17, 107–145 (2001)
8. Rezaei, M., Fränti, M.P.: Set matching measures for external cluster validity. IEEE Trans.
Knowl. Data Eng. 28(8), 2173–2186 (2016)
9. Rendón, E., Abundez, I., Arizmendi, A., Quiroz, E.M.: Internal versus external cluster valida-
tion indexes. Int. J. Comput. Commun. 5(1), 27–34 (2011)
10. Reichart, R., Rappoport, A.: The NVI clustering evaluation measure. In: Proceedings of the 13
Conference on Computational Natural Language Learning, pp. 165–173) (2009)
11. Rosenberg, A., Hirschberg, J.: V-measure: a conditional entropy-based external cluster evalu-
ation measure. EMNLP-CoNLL 7, 410–420 (2007)
12. Draszawka, K., Szymanski, J.: External validation measures for nested clustering of text doc-
uments. ISMIS Ind. Sess. 369, 207–225 (2011)
13. De Souto, M.C., Coelho, A., Faceli, K., Sakata, T., Bonadia, V., Costa I. : A comparison of exter-
nal clustering evaluation indices in the context of imbalanced data sets. Brazilian Symposium
on Neural Networks (2012)
14. Dalli A.: Adaptation of the F-measure to cluster based lexicon quality evaluation. In: Pro-
ceedings of the EACL Workshop on Evaluation Initiatives in Natural Language Processing,
pp. 51–56 (2003)
15. Rezaei, M., Fränti, P.: Set matching measures for external cluster validity. IEEE Trans. Knowl.
Data Eng. 28(8), 2173–2186 (2016)
16. Lei, Y., Bezdek, J., Chan, J., Vinh, N., Romano, S., Bailey, J.: Extending information-theoretic
validity indices for fuzzy clustering. IEEE Trans. Fuzzy Syst. 25(4), 1013–1018 (2017)
17. Cao, J., Lin, Z., Huang, G.: Self-adaptive evolutionary extreme learning machine. Neural Pro-
cess. Lett. 36, 285–305 (2012)
18. Mary, S., Sivagami, A., Rani, M.: Cluster validity measures dynamic clustering algorithms.
ARPN J. Eng. Appl. Sci. 10(9), 4009–4012 (2015)
19. Jang, J.: ANFIS: adaptive-network-based fuzzy inference system. IEEE Trans. Syst. Man
Cybern. 23, 665–685 (1993)
20. Yeh, C., Yang, M.: Evaluation measures for cluster ensembles based on a fuzzy generalized
Rand index. Appl. Soft Comput. 57, 225–234 (2017)
21. zu Eissen, B., Wißbrock, F.: On cluster validity and the information need of users. ACTA Press,
pp. 216–221 (2003)
22. Wu, K., Yang, M.: A cluster validity index for fuzzy clustering. Pattern Recog. Lett. 26(9),
1275–1291 (2005)
23. Halkidi, M., Batistakis, Y., Vazirgiannis, M.: On clustering validation techniques. J. Intell. Inf.
Syst. 17, 107–145 (2001)
Adopting Big Data Analysis in the Agricultural Sector … 153

24. Petrovic, S.: A comparison between the silhouette index and the davies-bouldin index in
labelling ids clusters. In: Proceedings of the eleventh Nordic Workshop of Secure IT Systems,
pp. 53–64 (2006)
25. Brownstein, J., Freifeld, B., Madoff, L.: Digital disease detection: harnessing the web for public
health surveillance. New England J. Med. (2009)
26. Manyika, J., Chui, M., Brown, B., Bughin, J., Dobbs, R., Roxburgh, C., Hung Byers, A.: Big
data: the next frontier for innovation, competition, and productivity (2011)
27. Rindler, A., McClowry, S., Hillard, R.: Big data definition (2012)
28. Cukier, K., Mayer-Schönberger, V.: Big data: a revolution that will transform how we live,
work, and think, p. 22 (2013)
29. Gandomi, A., Haider, M.: Beyond the hype: big data concepts, methods, and analytics (2015)
30. White, T.: Hadoop: the definitive guide (2009)
31. Wolfert, S., et al.: Big data in smart farming. Retrieved from Agricultural Systems (2017)
32. Greengard, Samuel: Weathering a new era of big data. Commun. ACM 57(9), 12–14 (2014)
33. Mining Indonesian tweets to understand food price crisis. Feb 2014. Web. Apr 2017
34. Rosenblatt, F.: The perceptron: a probabilistic model for information storage and organization
in the brain. Psychol. Rev. 65, 386–408 (1958)
35. Riedmiller, M., Braun, H.: A direct adaptive method for faster back propagation learning: the
RPROP algorithm. In: Proceedings of the IEEE International Conference on Neural Networks,
pp. 586–591 (1993)
36. Hecht-Nielsen, R.: Counterpropagation Networks. Appl. Opt. 26, 4979–4983 (1987)
37. Melssen, W., Wehrens, R., Buyden, L.: Supervised Kohonen networks for classification prob-
lems. Chemom. Intell. Lab. Syst. 83, 99–113 (2006)
38. Hopfield, J.J.: Neural networks and physical systems with emergent collective computational
abilities. Proc. Natl. Acad. Sci. 79, 2554–2558 (1982)
39. Kohonen, T.: The self-organizing map. Proc. IEEE. 78, 1464–1480 (1990)
40. Huang, G., Zhu, Q., Siew, C.: Extreme learning machine: theory and applications. Neurocom-
puting 70, 489–501 (2006)
41. Hansen, M., Smith, M., Smith, L., Salter, M., Baxter, E., Farish, M., Grieve, B.: Towards
on-farm pig face recognition using convolutional neural networks. Comput. Ind. 98, 145–152
(2018)
42. Ferentinos, K.: Deep learning models for plant disease detection and diagnosis. Comput. Elec-
tron. Agric. 145, 311–318 (2018)
43. Ju, Y.: Conversion of remote sense picture into data array (2018)
44. Goodfellow, I., Bengio, Y., Courville, A.: Deep learning. MIT Press, pp. 216–261 (2016)
45. Stuart Ward, J., Barker, A.: Undefined by data: a survey of big data definitions (2013)
46. Lukman, E.: Indonesia is social: 2.4% of world’s twitter posts come from Jakarta. Tech in
Asia—Connecting Asia’s Startup Ecosystem, p. 12 (2017)
47. Greenberg, J., Elliott, C.: A cold cut crisis: listeriosis, maple leaf foods, and the politics of
apology. Can. J. Commun. 9, 189–204 (2009)
48. Noyes, K.: Big data poised to change the face of agriculture. Fortune, 02 June 2014 (2017)
49. Kamilaris, A., et al.: Estimating the environmental impact of agriculture by means of geospatial
and big data analysis: the case of catalonia. Science to Society (2018)
50. Sonka S.: Big data and the ag sector: more than lots of numbers. Int. Food Agribus. Manag.
Rev. (2014)
51. About GODAN. About GODAN|GODAN. GODAN (2017)
52. Blue river technology. http://www.bluerivert.com/
53. Cerqueira, R.F., Mantripragada, K.: Estimating rainfall precipitation amounts by applying com-
puter vision in cameras. U.S. Patent No. 9,436,997 (2016)
54. Kumar, N., et al.: Leafsnap: a computer vision system for automatic plant species identification.
Computer Vision ECCV Lecture Notes in Computer, pp. 502–516 (2017)
55. Treinish, L., Praino, A.P.: Applications and implementation of a mesoscale numerical weather
prediction and visualization system. In: Proceedings of the 20th Conference on Weather Anal-
ysis and Forecasting/16th Conference on Numerical Weather Prediction (2004)
154 R. Kashef

56. Zulkarnain, H.: IBM research: using weather modeling and data analytics to save natural
resources (2017)
57. Nowcasting food prices in Indonesia using social media signals (2017)
58. Treinish, A., Anthony, P.: The role of meso-γ-scale numerical weather prediction and visualiza-
tion for weather-sensitive decision making. Proc. Forum Environ. Risk Impacts Soc. Successes
Chall. (2006)
59. Woodie, A.: Farmer’s plant for hyper-local forecasts with IBM’s deep thunder. Datanami. Tabor
Commun. 8 Apr 2017
60. Weston, J.: Support vector machine (and statistical learning theory). Princeton Tutorial. Prince-
ton, USA. Tutorial
61. Zikopoulos, P., Eaton, C.: Understanding big data: analytics for enterprise class Hadoop and
streaming data (2011)
62. Meilă, M.: Comparing clusterings: an information based distance. J. Multivar. Anal. 98(5),
873–895 (2007)
63. Linnainmaa, S.: Taylor expansion of the accumulated rounding error. BIT 16, 146–160 (1976)
64. Broomhead, D., Lowe, D.: Multivariable functional interpolation and adaptive networks. Comp.
Syst. 2, 321–355 (1988)
65. Specht, D.: A general regression neural network. IEEE Trans. Neural Netw. 2, 568–576 (1991)
66. Effland, T., et al.: Discovering foodborne illness in online restaurant reviews. J. Am. Med. Inf.
Assoc. (2018)
67. Proudly Farmers First. farmers business network (2017)
68. Solano, B.: Our story. Farmobile|Press Page. Farmobile, n.d. Web. 13 Apr 2017 http://press.
farmobile.com/
69. Sonka, S.: Big data and the ag sector: more than lots of numbers. Int. Food Agribus. Manag.
Rev. 17(1), 1–20 (2014)
AI-Based Yield Prediction and Smart
Irrigation

Deepak Sinwar, Vijaypal Singh Dhaka, Manoj Kumar Sharma

and Geeta Rani

Abstract This chapter presents different techniques and applications of Artificial

Intelligence for yield prediction and smart irrigation. Timely prediction of irrigation
requirements and crop yields is necessary for farmer’s welfare and satisfaction. The
beforehand prediction significantly contributes to minimizing production cost and
maximizing crop yields. The precise prediction of crops’ yields is also useful for
government, as it is effective in planning various schemes, transport needs, buy-
ing mechanisms, storage infrastructure, and liquid position of the economy before
actual selling of crop by farmers to market. This chapter acknowledges the past break-
throughs and emerging Artificial Intelligence-based techniques in precision farming
specifically for yield prediction and smart irrigation. Artificial Intelligence-based
system provides sufficient information about crop yields at an early stage and its
associated smart irrigation management system is effective in the judicious use of
essential resources such as water and energy for agriculture.

Keywords Yield prediction · Smart farming · Precision agriculture · Smart

irrigation · IoT · Temperature monitoring · Soil moisture · Crop · Artificial
intelligence

1 Introduction

In recent years, an explosion in applications of Artificial Intelligence (AI) specifically

in the domain of health care, weather forecasting, space programmes, automation,
agriculture, etc. has been observed. Across the globe, more than half of the popula-
tion is involved in agricultural practices. In the arena of agriculture, AI is adopted for
the planning of farming activities, sales of agricultural products, weather forecasting,

D. Sinwar (B) · V. S. Dhaka (B) · M. K. Sharma · G. Rani

Department of Computer and Communication Engineering, School of Computing and
Information Technology, Manipal University Jaipur, Dehmi Kalan, Jaipur, Rajasthan, India
e-mail: [email protected]
V. S. Dhaka
e-mail: [email protected]

© Springer Nature Singapore Pte Ltd. 2020 155

P. K. Pattnaik et al. (eds.), Internet of Things and Analytics
for Agriculture, Volume 2, Studies in Big Data 67,
https://doi.org/10.1007/978-981-15-0663-5_8
156 D. Sinwar et al.

and yield prediction. Farmers are taking advantages of AI and getting higher pro-
duction than traditional farming techniques. In traditional farming, cultivation was
manual process irrigational needs of a crop and its yield was highly dependent on
climate factors throughout the globe. The onset of inventions in industrial revolu-
tion and improvement in mechanisms involved in agricultural practices gave relief
to miserable life of farmers. But still, it is a labor-intensive practice and dependent
on mercy of monsoon.
Thus, there is a need of developing a system that automates the agricultural prac-
tices such as cultivation and irrigation, predicts yields, and help in better planning for
systematizing the demand and supply of agricultural products throughout the globe.
A blend of AI and IoT technologies intelligently gathers relevant information about
different crops, their growth rate, and irrigation requirements. These technologies
are effective in monitoring frequent climate change, weather forecasting, nutrition
deficiency in plants, plants’ health monitoring, pest control, weed management, etc.
Researchers have covered a long way in developing subsidiary technology such as
geo-positioning satellites (GPS), satellite imagery and sensors for monitoring of
yield affecting parameters such as moisture, soil pH level, temperature, etc. Target
experimentation at various stages of cultivation such as water resource management,
weed elimination, diseases prediction, pest control, estimation of yield production
and effective storage of agricultural products is still starving. These challenges can
be addressed by applying the techniques of AI and data science.
A voluminous data about agriculture is collected using various sensors such as
remote sensors, proximity sensors, temperature/humidity sensors, etc. Advanced
Internet facilities disseminate this data to various channels such as the cloud. The
introduction of online data storage at cloud, advancements in techniques of data
processing, and increase in computation power become effective in dealing with
challenges identified in agriculture. Use of advanced techniques makes agricultural
predictions more precise, reliable, and useful.

1.1 Scope of AI in Agriculture

Traditional agricultural practices used by farmers are not adequate to fulfill the
increasing demand. To serve this increasing demand, farmers need to adopt the latest
advancements in agriculture such as use of state-of-the-art tools/machinery and AI-
based techniques. It has been observed that most of the farmers are unaware of uses
of pesticides specifically in what amount and when. Due to these practices, fertility
of the soil, as well as yield production, degrades in an intensified way. A plethora
of researchers has proved the increased gain in yield growth with the help of AI and
automation. AI can contribute to agriculture in versatile ways. Authors explain the
inclusion of AI in the following seven agricultural practices.
AI-Based Yield Prediction and Smart Irrigation 157

1.1.1 Weather Forecasting

One of the important advantages of AI in agriculture is weather forecasting. With the

help of latest AI-based state-of-the-art weather forecasting techniques, it becomes
easier for farmers to take appropriate decisions in planning respective crops. Nowa-
days, many weather forecasting devices are available for predicting the weather.
Sehgal et al. [1] presented a visual tool (ViSeed) based on long short-term memory
(LSTM) [2] which can be used for weather and soil predictions. With the help of such
farmer-friendly tools, one can plan his/her crop activities, crop types in an efficient
way.

1.1.2 Smart Irrigation

To cope with the shortage of water, there is a critical need for some smart irrigation
systems that can irrigate more areas with low consumption of water. A comprehensive
review of various techniques of smart irrigation has been presented in Jha et al. [3].
However, there is an availability of various low water consumption-based irrigation
techniques, for example, sprinkler systems and drip irrigation systems; but these
systems need human intervention up to a great extent. There is a scope to add features
to existing systems to develop smart irrigation systems. The system continuously
monitors the level of water in a crop, compares the water content available in soil
and crop plant with standard need of water. It automatically starts sprinklers or drips
as per water requirements of crop. Arif et al. [4] presented ANN-based models for
predicting soil moistures.
A Neural Network-based irrigation system (Neuro-Drip) has been presented by
Hinnell et al. [5]. On the other hand, an IoT-based smart irrigation management
system has been presented by Goap et al. [6], Nawandar and Satpute [7]. AI in this
field has contributed a lot in collaboration with the Internet of Things (IoT). For smart
irrigation, there is a requirement of gathering of information about level of moisture
present in the soil, water content in plants, humidity in atmosphere, temperature, etc.
This information can be gathered using soil moisture sensors, temperature sensors,
humidity monitoring sensors, etc. These sensors are connected to low price Arduino-
based systems for storage of gathered information and executing analysis algorithms
for predicting the water requirement of crop at a particular time.

1.1.3 Crop Disease Prediction and Health Monitoring

Traditional methods of monitoring crop health are time-consuming and inefficient

especially for larger areas of thousands of acres. Many researchers have developed
many AI-based architectures to overcome the challenges identified in the arena of
traditional agriculture. Some architectures are surveyed in Kamilaris and Prenafeta-
Boldú [8].
158 D. Sinwar et al.

AI-based techniques with inclusion of image processing, deep learning, and data
analysis provide an easy and effective way for disease prediction and health moni-
toring. The system captures crop images using high definition(HD) camera-enabled
drones, unmanned aerial vehicles (UAVs), or satellite imagery. The captured images
are used as dataset for training of Convolution Neural Networks (CNNs) [9], a class
of Artificial Neural Network (ANN). CNNs extract useful features from fed images
and make predictions about disease(s) in crops. The system is effective in continuous
monitoring of health of plants and hence gives better solutions for calculating amount
of pesticides to be used and time to use a pesticide.

1.1.4 Crop Readiness Identification

AI-based system captures images of a crop and analyzes them for determining the
crop readiness in a particular area for harvesting. The crops can be categorized into
different categories on the basis of readiness and other quality parameters before
actually sending them to market. For categorization, use of various pattern clustering
techniques viz. K-means, fuzzy C-means (FCM), expectation maximization (EM),
and hierarchical clustering plays an important role.

1.1.5 Yield Prediction

Yield prediction is an area of interest for researchers for past many decades. Yield
prediction requires yield mapping devices, which are still not easily available to
farmers [10]. CNNs can solve the problem of yield prediction in an economic and
easy way. Many researchers [10–12] have developed models to predict crop yields
using AI-based methods which use RGB/normalized difference vegetation index
(NDVI) images. Their experimental results on publically available dataset prove the
usefulness of their models. A comparative analysis of CNN-based models with AI-
based models clearly indicates that CNNs are more advantageous than traditional
Machine Learning (ML) and AI-based techniques [13]. Authors will discuss the
detailed process of predicting crop yields using CNN in Sect. 2.

1.1.6 Weed/Pest Management

Traditionally, weed management was accomplished with a combination of many

techniques viz. mechanical weed control, crop rotation, herbicides, etc. [14]. In
addition to the above, various biochemical-based combined weed and pest control
solutions are also available in the market, but they all results in reducing the yield
productivity. The reduction in crop yield occurs due to the spraying of pesticides
frequently and in a constant amount in a field. AI provides an intelligent solution for
this challenge.
AI-Based Yield Prediction and Smart Irrigation 159

Recently, one AI-based autonomous robot called “Agbots” has been developed by
McAllister et al. [14], which can do weed management in the field in an interesting
manner. The idea behind AI-based weed management is to identify weeds automati-
cally (by examining camera images obtained from autonomous robots) and perform
corrective actions (mechanically weed removal or spraying herbicides) accordingly.
Many other AI-based techniques are also available such as See and Spray (by Blue
River Technologies [15]), which uses AI to identify and spray individual plants in
milliseconds.

1.1.7 Precision Farming/Agriculture

The aim of precision agriculture (PA) is to increase yield production as well as quality
by simultaneously reducing the overall cost and environmental pollution [16]. The
quantity and quality of a crop depend on many parameters such as soil, weather,
irrigation, etc. So, there is a requirement to monitor all these parameters at a regular
interval of time. Traditional monitoring techniques are not adequate in accurate and
efficient monitoring of these parameters.
So, there is vital demand for an automated system which can perform monitoring
of parameters in an effective way. In modern era, proximity sensing and remote sens-
ing dominate the field of agriculture and effectively monitor a plethora of parameters
required for better prediction and planning of agricultural practices. Proximity sens-
ing specifically deals with soil using high-resolution data. Remote sensing provides
the geographical sensing of fields using various sensors. Thermal remote sensing
provides some additional information such as temperature, water status, etc. It also
helps us in getting many vegetation indices. Many researchers attempted to present
several vegetation indices on the basis of various parameters. Xue and Su [17] pre-
sented a systematic reviewed of many vegetation indices. AI in collaboration with
these techniques can provide cost-effective solutions for PA.

2 Yield Prediction

Manual record-keeping and monitoring trends in agricultural activities and yields

are a tedious and time-consuming task. There are fair chances that farmers are not
available all the time to keep track of all agricultural activities. Thus, it results in
collection of incomplete and inconsistent data. AI-based solutions are effective to
overcome such situations. In the era of technology, this is mandatory to motivate
and train farmers for adoption of new tools and techniques. This is highly desirable
to reduce the manual efforts, improve quality of crops, and to meet the increasing
demand of food for increasing population.
Precision agriculture provides an intelligent solution to the challenges observed in
the field of traditional agriculture. Increase in awareness and education make farmers
to think about adoption of precision agriculture. Farmers are willing to adopt new
160 D. Sinwar et al.

strategies of precision agriculture but precise yield prediction is still a challenge due
to shape irregularities, illumination conditions, etc.
Continuous monitoring of each activity involved for a crop from the stage of
sowing seeds till harvesting contributes toward precision agriculture. The planning
of timeline for agricultural activities on the basis of past experience and collected
record is demonstrated in Nevavuori et al. [10].
Farmers are preparing timeline and maintaining record of following parameters
for better prediction.
• Soil fertility and improving fertility by using manures, compost, organic and inor-
ganic fertilizers, bio-fertilizers, etc.
• Growth rate of a crop.
• Irrigation requirements in terms of amount of water required and time and fre-
quency of irrigation.
• Determining precise amount of pesticide and time of spraying pesticide.
• Identifying type(s) of weed growing with a particular crop, growth rate of weed
and determining methods to destroy weed plants without harming the main crop.
• Season wise and crop-wise yield estimation.
• Summarizing timeline for future, based on accumulated knowledge and experi-
ence.

2.1 Machine Learning Techniques for Yield Prediction

Review of related literature reveals the existence of a significant research work in

the field of Machine Learning (ML) techniques for agriculture. The categories of
research work are specifically in yield prediction, disease prediction, crop quality
prediction, yield estimation, price estimation and weed detection and eradication. In
this section, authors give a brief summary of existing works in the domain of yield
prediction using ML-based techniques.
Liakos et al. [18] demonstrate different ML models used for solving real-
world problems. Most commonly used ML models are Artificial Neural Networks
(ANNs) Deep Learning (DL), Support Vector Machine (SVM), Decision Trees (DT),
Bayesian Models (BM), Ensemble Learning (EL), and Dimensionality Reduction
(DR). There are numerous ML techniques available based on these ML models.
These models are based on one or more ML techniques namely Convolution Neu-
ral Networks (CNNs), BackpPropagation Network (BPN), Feedforward Network
(FFN), Hopfield Network (HN), Multilayer Perceptron (MLP), Radial Basis Function
Network (RBF), Deep Boltzmann Machine (DBM), Deep Neural Networks (DNN),
K-Nearest Neighbor (KNN), Linear Discriminant Analysis (LDA), Multiple Linear
Regression (MLR), Generative Adversarial Networks (GAN), and Ensemble Neural
Networks (ENN).
The process of yield prediction using Machine Learning starts with capturing
images of crop. The image capturing process can be carried out by handheld devices
AI-Based Yield Prediction and Smart Irrigation 161

Training Phase
Pre-processing,
Segmentation, Trained
Input Image Neural Network
Feature Model
Extraction, etc.
Re-Training
(if estimated output≠ Ground Truth)

Testing Phase
Pre-processing,
Segmentation, Yield
Input Image Trained Model Prediction
Feature
Extraction, etc.

Fig. 1 Architecture of a general yield prediction system

such as mobile phone cameras, digital cameras, and satellite imagery, etc. After cap-
turing, these images are used for providing training to suitable Neural Networks such
as ANNs, MLP, and CNNs. The trained networks are used for making predictions
in a desired domain. In case there is a mismatch in desired output and ground truth,
then there is a need to retrain the model. The retraining continues until the model
becomes capable of making precise predictions of yields. Figure 1 demonstrates the
architecture of yield prediction using ML.
Let us look at some of the research works carried out by researchers in the field
of crop yield prediction using ML techniques.
Wang et al. [19] developed a computer vision-based system which can automat-
ically estimate the number of apples in an apple orchard. The system is accurate
as well as faster in nature. This system contains a vehicle for moving in the field.
The vehicle dismounted with high precision cameras on metal frames. Both vehicle
and cameras are calibrated accurately according to coordinate frames. Apple regions
are identified with the help of pixel properties. Their model used hue, saturation,
and value for identification of specular reflections. Experimental results and analysis
show that their model is able to calculate the number of apples accurately and rapidly.
Ramos et al. [20] proposed a method for counting the coffee fruits. Their machine
vision system (MVS) is capable of counting and classifying coffee fruit. Harvestable
and non-harvestable fruits were categorized in this work. They captured images
via a mobile-based system and processed the images using linear estimation model
to calculate the number of coffee fruits on a branch. MVS has five stages as: (i)
acquisition, (ii) segmentation, (iii) fruit boundary/outline detection, (iv) adjustment
of ellipses, and (v) detection, classification, and counting of fruits. Promising results
were presented on the basis of experimental analysis and validation strategies.
162 D. Sinwar et al.

Senthilnath et al. [21] presented a system based on Bayesian Information Criterion

(BIC) for detection of Tomatoes in the image. They have used unmanned ariel vehi-
cles (UAVs) for obtaining high-resolution RGB images. For categorizing pixels into
two categories as “tomatoes” and “non-tomatoes,” spectral clustering is used using
K-means, self-organizing maps (SOM), and expectation maximization (EM). Mor-
phological operations are used for segmentation. Their experimental results showed
that the performance of EM was found better than K-means and SOM in the counting
of tomatoes.
Pantazi et al. [12] presented performance comparison of three ML techniques
(counter-propagation ANNs (CP-ANNs), supervised Kohonen networks (SKNs) and
XY-fused networks (XY-Fs)) for Wheat yield prediction based on satellite imagery
and online multilayer soil data. As we know that to predict the crop yield, there is a
need for various parameters. In this system, researchers have used one such parameter
called normalized difference vegetation index (NDVI), mentioned as follows:

NDVI = (NIR − R)/(NIR + R) (1)

where NIR is near-infrared and R is red bands, which enables the accurate estimation
of yield per area. On the other hand, to measure the soil properties, they have used
vis-NIR spectrophotometers. Experimental results consisting of field maps show
that the performance of SKNs was found best among three SOM-based models in
predicting wheat yield crop.
Amatya et al. [22] developed a machine vision-based system for automatic har-
vesting of Cherry fruits because manual harvesting is very labor exhaustive which
causes more than 50% of production cost. A Bayesian classifier is used to classify the
detected branches, leaves, cherry, and background. After getting information about
branches and cherry, we can harvest them accordingly. The system is capable of
detecting branches with an accuracy of nearly 90%.
Sengupta and Lee [23] presented an SVM-based model for detection of immature
green Citrus fruit in varying lighting conditions. The system is able to detect nearly
80% of citrus fruits in the tree canopy. In addition to the above, lots more work in the
area of yield prediction is available. Liakos et al. [18] presented a systematic review
of various Machine Learning techniques for agriculture. They have classified these
techniques into different categories such as yield prediction, disease detection, weed
monitoring, etc.

2.2 Remote Sensing-Based Yield Prediction

Remote sensing brings information about a field, crops, or an object without actually
visiting the field. The information is captured with the help of different sensors and
is made available in raw, processed, or analyzed form. It works on the principle of
properties of objects. The properties may be chemical, physical, structural, energy
AI-Based Yield Prediction and Smart Irrigation 163

Data Capturing Data Information

Interpretation Production (Graphs, Decision
(Through Satellites, (Processing/ Statistics, Maps, GIS Making
UAVs, Sensors etc.) Filtering etc.) datasets etc.)

Fig. 2 The Remote Sensing process

Table 1 Various active and passive sensors for remote sensing [24]
Active sensors Passive sensors
Laser altimeter Accelerometer
Lidar Hyperspectral radiometer
Radar Imaging radiometer
Ranging instrument Radiometer
Scatterometer Sounder
Sounder Spectrometer
Spectroradiometer

emissions, etc. Sensors sense these properties and send to a storage system. In stor-
age systems, the information is organized in a structured way. This information is
analyzed to make decisions for deciding further actions required. Remote sensing
techniques have numerous applications in agriculture viz. yield forecasting, damage
identification, harvesting time prediction, irrigation estimation, and nutrient require-
ment prediction, etc. Figure 2 shows the process of remote sensing. It comprises of
four basic steps viz. data capturing, data interpretation, information production, and
decision making.

2.2.1 Data Capturing

The process of capturing data is done in either of two ways, i.e., active or passive. In
the case of active remote sensing, sensors are focusing directly on objects and send
the sensed information to the control station which is located on the ground. Active
sensors make uses of their own energy sources for object illumination.
Passive remote sensing follows the concept of reflection. Objects are exposed
to sunlight and then reflected wavelengths are sensed. The sensed information is
further provided to control stations. The electromagnetic spectrum is being used for
this purpose. List of some active and passive sensors obtained from [24] is given in
Table 1

2.2.2 Data Interpretation

After capturing the required data, the process of processing that data needs to be
done in order to obtain the required information. This processing may take a while
164 D. Sinwar et al.

depending upon the operations/statistical techniques being applied. This process is

intended also for filtering of only relevant data which needs to be considered for the
next phases. Various mathematical/statistical techniques are being applied in order
to get the required information.

2.2.3 Information Production

The next step after interpreting data is the production of the required information. As
we know that the processed data can be presented in the form of various visual rep-
resentations such as graphs, charts, maps, GIS data, etc. We can get various types of
information with the help of remote sensing data such as yield modeling/forecasting,
nutrient deficiency graphs, area estimation, soil mapping, etc. These graphs/charts
are the best way to represent the gathered information, by which we can take effective
and timely decisions.

2.2.4 Decision Making

With the help of various visual representations, one can take effective decisions. The
objective of these decisions is very clear: maximizing the net profit by increasing the
quality as well as quantity of the crop. Also, there is a need to identify risk factors in
achieving this objective. Nowadays, the process of decision making becomes easier
with the help of various AI-based techniques. Once we have obtained the required
information, we can present this information to the information scientists in order
to save them in the knowledge base. On the basis of certain parameters AI/Machine
Learning (ML) systems are able to predict/provide efficient decisions (such as the
necessity of water treatment, moisture prediction, irrigation requirements, pesticide
requirements, etc.) which makes farmers away from critical thinking process and
obtaining maximum profits out of minimum/required efforts. Traditional ML tech-
niques use feature extraction in the earlier stages, and on the basis of these features,
different tasks (crop classification, yield prediction, etc.) were carried out [10]. But
traditional ML techniques were not sufficient in producing the optimal decisions.
Nowadays, it becomes possible to provide optimal decisions with the help of some
advanced paradigms such as Deep Learning paradigm. Among various Deep Learn-
ing paradigms, Convolutional Neural Networks (CNNs) prove to be efficient in image
analysis/classification. By combing Remote Sensing with Deep Learning techniques,
we can have effective agricultural decisions that can maximize the farmer’s profit.

2.3 Big Data-Based Yield Prediction

With the advancement of Information and Communication Technology (ICT), mas-

sive volumes of data are being generated. Big data techniques play important roles in
AI-Based Yield Prediction and Smart Irrigation 165

analyzing these massive volumes for smart farming and yield prediction. The process
of big data starts by implementing “data chain,” which is the organized sequence of
activities from capturing to decision making/marketing [25] mentioned as follows:
1. Capturing of data (Identifying sources of data, types, and purposes)
2. Storage requirements (Identifying storage location, permissions/authenticity,
etc.)
3. Transfer from one location to another
4. The transformation from one form to another
5. Analysis (Identifying methods, purposes, and output)
6. Decision Making (Identifying strategies, risk assessment, etc.)
Big data can play important roles in predicting the amount of yield prediction.
Here, the emphasis is on analysis rather than big or small. An agricultural data
without proper analysis is of no use until it is being analyzed properly. However,
big data in agriculture is still in an early stage, and there is a need to do more work
on big data with respect to agricultural applications. As mentioned above in the first
step, we may require the capturing of data from multiple sources. If data is gathered
from multiple sources, definitely it will not be in a uniform shape. It is the task of
data analysts to prepare and arrange data in a uniform shape which in future can be
transferred/transformed.
There are numerous data analysis techniques [26] available, but for analyzing
agricultural data, we have to focus on some specific techniques such as scalable
vector machines, K-means, NDVI, Wavelet/Fourier transformations, etc. In addi-
tion, to predict the crop yield, big data can do more with agricultural data such as
weather forecasting, soil/land analysis, food availability, etc. In this regard, Kamilaris
et al.[26] presented a systematic yet remarkable review of various big data techniques
for agriculture.

3 Smart Irrigation Systems

Efficient utilization of water is a challenging task especially for places where avail-
ability of water is a major concern. For gaining high yields, one has to irrigate the
field when needed or we can say that the exact amount and time of irrigation must
be known. Nowadays with the help of new innovations in technology, we are liv-
ing in the world of advanced irrigation systems called “smart irrigation.” The word
smart means, the sensors are able to sense the water requirements in plants. This
ability is achieved by combining multiple technologies viz. automation, sensors,
and knowledge (AI). Even though these systems are not costly in nature, due to
lack of awareness, most of the farmers are adopting a traditional way of irrigation
which results in inefficient utilization of water and low crop yield. Thanks to the
researchers who have developed such low-cost smart irrigation systems. One such
system is shown in Fig. 3.
166 D. Sinwar et al.

Fig. 3 AI-based smart irrigation system [35]

The automation behind these smart irrigation systems is achieved using smart
embedded systems. These smart embedded systems can be formed with the help of
various technologies/devices such as Raspberry Pi, Arduino, power unit, temperature
sensors, moisture sensors, Machine Learning, and Internet of Things (IoT) as shown
in Fig. 4.
Jha et al. [3] presented a review of such irrigation systems in a systematic way.
Smart irrigation systems are able to save a massive amount of water, which can be
utilized for other important purposes of mankind.
Smart irrigation systems also ensure the reachability of water to each and every
plant in an exact amount which results in maintaining the good health (by pre-
venting them from dehydration and excessive/irregular irrigation) of the plant and
makes away from diseases. It has been observed that some of the diseases in plants
are due to improper irrigation which is mostly due to the adaptation of traditional
irrigation systems. Continuous observations of sprinklers/drip modules are also nec-
essary at regular intervals to prevent them from failures. Sometimes these systems
need recalibrations due to degradation in performances or damages caused by ani-
mals/environment. In this situation, some part of the field is highly irrigated while
some part remains unirrigated. The main challenge in designing a smart irrigation
system is the nonuniformity of land and crop types. Some of the smart irrigation
systems are discussed in the upcoming section.
AI-Based Yield Prediction and Smart Irrigation 167

Fig. 4 Various devices for smart implementing smart irrigation systems: a Raspberry Pi, b Arduino,
c soil moisture sensor, d temperature sensor

3.1 Existing Smart Irrigation Systems

Many researchers work in the area of smart irrigation systems so far. Let us discuss
some of them as follows:
Suresh and Umasankar [27] developed a smart irrigation system which works on
the solar power supply. The system is able to minimize electricity costs by utilizing
solar power. It is composed of an automated water flow system which works by
sensing the moisture level in the soil.
The intention behind the development of this system is to save both electricity
as well as water in the Indian scenario. The block diagram of solar-powered smart
irrigation is given in Fig. 5, which consists of solar panels, battery, converter mod-
ule, and water pump. The system works on solar-powered supply, which is used to
turn on the stepper motor to feed the water from well to storage tank on the basis
of sensed moisture. The storage tank is mounted with a water outlet valve which
is electronically controlled by moisture sensing circuitry. The moisture sensors are
mounted inside the field of where cultivation of the crop is going on. Sensors are
able to convert the moisture content to equivalent voltage which is compared with the
168 D. Sinwar et al.

Solar Panel Converter Water Pump

Battery Actuator Storage Irrigation

Tank

Moisture
Sensor

Fig. 5 Architecture of a solar-powered smart irrigation system [27]

reference voltage. The voltage difference level is used to find out the water require-
ments. Their experimental results showed that the system is capable of irrigating the
field with the minimum usage of water and energy.
Kehui et al. [28] designed a system using wireless sensor network for monitoring
of moisture and water height in the rice field. The system was tested on real time and
proved to be feasible for irrigating rice fields. The working of the system is explained
as follows:
1. Moisture data is calibrated to the data center.
2. The existing expert data available with the data center is being compared with
real-time sensed data.
3. Irrigation instructions are being sent to the base station on the basis of irrigation
requirements.
4. The base station sends these requirements to the irrigation control system (ICS).
5. ICS opens the electronic valve and closes after irrigation.
6. The process works in a cycle as per the irrigation requirements.
Vellidis et al. [29] developed a prototype for scheduling irrigation to cotton fields
on the basis of soil moisture and soil temperature as shown in Fig. 6. This smart
irrigation system works on the principle of smart nodes consisting of a combination
of a sensor array, RFID tag for sending information to the central computer, and
electronic circuitry. The tags used in these nodes have the capacity to transmit signals
up to the range of 0.8 km. The smart nodes will send the sensed information at
user-specified intervals. The information in each transmission is the unique node
identification and 12 bytes of sensor values (two temperature values and three sensor
values). The nodes are provided with a power backup of 9 V lithium batteries. The
microcontrollers are programmed for efficient usage of power during information
transmission and sensing moisture/temperature levels. The system was tested in a
real environment which offered real-time monitoring of water in the soil. According
to these values, users are able to schedule the irrigation requirements (which were
not automated in this system).
Nasiakou et al. [30] presented a software system for reducing the irrigation cost
by combining the smartness of renewable energy sources and irrigation as shown
in Fig. 7. The system contains common information and common bus which is
AI-Based Yield Prediction and Smart Irrigation 169

Fig. 6 Smart irrigation system [29], a smart sensor array for cotton field, b modified sensor array
for peanut

Fig. 7 Component view of energy-efficient irrigation system [30]

170 D. Sinwar et al.

connected with six components as irrigation modules, energy-consuming/producing

devices, power distribution system, energy markets, and auxiliary modules.
The main goal of designing this system is analyzing the capabilities of batteries
and renewable energy sources to handle both residential as well as agricultural load.
As shown in Fig. 7, the six basic components (irrigation module, power distribu-
tion module, energy consumption devices, energy-producing devices, energy mar-
kets, and auxiliary module) of their systems are connected to a common bus called
information and communication bus. Every module has its designated purpose in
contributing to the irrigation system. Power distribution module is responsible for
distributing electricity flow to the whole system. It calculates the overall energy flow
in terms of the voltage vector of different nodes. Electricity generating module is
responsible for implanting various energy-producing devices, i.e., engines, turbines,
PV panels, and batteries. Energy consumption module on the other hand provides
possible energy-consuming devices which can be categorized into two categories,
viz. residential/household devices and commercial devices. The energy market is
responsible for organizing energy bids between producers and consumers. The main
module of the system, i.e., irrigation module is capable of irrigating plants in two
different ways, viz. automated and passive. All other components of the system such
as temperature sensors, soil moisture sensors, humidity sensors are implemented in
the auxiliary module of the system. Simulation of the system was achieved on R4
25:00 1 and IEEE-13 bus feeders equipped with 68 and 80 houses, respectively. The
simulation gives promising results in terms of satisfying the irrigation requirements
of both agricultural and residential design with low energy consumptions.
Navarro-Hellín et al. [31] presented a smart irrigation decision support system
(SIDSS) for estimating the weekly irrigation requirements using sensors and Machine
Learning techniques (PLSR and ANFIS). The system was tested on citrus plants
located in the southeast of Spain. Performance of the system was tested with human
expert values. A reduction of 22% irrigation requirements was observed using their
system as compared with weather values of the previous year (i.e., irrigation require-
ments of the year 2015 are predicted on the basis of 2014).
Figure 8 shows the prototype of the closed-loop irrigation system which is able to
send feedback to a typical irrigation advisor system (shown in dashed box). Earlier,

Weather Crop/ Soil Crops

Farmer
Information Characteristics

Expert Agronomist Irrigation

Report

Soil Sensors

Fig. 8 Closed-loop irrigation advisor system [31]

AI-Based Yield Prediction and Smart Irrigation 171

the decisions were taken by expert agronomist on the basis of weather information and
characteristics of soil/crop. The expert’s knowledge is being used in their proposed
system, which incorporates two Machine Learning algorithms along with expert
knowledge to take irrigation decisions as shown in Fig. 9. They have developed
their own wireless module (as shown in Fig. 10) for collecting information about
soil/crop from a lemon field. The device is equipped with GPRS/GSM modem for
accessing it from anywhere. The advantage of these modules is that they are powered
with a solar power system and are capable of providing sufficient information for
decision support systems. The collected information is being stored in the databases
for making irrigation decisions along with previous knowledge of experts using
Machine Learning algorithms.
Weather stations, on the other hand, are sending the following information:
Relative humidity (RH), temperature (T), rainfall (RF), wind speed (WS), vapor
pressure deficit (VPD), dew point (DP), global radiation (GR). This information

Training Inputs Decision

Support System
Weather
Information ANFIS
Predictive Training Targets
Model Irrigation
Soil Sensors Reports
PLSR
Crop and Soil Predictive
Characteristics Model

Fig. 9 SIDSS with training inputs and targets [31]

Fig. 10 Wireless information collection device with GPRS/GSM modem installed in the southeast
of Spain for the lemon crop by Navarro-hellín et al. [31]
172 D. Sinwar et al.

is made available to the public on the SIAM website [32]. The goal of SIDSS is
to take final decisions on the basis of information collected from different sensors
and expert’s knowledge regarding water requirements or time requirements for con-
stant water irrigation. They have incorporated two Machine Learning techniques,
viz. partial least square regression (PLSR) and adaptive neuro-fuzzy inference sys-
tem (ANFIS). PLSR is a statistical model which is needed for pursuing knowledge
between predictor and response variables, whereas ANFIS is capable of generating
fuzzy inference rules on the basis of given input/output dataset. The system was
tested to three different locations and found to be suitable for scheduling irrigation
requirements on the basis of continuous monitoring of soil/crop characteristics along
with expert’s knowledge.
Recently, Goap et al. [6] developed an IoT-based solution for smart irrigation using
open-source technologies. The system is fully functional which takes advantages of
its integration with the cloud using web-based services. The system works on the
principle of sensed values (i.e., soil moisture, ultraviolet radiations (UV), soil/air
temperature, and humidity) of from sensor nodes. The proposed system is intelligent
enough in providing real-time irrigation decisions and web-based visualizations. The
block diagram of their system is shown in Fig. 11.
Figure 11 shows the architecture of IoT-based smart irrigation system. The system
is embedded with many components which are divided into three main units, viz.
sensor network, server, and application layer. Sensor nodes are continuously sensing

Fig. 11 The architecture of IoT-based smart irrigation system using open-source technologies [6]
AI-Based Yield Prediction and Smart Irrigation 173

the required values from the fields and sending them to the data processing server
for storing them to a centralized database system. The centralized database along
and other web-based data are collectively used for making irrigation decisions on
the basis of an intelligent algorithm. The algorithm makes use of both supervised
and unsupervised techniques which gives better accuracies while minimizing the
mean squared error for predicting the soil moisture level of upcoming days. Values
obtained from sensor units are being supplied to support vector regression (SVR)
(modified SVM) model for providing training. After getting training, SVR is used to
make predictions on soil moisture differences (SMD). For improving the accuracy
of predictions, K-means clustering is used. The final predicted values are used for
making irrigation decisions. The system gave promising results while tested on real
environment.
Gu et al. [33] presented a software-based solution called Root Zone Water Quality
Model (RZWQM) for scheduling irrigation requirements. The system is able to iden-
tify irrigation requirements on the basis of crop computed on the basis of agricultural
models because it seems harder to calculate this stress in real time. Irrigation timings
are suggested on the basis of water stress (WS) calculations by model, whereas irri-
gation depth is calculated on the basis of soil moisture threshold. The model captures
weather inputs from databases of weather websites (historical and four days ahead
of data) and on-site sensors. Water requirements for irrigations are calculated on the
basis of Eq. (2).
⎧ ⎫
⎨N
  ⎬
IR(t0) =K θ( f ci) − θ(t(0)i) · D(i) − P(t0+4d) (2)
⎩ ⎭
(i=1)

where
IR(t0) is the required water supply for irrigation
i represents layers of soil
N represents the deepest rooting depth derived from crop biomass
θ (fci) represents volumetric θ at the field capacity of ith layer of root
(t (0)i ) represents volumetric θ on the irrigation day at ith layer of root
K represents the proportion of irrigation losses which needs to be done again
P(t0+4d) represents expected accumulative rain and data of the next four days.
All existing irrigation calibrations are being removed and RZWQM is then recali-
brated on the basis of calculated WS. GUI of RZWQM irrigation scheduling software
is shown in Fig. 12. The system is able to save the water requirement with a trivial
reduction (0.03–3.81%) in crop yield. Experimental results showed that the system
is able to save water and maintain the crop yields.
Nawandar and Satpute [7] proposed an IoT and Neural Network-based low-cost
smart irrigation system. IoT is used for connecting various devices, whereas Neural
Networks are used for irrigation decision-making purposes.
The system is user-friendly in the nature that it provides various crop-related
details like estimated irrigation requirements, soil statistics, zone-wise irrigation
alert, and remote data monitoring through message queue telemetry transport
174 D. Sinwar et al.

Fig. 12 The interface of RZWQM scheduling software [33]

(MQTT) protocol. The architecture of the proposed low-cost smart irrigation system
is shown in Fig. 13.
The system consists of a unified sensor pole (USP) which is called as brain of this
system. The purpose of including USP is to take intelligent decisions about irrigation
process. USP senses real-time data (i.e., temperature, humidity, and soil moisture)
from smart nodes and sends them to both irrigation unit (IU) and sensor information
unit (SIU). SIU is capable of logging sensor data to website through 802.11x. A

Fig. 13 Architecture of IoT and Neural Network-based low-cost smart irrigation system [7]
AI-Based Yield Prediction and Smart Irrigation 175

Fig. 14 Installation of IoT-based smart irrigation system [7]: a plant, b moisture sensor, c humidity
and temperature sensor, d USP

HTTP server hosts these data to a webpage for remote monitoring. Installed module
is shown in Fig. 14, which costs ≈$13. USP is responsible for providing sensed
values at various regular intervals. The position of smart node is decided on the
basis of water availability. Low water availability spaces cause massive logging of
same data at regular intervals. So, the area which is having average availability of
water is chosen for the installation of these sensors, because it is able to sense the
water requirements of nearby areas. IU on the other hand is responsible for making
intelligent decisions about irrigation requirements. USP has the capability to work
in either admin or user mode. It works in admin mode until it times out. USP is
connected with irrigation unit which instructs stepper motor via relay to start or stop.
MQTT broker sends sensor values through messages in an energy-efficient way.
Examples of some famous MQTT brokers are Mosca, RabbitMQ, HiveMQ, Cloud-
MOTT, etc. The principle of sending message is publish–subscribing mechanism, in
which sensor nodes are publishing data to MQTT on topics. Subscribes on the other
hand are end users, which use the published topics. For taking intelligent irrigation
decisions, a Neural Network is provided training of real-time data as well as crop
database inputs. The system showed promising results after installation. Figure 15
shows the crop growth after installation.
176 D. Sinwar et al.

Fig. 15 Crop growth after

installation of IoT-based
low-cost smart irrigation
system [7]

3.2 Architecture of Proposed Smart Irrigation

Systems/Automated Irrigation Infrastructure

As discussed in the previous section, various smart irrigation systems have been
proposed by many researchers which are overcoming the drawbacks of traditional
irrigation systems (drip irrigation, direct pump/canal irrigation, sprinkler irrigation,
tube well irrigation) that need continuous monitoring of field. Most of these smart
irrigation systems are capable of automated irrigation of crop field on the basis of sen-
sor values. Sensor nodes are the most essential part of these systems. Other important
components include but not limited to: controller unit, wireless medium, GSM mod-
ule, Arduino, display unit, relay, Raspberry Pi, water pump, sprinklers/drip modules,
etc. The difference in these systems arises due to two things, viz. variety/capability of
sensor nodes and their organization with other components. Temperature sensors are
responsible for observing the amount of temperature adequate for a particular crop,
whereas moisture sensors are capable of sensing level of moisture in soil. It is the
responsibility of system designer to obtain the values from these sensors continuously
or at some specified intervals. Some systems were found to be energy efficient in
the sense that wireless nodes must not sense and send the information continuously;
rather, they were calibrated to do these tasks at user-specified interval thresholds.
A controlled unit is continuously observing these sensed values and enables water
pump operations (On and Off). The positioning of sprinklers/drip modules should
be mounted specifically according to the plant/crop and land structure. There is no
general irrigation system available which can satisfy the irrigation requirement of
all types of crops/fields, because the structure and irrigation needs are not same for
all crops. Considering these versatilities in the field/crop, intelligent irrigation sys-
tems should be capable of making decisions on real-time sensed values as well as
AI-Based Yield Prediction and Smart Irrigation 177

Solar Power Cloud

Smart valve/
Smart Sensor Arduino Water Pump Smart water meter
nodes

Previous Sprinklers/
Raspberry pi Drip irrigation
Knowledge

Fig. 16 The architecture of the proposed smart irrigation system CS-HYSIS

analyzing the previous knowledge. After observing various irrigation systems in the
previous section, an architecture of a smart irrigation system named cloud and solar
power enabled hybrid smart irrigation system (CS-HYSIS) has been proposed as
shown in Fig. 16.

3.2.1 Working Principle

As shown in Fig. 16, the system contains smart sensor nodes which are provided
power supply from solar panels. The choice of solar panels may vary according to
the weather conditions (if the system is being installed in a cloudy/rainy environment,
definitely advanced controllers for solar panels need to be installed). The sensor nodes
contain three sensor units mentioned as follows:
• Temperature sensor
• Soil moisture sensor
• Humidity sensor
These smart sensor nodes are directly connected to the Arduino controller. The
controller enables water pump to turn on or off on the basis of real-time sensor values
and existing knowledge (trained inside a Neural Network) from Raspberry Pi. The
concept of existing knowledge has been incorporated in this system to irrigate the
plants during unavailability/failure of sensor nodes. Because sensor nodes may stop
working due to various reasons, i.e., power failure, fault in a part of sensor, physically
damaged by any animal(s), etc. This feature prevents plants from dehydration in
case of failure of smart sensor nodes. Existing knowledge can be obtained from
previous data or from agricultural expert. Pretrained network plays an important role
in providing overall predictions about irrigation requirements. The controller is able
to make decisions either on the basis of real-time values (obtained from smart sensor
units)/pretrained network or from both. It analyzes both existing pieces of knowledge
and real-time values in order to make appropriate decisions (turning on or off the
water pump) on the basis of user-specified thresholds (i.e., moisture level, humidity
level, and temperature).
178 D. Sinwar et al.

The most important part of this system is that it is able to send all irrigation-related
data to the cloud, for making analysis and future purposes. User can see the current
and past irrigation usages anytime on a smart web-based or app-based portal. The
interface may provide several additional features, i.e., start/stop irrigation, total water
consumptions in a particular amount of time/area, real-time sensor values, etc. Smart
valve/smart meter will count the units of water consumptions and send these details
to the cloud. With the help of this methodology, we can make sense about total water
consumption/requirements for a particular crop and monthly/weekly consumption
of water. These values can be used for making future plans/decisions about water
requirements.
The position of sprinklers and drip modules is calibrated manually according to
the situations of different crops/fields. Nowadays, the sprinklers are so smart in the
nature that they can itself irrigate the field smartly. Traditional sprinklers are not
able to irrigate the fields properly. Wi-Fi enabled sprinklers provide the facility for
controlling the sprinkler nozzles using smartphones. The direction of the nozzles
and speed of water spraying can be controlled through smartphones. Even some
smart sprinklers are able to make their own irrigation schedules as well. For more
information on smart sprinklers, please refer to the review of latest sprinklers which
is provided in Top 8 Best Smart Sprinkler Review [34]. The choice of sprinkler
systems is also challenging for implementing smart irrigation systems in order to do
optimized irrigation.
The proposed system is able to achieve precision agriculture by doing optimized
irrigation and can provide several advantages as follows:
• Cost minimization:
– Optimum utilization of energy (obtained through solar power systems
– Optimum utilization of water (sprayed through smart sprinklers as per require-
ments).
• Increase in crop growth:
– Achieving good health of plants by providing them optimized irrigation in order
to prevent them from dehydration and excessive/irregular irrigation.
• Reduction in manpower requirements for irrigation.
• Providing complete irrigation data through cloud computing.
• Efficient utilization of existing knowledge through Artificial Intelligence.

4 Conclusion

Yield prediction and smart irrigation systems using AI-based techniques are demand-
ing in precision agriculture. This chapter examines the critical parameters for predict-
ing both crop yields and irrigation requirements in a systematic way. After carefully
AI-Based Yield Prediction and Smart Irrigation 179

examining many existing systems, we have provided effective prototypes and solu-
tions to meet the expectations of the current scenario. It has been observed that
AI-based techniques play a major role along with other hardware components such
as Raspberry Pi, soil moisture sensors, temperature sensors, humidity sensors, etc.
in predicting the crop yields and irrigation requirements. If both yield prediction and
smart irrigation systems are well equipped with AI-based technologies, then these
can prove the effectiveness in minimizing the overall agricultural cost, increasing
growth in the economy and minimizing wastage of essential resources such as water,
energy. These techniques are effective in reducing human efforts and fasten the plan-
ning of agricultural practices. But minimizing the cost of system deployment and
training of uneducated and non-techno-savvy farmers is still challenging.

References

1. Sehgal, S., et al.: Crop planning using stochastic visual optimization, 2017. IEEE Visual. Data
Sci. VDS 2017, 47–51 (2018)
2. Hochreiter, S., Schmidhuber, J.: Long short-term memory. Neural Comput. 9(8), 1–32 (1997)
3. Jha, K., Doshi, A., Patel, P., Shah, M.: A comprehensive review on automation in agriculture
using artificial intelligence. Artif. Intell. Agric. 2, 1–12 (2019). https://linkinghub.elsevier.com/
retrieve/pii/S2589721719300182
4. Arif, C., Mizoguchi, M., Mizoguchi, M., Doi, R.: Estimation of soil moisture in paddy field
using artificial neural networks. Int. J. Adv. Res. Artif. Intell. 1(1), 17–21 (2013)
5. Hinnell, A.C., et al.: Neuro-drip: estimation of subsurface wetting patterns for drip irrigation
using neural networks. Irrig. Sci. 28(6), 535–544 (2010)
6. Goap, A., Deepak Sharma, A.K., Shukla, Rama Krishna, C.: An IoT based smart irrigation
management system using machine learning and open source technologies. Comput. Electron.
Agric. 155(May), 41–49 (2018). https://doi.org/10.1016/j.compag.2018.09.040
7. Nawandar, N.K., Satpute, V.R.: IoT based low cost and intelligent module for smart irrigation
system. Comput. Electron. Agric. 162, 979–90 (2019). https://doi.org/10.1016/j.compag.2019.
05.027
8. Kamilaris, A., Prenafeta-Boldú, F.X.: Deep learning in agriculture: a survey. Comput. Electron.
Agric. 147, 70–90 (2018)
9. Lecun, Y., Bottou, L., Bengio, Y., Haffner, P.: Gradient-based learning applied to document
recognition. Proc. IEEE 86(11), 2278–2324 (1998)
10. Nevavuori, P., Narra, N., Lipping, T.: Crop yield prediction with deep convolutional neural
networks. Comput. Electron. Agric. 163(April), 104859 (2019). https://linkinghub.elsevier.
com/retrieve/pii/S0168169919306842
11. Crane-Droesch, A. Machine learning methods for crop yield prediction and climate change
impact assessment in agriculture. Environ. Res. Lett. 13(11) (2018)
12. Pantazi, X.E., et al.: Wheat yield prediction using machine learning and advanced sensing
techniques. Comput. Electron. Agric. 121, 57–65 (2016). https://doi.org/10.1016/j.compag.
2015.11.018
13. You, J., et al.: Deep gaussian process for crop yield prediction based on remote sensing data.
31st AAAI Conference on Artificial Intelligence, AAAI 2017: pp. 4559–4565 (2017)
14. McAllister, W., Osipychev, D., Davis, A., Chowdhary, G.: Agbots: weeding a field with a team
of autonomous robots. Comput. Electron. Agric. 163(September 2018), 104827 (2019). https://
doi.org/10.1016/j.compag.2019.05.036
15. Chostner, B.: Smart Machines for Weed Control and Beyond (2017)
180 D. Sinwar et al.

16. Chlingaryan, A., Sukkarieh, S., Whelan, B.: Machine learning approaches for crop yield pre-
diction and nitrogen status estimation in precision agriculture: a review. Comput. Electron.
Agric. 151(May), 61–69 (2018). https://doi.org/10.1016/j.compag.2018.05.012
17. Xue, J., Su, B.: Significant remote sensing vegetation indices: a review of developments and
applications. J. Sens (2017)
18. Liakos Konstantinos, G., et al.: Machine learning in agriculture: a review. Sensors (Switzerland)
18(8), 1–29 (2018)
19. Wang, Q., Nuske, S., Bergerman, M., Singh, S.: Automated crop yield estimation for apple
orchards. Exper. Robot. Springer Tracts Adv. Robot. 88, 745–758 (2013). http://link.springer.
com/10.1007/978-3-540-77457-0
20. Ramos, P.J., Prieto, F.A., Montoya, E.C., Oliveros, C.E.: Automatic fruit count on coffee
branches using computer vision. Comput. Electron. Agric. 137, 9–22 (2017). https://doi.org/
10.1016/j.compag.2017.03.010
21. Senthilnath, J., et al.: Detection of tomatoes using spectral-spatial methods in remotely sensed
rgb images captured by UAV. Biosys. Eng. 146, 16–32 (2016). https://doi.org/10.1016/j.
biosystemseng.2015.12.003
22. Amatya, S., et al.: Detection of cherry tree branches with full foliage in planar architecture for
automated sweet-cherry harvesting. Biosys. Eng. 146, 3–15 (2015). https://doi.org/10.1016/j.
biosystemseng.2015.10.003
23. Sengupta, S., Lee, W.S.: Identification and determination of the number of immature green
citrus fruit in a canopy under different ambient light conditions. Biosyst. Eng. 117(C), 51–61
(2014). http://dx.doi.org/10.1016/j.biosystemseng.2013.07.007
24. NASA. Remote Sensors. EARTHDATA powered by EOSDIS. https://earthdata.nasa.gov/learn/
remote-sensors, 9 Aug 2019
25. Wolfert, S., Ge, L., Verdouw, C., Bogaardt, M.: Big data in smart farming—a review. Agric.
Syst. 153, 69–80 (2017). https://doi.org/10.1016/j.agsy.2017.01.023
26. Kamilaris, A., Kartakoullis, A., Prenafeta-Boldú, F.X.: A review on the practice of big data
analysis in agriculture. Comput. Electron. Agric. 143(September), 23–37 (2017). https://doi.
org/10.1016/j.compag.2017.09.037
27. Harishankar, S., Umasankar, V.: Solar Powered Smart Irrigation System (September 2015)
28. Kehui, X., Xiao D., Luo X.: Smart water-saving irrigation system in precision agriculture based
on wireless sensor network, pp. 170–75 (2010)
29. Vellidis, G., et al.: A real-time wireless smart sensor array for scheduling irrigation. Comput.
Electron. Agric. 61, 44–50 (2008)
30. Nasiakou, A., Vavalis, M., Zimeris, D.: Smart energy for smart irrigation. Comput. Electron.
Agric. 129, 74–83 (2016). https://doi.org/10.1016/j.compag.2016.09.008
31. Navarro-hellín, H., et al.: A decision support system for managing irrigation in agriculture.
Comput. Electron. Agric. 124, 121–131 (2016)
32. SIAM: agricultural information system of Murcia (2019). http://siam.imida.es/apex/f?p=101:
1:138900487927405
33. Gu, Z., et al.: Development of an irrigation scheduling software based on model predicted crop
water stress. Comput. Electron. Agric. 143(January), 208–221 (2017). https://doi.org/10.1016/
j.compag.2017.10.023
34. Top 8 Best Smart Sprinkler Review (2019). https://www.helpadvisors.org/best-smart-sprinkler-
systems/, 22 Aug 2019
35. Microsoft Cloud Perspectives (2018). Feeding the World with AI-driven Agriculture Innovation
IoT-Based Smart Tree Management
Solution for Green Cities

Bilal Shabandri, Sahith Reddy Madara and Piyush Maheshwari

Abstract This paper discusses key challenges faced by transformation of smart

cities and integration of green technologies, initiatives, and existing systems that are
in place to solve for these challenges. With overpopulation, fast urbanization, and
environmental changes, powerful administration of trees and plants is required like
never before. We present an Internet of things (IoT)-based smart tree management
architecture for cities which is able to track various characteristics of a given tree
such as the air quality, sunlight level, sound pollution level, and other important char-
acteristics for efficient city planning, afforestation plans, and higher quality produce.
We divide the solution into three sections: The Sensor Node, The Cloud Integration,
and User Experience. We have developed a working prototype using a Raspberry Pi
and a mobile application consisting of an intuitive user interface that lets users and
farm owners get information about a certain tree and view its information and alerts.
We utilize low-power wide-area network modules for the connectivity of the various
sensors to the cloud. A map view is also developed which consists of various fil-
ters that let city authorities understand the distribution of healthy and damaged trees
across the city accompanied by an extensive customized dashboard. A prototype of a
green smart tree management system based on the Internet of things is discussed and
implemented. The paper describes the functions and actual characteristics of each
system component in detail. The experimental results suggest the feasibility of the
application to enable the development of smart cities across the world which will play
a key role in decreasing pollution levels across the city and increasing afforestation

B. Shabandri
Department of Electronics and Telecommunication,
Amity University Dubai, Dubai, United Arab Emirates
e-mail: [email protected]
S. R. Madara (B)
Sharjah Academy for Astronomy, Space Sciences and Technology,
University of Sharjah, Sharjah, United Arab Emirates
e-mail: [email protected]
P. Maheshwari
Department of Computer Science and Engineering,
Amity University Dubai, Dubai, United Arab Emirates
e-mail: [email protected]
© Springer Nature Singapore Pte Ltd. 2020 181
P. K. Pattnaik et al. (eds.), Internet of Things and Analytics
for Agriculture, Volume 2, Studies in Big Data 67,
https://doi.org/10.1007/978-981-15-0663-5_9
182 B. Shabandri et al.

in areas that require it the most. This solution can be useful in spreading awareness
among citizens on the importance of afforestation and proper management of trees
in the city and can help farmers in rural areas to avoid hours of manual inspection
and labor spent in inspecting the health of their farms.

Keywords IoT · Tree management · Cloud computing · Green city · Smart tree
system

1 Introduction

A city can be classified as smart and green only through vital arrangement of informa-
tion and communication technology foundations and administration to accomplish
maintainability strategy goals. However, this requires that trees must be included.
Not only do trees just constitute green space valuable to differentiate urban con-
tamination impacts or give ecosystemic advantages to inhabitants, but also they can
likewise be utilized as bioindicators and their association in correspondence systems
can symbolize to a huge commitment to construct/build a keen, green city. Smart tree
management utilizes sensors to screen the strength of trees over the city to monitor
the air contamination and advance planting of more trees with the assistance of web
and mobile application. With a quickly evolving atmosphere, scene, and city frame-
work because of human maneuvers, successful administration of plants is required
like never before [1]. Advancements, for example, radio-frequency identification and
remote wireless systems can interchange objects to residents as well as trees. The
smart city idea can be viewed as an expansion of smart gadgets that discourse singu-
lar needs and smart frameworks that address the necessities of a household/gathering
of building clients. Greening metropolitan spaces is a key movement that practically
each management seeks after [2].
The protection of environment is an originator to a sustainable, adjusted, and
comprehensive development. The responsibility to ensure a high caliber and pros-
perous life for the future age cannot be met unless the present pressing environmental
issues are tended to in an appropriate and auspicious way. We believe that a con-
nected smart management solution for trees over the city is a standout among the
most naturally cordial approaches to enhance the air quality and biology in urban
zones, in this way empowering urban areas to have much lower carbon impressions.
This smart tree can be utilized for the tourists and residents. An augmented reality
encounter lets users simply point to a specific tree and view its health and the air
quality details. There is additionally an option to alert the authorities if a specific tree
is obstructing the street or causing inconvenience. With a connected tree framework,
a campaign for boosting green propensities can be remunerated. Moreover, users can
assert different sorts of saplings to upgrade greenery and cultivate an adoration for
nature in the psyches of new ages. Consequently, users can examine specific trees
with their mobile applications and win tree saplings when they answer a couple of
trivia questions concerning the area around a tree. The need for this solution stems
IoT-Based Smart Tree Management Solution for Green Cities 183

from the continuous temperature rise caused by a dangerous atmospheric deviation.

This adds to the water deficiency of the United Arab Emirates, due to dry season,
rising sea level, and aridity. Ecological issues in the UAE are caused by the misuse
of characteristic assets, quick populace development, and high energy necessities.
The farmland of the United Arab Emirates, portrayed with its enormous dried land,
accidental precipitation, and high temperatures are as of now getting to be bone-dry
[3, 4].
In the recent years, the number of deaths due to respiratory diseases has become
significant. A major contributor to this statistic is due to abundant unmonitored pol-
lution across urban cities. Figure 1 obtained from the UAE Ministry of Health shows
the main causes of deaths in the United Arab Emirates. Respiratory diseases rank
extremely high and are responsible for majority of deaths in the country [5]. Globally,
epidemics caused by contamination were responsible in 2015 for an expected 9 M
premature deaths—16 percent of all casualties around the world, which is three times
greater number of deaths than AIDS, tuberculosis, and intestinal sickness combined,
and fifteen times more than all wars and different types of violence. It kills a greater

Fig. 1 Circuit diagram and components of prototype

184 B. Shabandri et al.

number of individuals than smoking, poverty, and natural disasters. In a few nations,
it accounts for one out of four demises [6].
The rest of the paper is organized as follows. Section 2 describes the related work
carried out in the field of IoT and green cities. Section 3 discusses the prototype
implementation and is subdivided further into the hardware components and mobile
application. Section 4 explains the system architecture and details the implementation
of such a system on an industrial scale along with test results and screens from the
user interface of the mobile application. Finally, Section 5 presents our conclusion
and future prospects with the project.

2 Related Work

Before delving into the points of interest of our Internet of things-based smart tree
management framework, we present an audit of a portion of the current framework
in vogue relating to smart tree system. Enormous research has been conducted in
the improvement of models in the extensive rural area. Applying models on farms
has been generally sporadic despite the potential favorable circumstances that could
result. In spite of the fact that Internet of things will be an essential innovation for the
smart cultivation itself, their utilization in conjunction with future Internet (FI) offers
a reason for another age of Farm Management Information Systems empowering keen
ranches end up plainly dynamic hubs in business to business (B2B) arrangements
and farming worth chains [7].
Conventional garbage administration utilizing human is not extremely powerful
with any legitimate supervision. Research has been done in utilizing IoT-based inno-
vation in checking the status of canister toward gathering the garbage once threshold
reached. There were certain downsides in the current IoT-based framework which
prompted the improvement of IoT-based smart garbage management system utilizing
ultrasonic sensors. The venture created toward savvy garbage management got parcel
of scope for future improvement. One of the key parts of smart city is smart garbage
management which includes checking garbage in city toward expulsion appropriately
and furthermore arranging off. There is no appropriate supervision from corporation
toward checking the movement of garbage accumulation [8].
Radio-frequency identification innovation represents one of the most standouts
among the most fascinating devices in machine-driven and designed parts. It signifies
another framework to build process proficiency in supply chains. While picking a tag
for a tree radio-frequency identification framework, it is important to consider differ-
ent perspectives, including size and shape, term, protection from outer physical and
compound components, and separation from the per-user/reader, breakdowns close
metals and fluids, satisfaction of nearby directions, and memory limit. When man-
aging tiny trees, for example, those commonly found in nurseries, radio-frequency
identification label imbedding inside organs of little size requires particular strate-
gies, an echo in regard to labels imbedding in Citrus spp. at nursery arrange is a case
of a plant labeling system that was dependable, rock-solid, and secure. Regardless of
IoT-Based Smart Tree Management Solution for Green Cities 185

whether not straightforwardly related to plants, radio-frequency identification labels

could be executed in biomonitoring frameworks with a specific end goal to ensure a
constant information correspondence [9].
Soils are vital to the capacity of every single imaginable biological system and to
nourishment and fiber generation. An ignored part of soils is their capability to mod-
erate ozone-depleting substance discharges. Albeit demonstrated practices exist, the
usage of soil-based ozone harming substance alleviation exercises is at a beginning
time and precisely evaluating discharges. This will encourage the execution of atmo-
sphere smart soil administration arrangements, through top and trade frameworks,
item inventory network activities for low-carbon customer items, and national and
worldwide greenhouse gas moderation approaches; it will likewise advance more
economical and atmosphere strong agrarian frameworks [10].
Previous work in the field has been done such as the work of Chua [11], which
provides a tree MIS used to improve the efficiency of specialist consultants such
as trained company personnel. It is geared more toward areas with high pedestrian
and traffic flows, such as crowded streets and it focuses on risk mitigation due to a
collapse of trees in urban areas. Our work is geared toward easing the work of farmers
in rural areas and we achieve this by employing additional sensors that give a more
holistic picture of a particular tree in the farm. As the worldwide population rises,
progressive decrease of energy sources and their resulting higher cost, environmental
change and air contamination are a portion of the primary issues that the urban areas
without limits should adapt to survive, changing into smart cities and concentrating
on green information technology and smart mobility.
Trees and individuals can be present together in urbanized spaces and can be
beneficial to each other. Trees can control microclimates, for example, air tempera-
ture and humidity diminish the power of bright light through screening and deliver
an abnormal state of oxygen while engrossing carbon dioxide during the day. The
development of trees is compelled by limited planting and confined headroom or is
influenced by development activities. In this manner, individuals have a commitment
to enable trees to develop in urbanized spaces by, for instance, checking their status
and actualizing well-being therapeutic activities [12].

3 Prototype Implementation

3.1 Hardware Components

The proposed method consists of various sensors which are connected to the Rasp-
berry Pi 3. For this prototype, we show a Pi connected to the Internet via an Arduino
Shield consisting of narrowband IoT (NB-IoT) [13] and LTE Cat-M1, category M1
(LTE-M) [14] modules for connectivity. The NB-IoT is a development of the LTE
framework and works with a transporter data transfer capacity of 180 kHz [13–15].
186 B. Shabandri et al.

Figure 1 consists of various modules and sensors used for the prototyping device.
It includes the DHT11 temperature and humidity sensor, MQ135 gas sensor, YL-69
moisture sensor, LDR, SEN-09198 piezo vibration sensor, and an MCP3008 for 8
channel analog input as the Raspberry Pi does not directly support analog inputs.
(a) Raspberry Pi 3 Model B—The Raspberry PI3 Model B+ retains the same
mechanical footprint as its predecessors while improving in performance. Its
1.4 Ghz quad core 64 bits processor coupled with 1 GB of RAM ensures unprece-
dented speed. It has Wi-Fi 2.4 and 5 Ghz wireless connectivity as well as Blue-
tooth 4.2. It comes with a faster RJ45 plug, four USB ports and a Micro SD
card slot in a compact format.
(b) Sound Sensor—To detect sound pollution in the vicinity, a module is used to
identify the point in time in which the sound has crossed a selected threshold.
Sound is detected via a microphone and fed into an LM393 op-amp. This thresh-
old is set using the in-built variable resistor. To display the crossing of the limit,
an LED on the module illuminates and the resulting output is set to low.
(c) Piezo Vibration Sensor—To detect if tree has sudden movements indicating
impacts, a piezoelectric sensor (7BB-20-6L0) is used. It works as a transducer
by measuring changes in weight, speeding up, temperature, strain, or power and
changing over them to an electrical charge.
(d) LDR—To detect if the tree is receiving enough sunlight, a light-dependent
resistor (LDR) is used. An LDR is sometimes called a photoresistor or a pho-
todetector as it changes resistance based on the amount of light that falls on
its surface. It is produced using semiconductor, cadmium sulfide (Cds), or cad-
mium selenide (CdSe). These two substances are semiconductors covered in a
clay sheet as a base.
(e) Temperature Sensor—To ensure that adequate water is reaching the tree,
DHT22 is used as a sensor for measuring temperature and humidity. The esti-
mation unit will be exhibited in a computerized flag/signal form.
(f) Air Quality Sensor—To detect air quality and pollution levels around the tree,
the module fills in as an air quality detection gas sensor, this is delicate to gas
unsafe to human, connected to measure NOx , alchohol, NH3 , benzene, CO, CO2 ,
etc. The module is additionally utilized for controlling climate conditions and
air cleaners in building structures. The estimation/measurement unit is intro-
duced as an analog flag/signal. In the present work, three sensor shapes were
utilized, comprising of MQ-2, MQ-135, and MQ-136 due to its gas measurement
contrasts.
(g) Moisture Sensor—To detect the soil moisture level, the SEN13322 is used.
It has two large/expansive uncovered cushions which work as probes for the
sensor, together acting as a variable resistor. The more water that is in the
dirt/soil means the better conductivity between the cushions will be and will
bring about a lower opposition/resistance, and a higher SIG out.
(h) Carbon Sensor—To ensure that the tree is intaking carbon dioxide ensuring
tree’s health.
Combining all the sensors, connectivity. Figure 2 depicts a test setup of the smart
IoT-Based Smart Tree Management Solution for Green Cities 187

Fig. 2 Test set up of the IoT monitoring device

tree monitoring device connected with the various sensors sending data to the cloud
platform through the NB-IoT shield to create a map view of the pollution points
across the city.

3.2 Mobile Application

A mobile application has been developed for the user which consists of three main
features or tabs, namely—‘Stats,’ ‘Schedule,’ and ‘Support.’ In the Stats tab, Fig. 3a
shows the dashboard of the mobile application with a custom user interface giving
options to monitor the water, air, sunlight, or overall health of the farm. Figure 3b
shows intelligent insights and predictive analytics based on various parameters col-
lected from the data from the individual trees. Figure 2c shows an in-depth view of
a selected tree in the farm and displays useful information such as alerts, monthly
reports, location, and infections.
The Schedule tab allows the farmer or the user to set reminders on his phone to
visit a specific tree so that he can water it, harvest it, or apply medication to it. The
Support tab allows the user to get in touch with the equipment providers through a
live chat in case a particular device outputs incorrect/inaccurate value.
Smart trees can likewise be used for city authorities. An outline of all the associated
trees is given on an adaptable guide that gives clients a chance to channel consid-
ering air quality, populace density, water level, light level, and different parameters.
This information can enable the specialists to draft afforestation intends to quicken
greenery over the city. There is additionally a chatbot for the city specialists to screen
188 B. Shabandri et al.

Fig. 3 Mobile application: a dashboard screen, b intelligent insights screen, c detailed tree-specific
insights

and team up more effectively. Moreover, there is an Amazon Alexa application to

discover the health of trees at a specific area with the use of voice. This information
can enable the authorities to draft afforestation plans to quicken greenery over the
city. We are using trees to reduce air pollution as well as sound pollution which pro-
tects the overall health of the population; on top of everything, we are monitoring the
trees when they are about to fall through movement sensor, to avoid hazard accidents
where it might hit a car or injure a person.

4 System Architecture

Now, we present a smart tree management (STM) solution based on Raspberry Pi

3, LPWAN modules and sensors. Figure 3 shows the system architecture of STM
which consists of a Raspberry Pi 3, carbon sensor, temperature sensor, humidity
sensor, light-dependent resistor (LDR), air quality sensor, piezo vibration sensor,
and sound sensor.
Figure 4 represents the distributed implementation system for the smart tree man-
agement solution. Each tree is equipped with NB-IoT and LTE-M sensors that com-
municate directly with the IoT cloud platform. The issue with traditionally creating
IoT-Based Smart Tree Management Solution for Green Cities 189

Fig. 4 Distributed implementation of monitoring devices

multiple gateways for assets was that devices that can connect to 4G are quite expen-
sive in an industrialized IoT use case with thousands of sensors constantly sending
and receiving data. Moreover, the process of connecting via Wi-Fi or a gateway is
quite complicated. With the help of LTE-M and NB-IoT solutions [16], such low-
density sensors and devices can communicate directly with the cloud and provide
for a plug-and-play solution that allows for greater scalability of the system.
Figure 5 shows the block diagram and working of the entire solution. Data flows
from thin devices directly to the IoT cloud server where information is analyzed and
segregated. It is then transmitted to a mobile application through the IBM Cloud
[17] API’s with an intuitive and clean UI for users to take action. The working of the
system is divided into three parts: (A) Sensor Node, (B) IoT Cloud Server, and (C)
User Interface.

4.1 Sensor Node

This consists of the industrial device that measures various essential parameters
about the health of a tree. This sensing is done with the help of an onboard MCU that
190 B. Shabandri et al.

Fig. 5 Block diagram of an industrialized system

performs necessary calculations, regularizations, and data cleaning. This data is then
sent to the IoT cloud platform directly through the LPWAN communication unit.
• Sensing Unit: The custom industrialized monitoring device consists of environ-
mental sensors and NB-IoT/LTE-M modules. The lower portion of the device
consists of a moisture sensor that is pushed into the ground to help monitor the
amount of water received by a specific tree. The sensor is also connected to LED
indicators that help to adjust the sensitivity of detection so that each device can be
manually calibrated to the specific amount of water required for being detected as
‘watered.’ The devices are designed to be modular and ‘plug-and-play’ compatible
with a variety of types of trees. The device is powered by batteries and can have a
battery life of up to 10 years.
• Processing Unit: LPWAN applications are best suited for devices with non-critical
data with low processing power. Thus, a device with minimal processing power
is sufficient for this particular use case. The unit receives the values from all the
sensors and stores them in the appropriate variables and prepares for Message
Queuing Telemetry Transport (MQTT) which is used for transfer of the data to
the cloud [18].
• Communication Unit: The device communicates via the MQTT protocol with
the help of subscribe and publish keys.
IoT-Based Smart Tree Management Solution for Green Cities 191

Fig. 6 Prototype unit consisting of various sensors

• The interval of packet transmission can be adjusted in the device management

platform so that battery life can be optimized. The GNSS capability in the NB-
IoT/LTE-M module provides autonomous geospatial positioning with global cov-
erage. During installation, the devices require a one-time setup for initializing the
name of the tree, setting the exact coordinates, and adjusting the sensitivity of
various parameters which is calibrated by a trained professional (Fig. 6).

4.2 IoT Cloud Server

The process of cloud enablement includes device authentication and identity, setting
up device templates, device dashboards, alert monitoring, and managing the database.
This can be achieved by setting up three components:
• Node-RED application
• Internet of things platform service
• Cloudant NoSQL database
Figure 7 describes a high-level overview of the system and lists the various sensors
required for connectivity along with the various cloud services and frameworks used
in the implementation of the system.
192 B. Shabandri et al.

Fig. 7 Overview of prototype implementation

(1) Setting up Device Identity

For the prototype, we used Node-RED on the IBM Watson IoT platform which allows
for the management of multiple IoT devices. Custom function nodes are used for
parsing the incoming data streams and to ensure that messages are not lost if there is
a huge influx of data. In the device information section (Fig. 8), we configure detailed
information regarding the device identity such as serial number, hardware version,
manufacturer, firmware version, etc. This metadata can be useful for device updates
and debugging purposes. After this, we define an authentication token that is either
system generated or manually defined by us.

Fig. 8 Updating device information and metadata

IoT-Based Smart Tree Management Solution for Green Cities 193

Fig. 9 Cloudant NoSQL DB for data storage and analysis

(2) Connecting to the Database

As shown in Fig. 9, IBM Cloudant is a NoSQL JSON record store that is
upgraded/optimized for handling of overwhelming workloads of simultaneous reads
and writes in the cloud; a workload that is regular of expansive, quickly developing
web and mobile applications. You can utilize Cloudant as a completely managed
DBaaS running on IBM Cloud.
For the database, we use a Cloudant NoSQL DB for all our data storage and
communication with the Watson IoT platform. One way to ensure that we define the
categories of data we are using from the database is by defining functions, which can
be considered as a form of a special document called ‘design document.’ We created
a design document with a simple view, which is a blueprint for how our functions
are defined and how the data will be categorized. This view grabs all documents with
the key corresponding to the timestamp defined for each tree/device and the value
corresponding to the entire document object. Instead of the timestamp, a unique ID
can also be generated for each data transfer request if the application requires.

4.3 Experimental Results

We used Python to establish an MQTT connection between the Raspberry Pi and the
Watson IoT platform. This can be achieved by downloading the ibmiotf library and
entering various credentials details such as deviceID, orgID, authentication token,
etc. This is followed by initializing the device client and connecting and sending a
data point with a value to the cloud as an event. Appropriate delays can be introduced
between message publish requests as required. Finally, this data is sent to a real-time
194 B. Shabandri et al.

IoT database (Fig. 5) where each message is stamped with a unique ID or timestamp.
This data can be downloaded as a CSV or directly connected to other cognitive
services for further analysis and pattern recognition. The database shows successful
readings for six values that are obtained from the various sensors placed in the tree.
Some columns have been omitted due to lack of relevancy. There is a jump in the
temperature values since hot air was pushed directly into the sensors to test the
functionality of the sensors. As we will see in the next section, data from this table
was analyzed with the help of various functions and tools that make it easier to predict
trends and observe all the problems with a specific tree.
(a) Customizing Data Flow and Functions
Node-RED is a programming apparatus for wiring together hardware, APIs, and
online services. A program-based editorial manager makes it simple to wire together
streams utilizing the extensive variety of hubs/nodes in the palette that can be sent to
its run time in a single click. Node-RED also provides us with an app URL that can
be used to display the front end in an active URL that can be accessed remotely on
any device. The IBM IoT node allows us to consume any data received by the IoT
platform service. The connection to the Cloudant database is also done at this node.
The data is then sent over to individual functions that contain single parameters
regarding the health of the tree, which is passed on to individual nodes that are
responsible for generating visual displays of the data in various forms such as line
charts, area graphs, pie charts, etc. (Fig. 10).

Fig. 10 Node-RED flow for dashboard view

IoT-Based Smart Tree Management Solution for Green Cities 195

4.4 User Interface

(a) Mobile Application

The mobile app interfaces with the IBM Bluemix platform by invoking REST API’s
with IBM API Connect used to discover APIs, test them, and get code snippets to
make it easier to call them from the mobile apps.

• Augmented Reality View: Our mobile application provides an interface for the
user where they can access an augmented reality view of smart trees in the public.
They must simply point their smartphone cameras to a tree, and they can get infor-
mation about the health of the tree for instance the amount of light it is receiving,
the moisture content, the air quality around it, and many other characteristics of
the tree (Table 1).
• Hazard Reporting: In case of a scenario where a certain tree is blocking the path
of vehicles or posing a threat to the pedestrians, the users have an option of simply
selecting the particular tree and reporting it as hazardous. This data is then verified
by the relevant local authorities and teams are quickly dispatched to solve the
problem.
• Tree Health Alerts: In addition to hazard reporting, users can also report trees that
look unhealthy—i.e., they are not watered regularly, they are affected by bugs, or
if there are polluting sources nearby. These reports are sent to relevant authorities
so that they can take the required action to solve the problem. The heat map of air
pollution levels across the city is shown in Fig. 11.
• Community Integration: Along with reporting, various campaigns can be
launched to encourage people to plant more trees. For instance, using the aug-
mented reality interface, every week a contest can be held where community
members have to take pictures of particular trees across the city to win saplings
that they can plant. Similarly, other interactive activities can create a closely knit
community that contributes to afforestation in the city.

(b) Web Application

The web application, as shown in Fig. 12, is meant as a control panel or dashboard
for city authorities or officials for monitoring tree hotspots in the city. This consists
of an overall map view of the city as well as a data analytics and monitoring section.
Unlike the mobile app, access to this application is restricted to certain users who
are provided with historical and real-time datasets of smart trees across the city.

• Map View with Filters: The map view consists of the culmination of all connected
trees across the city. This is displayed in the form of a heat map based on all the
measured parameters. With a simple toggle, city authorities in the city can detect
areas with maximum pollution, areas with the least trees, and regions with the most
complaints. Thus, individual trees can be monitored, and a variety of antipollution
campaigns can be launched to suggest the most appropriate area for tree plantation
cultivation.
 196

Table 1 Real-time sensor values

S. No. Humidity Temperature Intensity Air quality Carbon level Sound level Impact level
1 39 29 761 221.8994142 32.11447715 1.841570345 25.87355163
2 39 29 761 221.8994142 32.11447715 1.841570345 25.87355163
3 39 29 761 221.8994142 32.11447715 1.841570345 25.87355163
4 39 29 761 221.8994142 32.11447715 1.841570345 25.87355163
5 39 29 761 221.8994142 32.11447715 1.841570345 25.87355163
6 64 40 750 232.7778636 53.60739733 1.542790012 8.215529875
7 64 40 750 232.7778636 53.60739733 1.542790012 8.215529875
8 64 40 750 232.7778636 53.60739733 1.542790012 8.215529875
9 64 40 750 232.7778636 53.60739733 1.542790012 8.215529875
10 64 40 750 232.7778636 53.60739733 1.542790012 8.215529875
11 64 40 750 232.7778636 53.60739733 1.542790012 8.215529875
12 64 40 750 232.7778636 53.60739733 1.542790012 8.215529875
13 64 40 750 232.7778636 53.60739733 1.542790012 8.215529875
14 30 30 750 259.3853959 111.6718109 1.530899469 9.285961972
15 30 30 750 259.3853959 111.6718109 1.530899469 9.285961972
B. Shabandri et al.
IoT-Based Smart Tree Management Solution for Green Cities 197

Fig. 11 Heat map of air pollution levels across the city

Fig. 12 Dashboard with real-time tree health data

198 B. Shabandri et al.

• Data Analytics and Live Monitoring: The web app allows real-time information
viewing of tree health data across the city. Individual trees can be assessed at spe-
cific time frames and patterns can be obtained. This information can be extremely
useful for predicting the best environment for tree growth and good behavior from
citizens can be incentivized and measured with active goals such as reducing car-
bon levels in a specific timeframe or decreasing level of sound pollution in an area.
This sound data generated across the city can also be used to draft out a sound
density map of the city. This can be used in innovative ways of energy production
using sound.

5 Conclusions and Future Prospects

In this paper, a prototype of a green smart tree management system based on the
IoT is discussed and implemented. The paper describes the functions and actual
characteristics of each system component in detail. The experimental results suggest
the feasibility of the application to enable the development of smart cities across the
world which will play a key role in decreasing pollution levels across the city and
increasing afforestation in areas that require it the most. In the future, we can expand
the prototype to simulate an actual city infrastructure by extending the network of
connected trees and improvements can be made in the design and casing of the model
to avoid damage due to rain, harsh weather conditions, or human interference. Finally,
we conclude that this solution can be useful in spreading awareness among citizens
on the importance of afforestation and proper management of trees in the city and
can help farmers in rural areas to avoid hours of manual inspection and labor spent in
inspecting the health of the farm. Extended applications of the research could include
employing CCTV cameras to remotely monitor multiple sites in the city. Computer
vision and machine learning could be utilized to provide predictive analytics about
damage that could occur to the tree and preventive action can be taken before the
actual damage takes place.

References

1. Anguluri, R.: Role of green space in urban planning: outlook towards smart cities. Urban For.
Urban Greening 9(2) (2017)
2. Brown, P.R.: The role of citizen activists in urban infrastructure development, growing greener
cities: urban sustainability in the twenty-first century. In: Birch, E.L., Wachter, S.M. (eds.),
pp. 152–169. University of Pennsylvania Press (2011)
3. Kirk, S.: Energy performance resilience of UAE buildings to climate change. Int. J. Environ.
Sustain. 7(1), 90–97 (2018)
4. Hamdan, N.M., Alawadhi, H., Jisrawi, N.: Particulate matter pollution in the United Arab Emi-
rates: elemental analysis and phase identification of fine particulate pollutants. In: Proceedings
of the 2nd World Congress on New Technologies, Budapest, Hungary, 18–19 Aug 2016
IoT-Based Smart Tree Management Solution for Green Cities 199

5. Barakat-Haddad, C., Zhang, S., Siddiqua, A., Dghaim, R.: Air quality and respiratory health
among adolescents from the United Arab Emirates. J. Environ. Public Health 1, 1–13 (2015)
6. Al-Maskari, F., Bener, A., Al-Kaabi, A., Al-Suwaidi, N., Norman, N., Brebner, J.: Asthma
and respiratory symptoms among school children in United Arab Emirates. Allergie Immunol.
32(4), 159–163 (2000)
7. O’Grady, M.J.: Modelling the smart farm. Inf. Process. Agric. 4(3), 179–187 (2017)
8. Dwivedi, P.: IoT based smart garbage management system. Int. J. Adv. Trends. Comput. Sci.
Eng. 6(4), 71–76 (2017)
9. Luvisi, A., Lorenzini, G.: RFID-plants in the smart city: applications and outlook for urban
green management. Urban For. Urban Green. 9(2), 1–18 (2014)
10. Paustian, K.: Climate-smart soils, nature. Int. J. Sci. 532, 49–57 (2017)
11. To, W.M.: Tree MIS caring for ecological assets in smart cities. IEEE Comput. Soc. 18, 50–55
(2016)
12. Chua, M.: Tree management system. Proc. De La Salle Univ. Research Congress, no. SEE-III-
021 (2014)
13. Wang, Y.P.E.: A primer on 3GPP narrowband internet of things. IEEE Commun. Mag. 55(3),
117–123 (2017)
14. Chiti, F., et al.: Optimized narrow-band M2M systems for massive cellular IoT communications.
IEEE Global Commun. Conf. (GLOBECOM) (2016)
15. Nokia Networks.: Future Works LTE-M—optimizing LTE for the internet of things. novotech.
com/docs/default-source/default-document-library/lte-m-optimizing-lte-for-the-internet-of-
things.pdf
16. Rapeepat, R., et al.: NB-IoT system for M2M communication. IEEE Wirel. Commun. Netw.
Conf. Workshops (WCNCW) (2016)
17. Derhamy, H., et al.: A survey of commercial frameworks for the internet of things. IEEE 20th
Conference on Emerging Technologies and Factory Automation (ETFA) (2015)
18. Hunkeler, U., et al.: MQTT-S—A publish/subscribe protocol for wireless sensor networks. 3rd
International Conference on Communication Systems Software and Middleware and Work-
shops (COMSWARE 08) (2008)
Data Analysis and Prediction Using Big
Data Analytics in Agriculture

K. Aditya Shastry and H. A. Sanjay

Abstract Big data analytics analyzes large sized data to discover useful and
hidden patterns along with correlations/associations and other beneficial insights.
Even though rapid developments in technology have taken place, and traditional
agriculture practices are being followed in several countries (especially developing
countries). In this regard, big data analytics possesses the capability to positively
transform the agricultural domain. The agriculturists are heavily dependent on tra-
ditional practices to select suitable crop to grow and how to grow the crop in effi-
cient and productive way. These traditional methods usually comprise of walking
in fields, choosing soil samples for soil moisture analysis, and detecting plant dis-
eases by observing the leaves manually. It is in fact significant for the agriculturists
(especially farmers) to understand the importance of technology for gathering crop
information in large quantities for transforming the collected data into meaningful
knowledge. Integrating big data with the agricultural domain can bring about radical
changes in how agriculture is being done currently. The technique of growing crops
effectively can be well understood by the appropriate utilization of big data analytics.
The three extensive areas which are transforming agriculture are accurate crop fore-
casts, precision farming, and generating high-quality seeds. This chapter is composed
of four sections. The first section provides an overview of analytics on big data in
agriculture. This is followed by the second section which will highlight the different
big data techniques in agriculture. The third section describes two case studies that
apply techniques for big data analysis in the agricultural domain. The fourth section
analyzes the various research areas related to agriculture that can be explored by the
academicians and researchers. The summary of the chapter is provided at the end.

Keywords Big data analytics · Agriculture · Prediction · Analysis

K. Aditya Shastry (B) · H. A. Sanjay

Department of Information Science and Engineering, Nitte Meenakshi Institute
of Technology, Yelahanka, Bangalore, Karnataka 560065, India
e-mail: [email protected]
H. A. Sanjay
e-mail: [email protected]

© Springer Nature Singapore Pte Ltd. 2020 201

P. K. Pattnaik et al. (eds.), Internet of Things and Analytics
for Agriculture, Volume 2, Studies in Big Data 67,
https://doi.org/10.1007/978-981-15-0663-5_10
202 K. Aditya Shastry and H. A. Sanjay

1 Introduction

Agriculture represents the sector which is responsible for providing food to every
human being in the world. However, the people who do agriculture are not given
much importance. After failing institutions, time has indeed come for technology to
take over the change. Emerging technologies need to be incorporated in agriculture as
new issues or challenges are faced everyday by the different agriculture stakeholders
especially farmers.
Agriculture in many countries lacks institutional attention, supports from banks in
terms of loans and farmer welfare schemes, and suffers from a myriad of disasters like
depleting groundwater levels in rural areas, climate change, unpredictable monsoon
or lack of it, droughts, floods, unfair price-fixing policies of produce, migration of
farmers toward the cities for pursuing better paying jobs, and more.
Big data analytics analyzes large sized data to discover useful and hidden patterns
along with correlations/associations and other beneficial insights. Even though rapid
developments in technology have taken place, traditional agriculture practices are
being followed in several countries (especially developing countries). In this regard,
big data analytics possesses the capability to positively transform the agricultural
domain. The agriculturists are heavily dependent on traditional practices to select
suitable crop to grow and how to grow the crop in efficient and productive way. These
traditional methods usually comprise of walking in fields, choosing soil samples for
soil moisture analysis, and detecting plant diseases by observing the leaves manually
[1].
However, it is critical for the agriculturists (especially farmers) to understand
the importance of technology for gathering crop information in large quantities for
transforming the collected data into meaningful knowledge. Integrating big data with
the agricultural domain can bring about radical changes in how agriculture is being
done currently. The technique of growing crops effectively can be well understood
by the suitable utilization of big data analytics. The three extensive areas which are
transforming agriculture are highlighted below [2]:
• Accurate crop forecasts: The major apprehension of farmers is how the crops that
they have sown will grow. In previous years, crop forecasts have been inaccurate.
The modern farmers can predict the crop yields more accurately using technolo-
gies that utilize high-end algorithms. These algorithms analyze huge amount of
crop and weather information to arrive at the predictions. Farmers can nowadays
predict the yield even before a seed is sown in soil. The big data analytics delivers
valuable insights that empower the farmers to harvest at appropriate times and
reap maximum yields. The farmer’s stress about worrying about the crop can be
considerably reduced by accurate crop forecasting.
• Precision farming: It represents the modern method to grow crops. Several of the
agricultural methods may be automated using the Internet of things (IoT). Through
precision farming, many farmers particularly in developed countries are profiting
when compared to the farmers who are not using the technology. The quality of
crops has also improved by the use of precision farming.
Data Analysis and Prediction Using Big Data … 203

• Generating seeds of high quality and minimizing hunger: Latest studies demon-
strate that starvation in the world is increasing due to the exponential growth of
population and global warming as seen in Africa where around 20 million humans
are facing the real threat of starvation. Though social activists across the globe
are putting their effort toward avoiding the food crisis, the actual solution may
lie in the proper application of big data analytics in the agricultural domain. The
plant data is being analyzed from past decades by several researchers and agri-
cultural scientists in the hope of cultivating effective and healthy crops. Recent
innovations in the analytical techniques for big data have led to the faster, larger,
and improved growth of crops. Global hunger can be eliminated by the use of
chemically engineered seeds assisted by big data analytics.

2 Big Data and Related Techniques in Agriculture

This section discusses the role of big data in agriculture. Some of the popular methods
for analyzing big data are also discussed. Before explaining the big data techniques,
we briefly look into what is big data.

2.1 Role of Big Data in Agriculture

According to [3], the big data is characterized using the following five dimensions:
• V1 denotes the volume of data. V1 indicates the data size gathered for analysis.
• V2 signifies the window of time between which the data is relevant and useful. For
instance, certain data has to be analyzed within specific time durations else the data
loses its significance. Specific examples include identifying pests and diseases in
animals within their lifetime.
• V3 represents variety of data. The variety in data may occur due to multiple
data sources (e.g., videos, remote sensing data, images), multi-temporal (e.g.,
data collected during different times or dates), and multi-resolution (e.g., spatially
diverse image resolution data). Data coming from different application domains
also possesses variety.
• V4 denotes veracity of data. V4 comprises of the potential, quality, reliability,
accuracy, and overall confidence of data.
• V5 indicates valorization of data which is its ability to spread appreciation, inno-
vation, and knowledge.
Even though the five “V”s are able to describe big data, the big data analysis needs
not to satisfy these five dimensions [4]. Generally, big data can have less accuracy
and lack stability making the fourth V weaker. According to [5, 6], another “V” for
visualization can be added for meaningfully demonstrating or illustrating the data
using complex data structures and information-rich scenarios.
204 K. Aditya Shastry and H. A. Sanjay

Makers of decisions/ challenges in

society

Interests related to economic,

Domain for environmental, social, preferences,
making decisions Wisdom values, trade-offs, ethics, intangibles,
risks…

Knowledge Options, Scenarios, Impact Assessments,

Science
Info+Appl Decision Support Systems
domain

Data analysis &

Information integration, AI,
data + added meaning

Databases,
(Big) Data
Satellites
raw material

Fig. 1 DIKW hierarchy [1]

As per [7], big data deals more with the capacity to cross-reference, search, visu-
alize, and aggregate large dataset faster. It largely emphasizes the ability to mine
information and insights which were expensive and not technically feasible [1, 8].
The relationships between information, wisdom, knowledge, and data are depicted
in the DIKW hierarchy where D stands for data, I for information, K for knowledge,
and W for wisdom. It basically conveys the concept that information is described in
data terms, knowledge with respect to information, and wisdom in terms of knowl-
edge. Figure 1 depicts DIKW hierarchy.
Figure 1 depicts how a lower-level entity (data) is transformed into a higher-level
entity (information). The processes involved in the transformation is also identified
and described. The concept is to take decisions based on evidence which requires
“wisdom.” The wisdom is combined with knowledge available having fewer assets
such as values, ethics, interests, and preferences that are tangible. From data, the
knowledge base is derived. In this respect, data is considered as the raw entity to
generate information along with meaning. The enrichment of information is done by
creating knowledge utilizing and combining policy and decision applications such
as instance integrated models, assessments of impact or support systems for decision
making which are contextual in nature.
The systems related to agriculture and environment research are dynamic in nature
and comprise of complex communications among living species and products which
are perishable (like humans, plants, agricultural products, animals) and their envi-
ronment. The elaborate information related to entity status and behavior and their
environmental conditions is needed for describing such systems. These systems can
Data Analysis and Prediction Using Big Data … 205

hold actual status, as well as conditions related to history or future. The data related to
spatial and temporal entity behavior is critical since the living systems possess vari-
ability in spatial and temporal conditions. Furthermore, detailed knowledge about the
different subsystems is required for understanding these interactions. The subsystems
may include climate, water, biological, and soil.
With regard to big data, the DIKW hierarchy demonstrates the process of trans-
forming raw data which is less useful to knowledge which is more useful to the
end users. The knowledge represents the pieces of data which are useful for pro-
viding meaningful decisions. This reliable and timely information is utilized by end
users who assume that the resources have been combined using the most suitable
sources from pool of big data, inferred and processed according to their decision
context. Obviously, such assumptions face a vast challenge to the community com-
posed of ICT experts, experts in domain, and data scientists who are involved in this
work. More cooperative approach and expertise with hybrid bottom-up and top-down
approaches are required for making decisions regarding the present context in big
data analytics [2].
Table 1 depicts the association among the number of works done in different
agriculture areas and the three V’s of big data. The third column gives the number
of papers that are giving solutions in the respective agricultural areas. Majority of
the studies have involved topics related to availability of food and security, insur-
ance and finance of farmers, changes in climate and weather, and research based on
management of lands and animals. The average ratings for first three V’s related to
volume, velocity, and variety with respect to each agricultural area are provided by
the authors [1] from columns 4 to 6 in Table 1.
The authors [1] used low (L), medium (M), and high (H) indicators for estimating
the first three “V”s of big data based on survey of 34 papers. These papers were
compared among them, and relative rankings in the form of L, M, or H were given.
The papers that were surveyed had data volumes ranging from medium to high and

Table 1 General areas of agriculture in which big data is used [1]

Agriculture area V1 V2 V3 No. of papers
Climate and weather change M M H 4
Land H L M 5
Research in animals M H L 4
Crops M M L 3
Soil M L L 2
Weeds L H L 1
Availability of food and security M L M 4
Research in plant and animal varieties M L H 1
Decision making by farmers H M H 2
Insurance and finance for farmers H M M 5
Remote sensing H M M 3
206 K. Aditya Shastry and H. A. Sanjay

variety ranging from medium to low. The projects related to weeds and animals were
the exceptions that involved high velocity since urgent actions are required when
proof of diseases and weeds are found. High information variety was required by
projects related to climate change and weather along with the apps related to decision
making by farmers since the regular forecasting of change in climate, weather, etc.
aids the farmers in doing his routine tasks. The applications linked with remote
sensing consumed highest volume of data since images have large sizes. The projects
related to land management along with availability and security of food, analysis of
soil, and biodiversity required the lowest velocity.
Lastly, soil, research related to animals, weeds, crops, and remote sensing possess
the least variety. From this, we can infer that these agricultural domains do not need
or do not have sufficient access to variety of data for resolving the domain-specific
issues. Table 2 depicts the associations of agricultural applications with the “V”
features of big data for the surveyed papers.
Data in large volumes is required for applications related to forecasting of weather,
security of food, estimation of yields of crops, and mapping of land. High velocity is
required for applications related to identification of diseases in animals, poor nutri-
tion in plants and decisions on productivity of farmers along with the quality, and
safety of food which have to be performed in real time (near). These applications
require decisions which have to be taken tactically faster instead of strategic plan-
ning for longer terms (lower velocity). The fourth dimension of big data which is
the variety of data is required by applications related to the estimation of insurance
and productivity of farmers. The fifth V related to high data reliability is required by
applications dealing with diseases of animals/plants, culling of herd, improvements
in crop productivity, and forecasting of yields. The number of publications and com-
mercial applications related to analytics of big data in agriculture is less, and hence,
it can be observed that it is still in early development stage.

2.2 Sources of Big Data and Techniques

It is worthwhile to note the sources from which big data originates for analysis to be
performed. Some of the common big data sources are as follows:
• Agricultural fields, i.e., from sensors on ground related to devices for chemical
detection, sensors that are bio in nature, stations associated with weather, etc.
• Data collected from third-party companies and governmental organizations such
as yearbooks related to statistics, reports from governments, guidelines, and reg-
ulations from public bodies, alerts, etc.
• Distributed sources such as online repositories and services of web
• Data from sensors which are airborne like satellites, planes, aerial vehicles which
are unmanned
• Web data from companies through web services which are online, social media
feeds, etc.
Data Analysis and Prediction Using Big Data … 207

Table 2 Usage of big data in diverse agricultural applications [1]

Big data dimension Description of “V” Agricultural applications
V1 Large data volume Forecasting weather [9], culling of
daily herd [10], productivity
improvement of farmers [11],
protection and insurance of small
farmers [12], financing of farmers
[13], estimations of food security
using remote sensing [14],
classification of land use and land
cover [15], sharing of data for
earth observations [16]
V2 High velocity for time critical Prediction of weather,
applications fermentation of wine [9], quality
and safety of animal food [17],
discrimination of weed [18],
recognition of diseases in animals
[19], improvement in productivity
[11], financial transactions of
farmers in secluded regions [20],
data sharing of earth observations
[16]
V3 High variety for data originating Identification of management
from heterogeneous sources zones [21], estimation of food in
developing countries [22],
evaluation of wildlife population
[23], insurance and protection of
farmers [12], improvement of
farmers productivity [10],
understanding of farmers
sustainability and operational
efficiency [24], tolerance of crop’s
drought [25, 26], science of
climate [27]
V4 High veracity for applications Dairy herd culling [10], quality
where data reliability and accuracy and safety of animal food [17],
is critical weed discrimination [18], animals’
disease recognition [19], food
availability estimation in
developing countries [22], small
farmers’ insurance and protection
[12], farmers’ productivity
improvement [11], data sharing of
earth observations [16], evaluate
wildlife population viability [23]
208 K. Aditya Shastry and H. A. Sanjay

The data from above-mentioned sources differs in velocity and volume and are
mostly heterogeneous in nature. The data can be accessed differently like in web
services, repositories that are static, live feeds, archives, files, etc. The data is also
characterized by different formats and types.
The big data sources along with the related technologies are shown in Table 3 for
each agricultural domain.
Every application of agriculture requires diverse big data sources to resolve the
associated problems. The commonly used big data sources include static reposi-
tories, geospatial, and remote sensing using satellites for most of the agricultural
domains. The ground sensors are utilized by research related to soil, animals, and
crops. Similarly, weather stations are used for applications related to climate change
and weather, mapping of lands, decision making for farmers, finance, and insurance.
Most of the works have combined data from variety of sources to address specific
problems.

2.3 Tools and Techniques for Analyzing Big Data

in Agriculture

Column 4 of Table 2 provides particular approaches that are utilized in different

agricultural areas. Among the 34 papers surveyed by author [1], ML is utilized in
13 papers, cloud-based platforms is used in nine papers, image processing in eight
papers, simulation and modeling in seven papers, analysis based on statistics in six
papers, and vegetation indices (NDVI) in six papers. Apart from this, five papers used
online services and four papers in geographical information systems (GIS). The land
applications related to remote sensing among many others use hybrid techniques
except certain applications related to soil.
The following tools are utilized in the respective applications as per the survey
done by authors [1]
• For agricultural prediction, the machine learning tools are used by Ma et al. [28]
and Mucherino et al. [29].
• The authors [9] and [30] used clustering in agriculture.
• Armstrong et al. [31] and Meyer et al. [32] utilized classification techniques in
agriculture.
• The work by Chi et al. [3], Ishimwe et al. [33] used image processing techniques
like harmonic and Fourier analysis, decomposition of wavelets, and fitting of
curves in agriculture.
• Remote sensing techniques along with image processing were utilized in Karmas
et al. [6] in which the input to the ML model was the image processing output.
• Large-scale storing is possible in agriculture using cloud platforms along with
MapReduce as demonstrated in the work by Hashem et al. [5]. The Geographical
Information Systems (GIS) were utilized by Lucas and Chhajed [34] for geospatial
problems.
Data Analysis and Prediction Using Big Data … 209

Table 3 Big data sources with techniques for analyzing the big data in agriculture [1]
Agricultural domain Big data sources Big data analysis techniques
Climate and weather change Weather stations, surveys of Techniques of machine
historical static information learning related to scalable
(climate and weather data), vector machines (SVM),
remote sensing through analysis based on statistics,
satellites, data related to cloud platforms, GIS
geospatial analysis, MapReduce
analytics
Land Remote sensing using Techniques of machine
airplanes, radar data, data in learning related to SVM,
geospatial zone, historical K-means clustering, random
datasets linked to forests, randomized trees,
characterization of land, vegetation indices related to
phenology of crops, NDVI
temperature and rainfall,
cover maps of global tree,
elevation, sensors in cameras
related to imaging of
multispectral data, weather
stations
Research in animals Historical information of Methods of machine learning
animals and soils related to related to decision trees,
their physiological SVM, and neural networks
characteristics, ground
sensors related to grazing
action, intake of food, heat,
production of milk, weight),
camera sensors
Crops Sensors placed in ground Machine learning (SVM,
(metabolites), remote sensing clustering based on
(satellite), historical datasets K-means), Wavelet based
(utilization of land, filtering, Fourier transform,
information of land country NDVI vegetation indices
wise, and data of statistics on
yields).
Soil Sensors placed in ground Machine learning (K-means
(salinity, conductivity of clustering)
electricity, humidity),
cameras (photosensitive),
historical databases
Weeds Remote sensing (plane, NDVI vegetation indices, ML
drones), information based on techniques like ANN and
historical (digital library logistic regression) and image
related to plant and weed processing techniques
images, data specific to
plants)
(continued)
210 K. Aditya Shastry and H. A. Sanjay

Table 3 (continued)
Agricultural domain Big data sources Big data analysis techniques
Availability of food and Surveys, historical Machine learning (neural
security information and databases networks), analysis linked to
(e.g., CIALCA, ENAR, rice statistics, simulation and
crop growth datasets), GIS modeling, analysis related to
geospatial data, statistical networks, analysis of GIS
data, remote sensing geospatial data, image
(synthetic aperture radar) processing
Research in plant and animal GIS geospatial data, Statistics (Bayesian belief
varieties historical information and networks)
databases (SER database of
wildlife species
Decision making by farmers Datasets containing historical Platforms which are cloud
information (e.g., survey data based, services based on web,
of US government), remote applications of mobile,
sensing related to satellites statistical analysis, modeling,
and drones, weather stations, simulation, benchmarking,
human based sensors, data big data storage,
based on web, data linked to message-oriented middleware
geospatial domain such as
GIS, social media based feeds
Insurance and finance for Web-based data, historical Cloud platforms, web
farmers information, weather stations, services, mobile applications
humans as sensors (crops,
yields, financial transactions
data)
Remote sensing Remote sensing related to Platforms linked to cloud,
planes, drones and satellites, analysis of data based on
datasets containing historical statistics, analysis of
information such as datasets geospatial data using GIS
of MODIS surface maps, processing of images,
reflectance, datasets of indices of vegetation related
surface land containing to NDVI, systems capable of
images, datasets of WMO making decisions, storage of
weather, heights of reservoirs big data, portals based on
originating from the altimetry community and web,
of the reservoirs, data based analytics using MapReduce,
on web, data related to applications of mobile,
geospatial regions such as computer vision, and AI
maps and imaging
Data Analysis and Prediction Using Big Data … 211

Database management systems (DBMS) that implement array data model are
suitable for storing big data which are heterogeneous in nature. NoSQL platforms
can be utilized for managing and storing huge volumes of unstructured data. The big
raster datasets are stored using the array DBMS that are specifically created for this
purpose.
Furthermore, vegetation indices (VI) have been utilized frequently in remote sens-
ing applications for soils or crops mapping. The VI is represented as combinations
of reflectance surface at 2/more wavelengths used to focus on a particular vegeta-
tion property. The NDVI is the most used VI which denotes a graphical indicator
utilized for analyzing the measurements of remote sensing and to estimate whether
the observed target possesses green vegetation or not. The message-oriented middle-
ware assists the event-based systems for notifying alerts faster such as prediction of
natural hazard, etc.
Table 4 illustrates the common and specific software’s utilized in big data analysis
based on the analysis type to be done.
A new practice is to apply image analysis for approximating agricultural problems
on the images derived from remote sensing. There was a rapid increase in the usage
of Landsat satellite data since 2008 due to free availability of the data. Remote
sensing is widely used in agriculture since it is not destructive and information may
be collected methodically over large geographical boundaries. Recent works by De
Leeuw et al. [35] have been on the product delivery of insurance materials due to
damage of crops, estimation of risks due to fire and floods or from surplus rain and
drought [12].

Table 4 Popular software tools utilized for big data analysis in agriculture [1]
Category Software tools
Tools related to image processing Toolkits related to OpenCV library, IM, and
VTK
Tools linked to ML TensorFlow of Google, Weka tool, Oryx,
Flavia, SHOGUN, R, mlPy, Apache Mahout,
Mlpack, and Mlib
Platforms based on cloud for storing big Cloudera, EMC Corporation, IBM InfoSphere
information along with computation and BigInsights, IBM PureData system for
analysis analytics, Aster SQL MapReduce, and
Apache Pig
Databases associated with GIS systems and MiraMon, Autodesk, MapInfo, ArcGIS
big data
Middleware linked with message-oriented PostGIS, Apache HAWQ, Oracle GeoRaster,
systems MongoDB, HadoopDB, Google BigTable,
Cassandra, MonetDB/SciQL, Hive, Oracle
GeoRaster, SciDB
Simulation and Modeling RabbitMQ and MQTT
Tools related to statistics R, NorsysNetica, and Weka
Analysis related to time series data MATLAB, Stata, BFAST, RATS
212 K. Aditya Shastry and H. A. Sanjay

The coarse resolution satellites like the AVHRR, MODIS, MERIS, and SPOT-
VEGETATION have become the major data sources for covering wide areas. The
ancillary data, viz. GIS data, field sensors, and historical data, is combined with
remote sensing to considerably improve the analysis performed particularly for pre-
dictive applications like identification of crops, or differentiating between grasslands.
As per the survey made by Kamilaris et al. [1] relatively fewer works exist in the
domain of analytics of big data for agriculture, indicating that it is still in early stage.
However, immense potential exists in this domain due to the diverse stakeholders
involved and the rapid dynamics involved in the system. New start-ups like [20–38]
have come up for delivering service to the agriculture stakeholders via the utilization
of big data analytics. The authors [5] and [30] observe that new business models may
be developed in agriculture because of rapid progress in cloud technologies which
enable large storage and computation facilities.

3 Case Studies on Big Data Analytical Techniques Used

in Agriculture

This section describes the different case studies used for big data analytics in agri-
culture. Recent and relevant works are considered.

3.1 Case Study-1: Big Data Querying

The domain related to agriculture and environment consists of datasets which pos-
sess diverse and large formats for content, storage, and structure with respect to
research purpose. Recently, many datasets are made open source by making them
a part of cloud related to Linked Open Data (LOD) which has formed data sources
that are heterogeneous and distributed in nature. However, accessing the LOD is
not easy and is not able to meet with the ever-evolving changes in data science. It
requires extensive big data querying facilities. In this case study, we study a research
project called SemaGrow that was initiated by ER FP7 for providing efficient ways of
accessing distributed data. Its main objective was the development of infrastructure
and algorithms for querying the distributed data sources such and the nodes of LOD
in an effective and efficient manner. It utilized the concepts of Semantic Web such as
SPARQL query language and RDF that facilitated the reuse and data sharing among
applications and scientific communities. It permitted the data interconnection in the
cloud related to LOD.
The particular challenges of agricultural data were handled by the SemaGrow
system by utilizing different use cases through sequence of data pilots related to
the agricultural domain. The merging of big data related to spatiotemporal datasets
which were very dissimilar along with its discovery was some of the challenges
Data Analysis and Prediction Using Big Data … 213

handled by SemaGrow. The modeling of climatic conditions related to agriculture

was one of the use cases handled by SemaGrow which inherently dealt with the
adaptation of the system to climatic changes. Large files having multiple dimensions
were utilized for storing the parameters of climate that were needed for modeling the
regional parameters. For this case, the spatial coverage related to different regions
and coverage of temporal conditions for long term was utilized. In many research
experiments, scientists try to produce redundant data for experimentation purpose.
This leads to inefficiency of resource usage and poses serious problems in regions
with limited storage, computing, and networking facilities. The SemaGrow was able
to generate lightweight datasets utilizing efficient querying methods related to merg-
ing, thematic, temporal, and spatial conditions. For instance, the SemaGrow system
was able to generate a combined dataset of soil, temperature, and precipitation param-
eters for an agricultural scientist in a specific region of Ghana with limited storage
and networking facilities.
SemaGrow mainly aimed at providing seamless access to big heterogeneous data
sources to the agricultural scientists allowing them to combine and download hetero-
geneous datasets. The problems related to data integration were reasonably resolved,
thus reducing the time and labor involved in the manual integration of big data. The
system was able to provide valuable information on time to the stakeholders which
aided their decision making. It focused on the layers of data and information of the
DIKW model to reduce the gap among these levels by coordinating and processing
the big data sources at the information level. Subsequently, it was able to reduce the
efforts and resources needed for generating information from raw data and solving
the integration challenges among heterogeneous data sources.

3.1.1 Methodology

The system of Trees4Future Clearinghouse was modified to work utilizing the tech-
nologies of SemaGrow for better big data querying capabilities for real-world agri-
cultural applications. For this reason, the back-end of Trees4Future was substituted
with SemaGrow stack and RDF databases comprising of data and metadata. Hence,
the application was able to perform semantic queries on metadata and actual data. A
set of heterogeneous data nodes which are distributed can be efficiently queried by
the SemaGrow Stack which can be found at the URL https://github.com/semagrow/
semagrow. The Stack is a query processor related to the federated SPARQL. A
query planner is included that optimizes the execution of query by utilizing the
metadata about the nodes. The Sevod vocabulary located at the Web site http://www.
w3.org/2015/03/sevod was established in the project in which the VoID vocabulary
is extended by including statistical information similar to histograms of database.
The data nodes that are slow are handled by utilizing the reactive software model in
the Stack. Thus, without the knowledge about the heterogeneous schemas the Stack
provides querying of the triple stores present underneath in a transparent fashion.
214 K. Aditya Shastry and H. A. Sanjay

By utilizing the initiatives for data coordination taken by ISIMIP and AgMIP, the
triple stores were established. To demonstrate the influence of global warming at sev-
eral levels, the Inter-Sectorial Impact Model Inter-comparison Community Project
was established. The format of NetCDF files was utilized as per the conventions of
climate and forecast for making the I/O data from ISIMIP available. Better economic
and crop models are produced by AgMIP for assessing the impact of climate for sub-
sequent generations by connecting the communities related to economic, climate, and
crop. Current information technology is being utilized for this purpose. The JSON
format is utilized for representing data in AgMIP by utilizing the ICASA Variable
List with regard to its metadata. The triplification of limited datasets into triple stores
is performed for the data from dissimilar sources. This enables the SPARQL to query
the data.
Finally, to resolve spatial queries like point-in-polygon a spatiotemporal triple
store called Strabon present at URL http://strabon.di.uoa.gr is added to the combined
nodes. Middleware software is utilized for the connection between the front-end of the
demonstrator and the Stack instance of SemaGrow. It functions as a pre-processor of
results of raw results of query into a response that can be handled by the demonstrator.
It also functions as a translator for translating the requests of URL into the appropriate
queries of SPARQL for the Stack. To efficiently serve the requirements of the users
using the system, the Strabon creates usable NetCDF files.

3.1.2 Results

Less number of users used the demonstrator application for the purposes of testing.
The users gave positive feedback regarding its functionality. However, better perfor-
mance was expected by the users in case of searches related to metadata which took
approximately 5e30 s. Similarly, the download provided by the system was very slow
(taking days) and the users expected the download to complete within 30 min. This
requires up scaling the infrastructure to a massive level. However, research experts
are of the opinion that although the system takes several days, it is faster compared to
the manual approach which involves repeated consultation, cooperation with remote
and local experts which are still more time-consuming. Hence, the hours or days
taken by the fusion queries are faster than the manual process.

3.2 Case Study-2: Data-Driven Discovery

The Trees4Future project available at the URL www.trees4future.eu represents Euro-

pean research project whose objectives are to combine, develop, and enhance the
infrastructures related to key forest genetics and research in forestry. The forestry
data is made available to a broader audience of agriculturists and decision makers
through this project. To do their work, the researchers of forestry collect data using
their peer and scientific networks. However, recently the distribution of forestry data
Data Analysis and Prediction Using Big Data … 215

is being performed particularly in the domain of genetics. Nonetheless, data storage

for relevant data is done in silos including private or local databases. Also, documen-
tation of metadata is not done.
In several instances, data is not documented by researchers or data is made private
since the associated results of research are not published or due to the chance of data
misuse. Also, data sharing through networks which are trusted mostly offers no ben-
efits. Hence, valuable data for agricultural research purposes is not available easily.
Forestry modelers form a research community in the Trees4Future project. Integrated
modeling is required to meet the present day challenges related to agriculture such
as change in climate. Model combinations which explain several subdomains and
provide coverage to temporal and spatial resolutions are required for estimating the
climatic changes along with determining the adaptation methods for climatic con-
ditions. To establish these integrated models, the Trees4Future has linked ForGEM
[39], EFISCEN [40], and Tosia models. The estimation of responses with respect to
genetic adaptation on population and individual tree levels are done by the ForGEM
model. The development of resources with respect to forests is projected by the
EFISCEN model on European and regional scale. The impacts on the production
chains of forestry wood with respect to social impacts, environmental, and economic
changes are analyzed by the Tosia model.
The problems with integrated modeling are mainly related to heterogeneous data
that come to various sources such as genetics, soil, statistics, market, and climate. It is
hard to access this kind of data for forestry purposes. This problem is resolved by the
Trees4Future project which is capable of providing technical solution that aids the
publication, discoverability, and documentation of forestry data by establishing an
infrastructure for forestry data. Besides, this infrastructure is able to provide benefits
for data sharing and documentation related activities.
The issues related to quality and trust also needs to be handled in this project. From
the perspective of the data owner, the data should not be misused after sharing. From
the consumer point of view, correct and complete data should be provided. The data
variety issues are to be handled since data can originate from various heterogeneous
resources and subdomains. The metadata has to be properly linked with the actual
data semantically.

3.2.1 Methodology and Implementation

The development of data discovery and search was done for improving the data
access. The chief aim was to provide wide access to already documented datasets as
well as to the datasets that were inaccessible. Improved facilities were provided to
the end users to search the datasets easily and discover the published forestry data.
This search mechanism was user-friendly for both the end users of the system and
the data publishers/data owners.
The workflow established in Trees4Future project is illustrated in Fig. 2.
216 K. Aditya Shastry and H. A. Sanjay

Search of
semantics &
discovery

End Users
Semantic
Tagging

(external)
ontologies

Metadata
Harvesting catalogues

Editor related Data

to Metadata Providers

Fig. 2 Publication process workflow developed in Trees4Future [2]

• To aid metadata specific to forestry, a brief schema of metadata is required which

is centered on Dublin Core standard.
• The metadata records are published via the OAI-PMH protocol, which consists
of a metadata editor (online) and embedded repository for delivering an endpoint
that is standardized and harvestable.
• The datasets are conceptualized and linked with external ontologies like
AGROVOC, genetic traits by the forestry ontology.
• A mechanism for annotation, harvesting, and triplification for assisting the gather-
ing of metadata using the metadata forestry schema and metadata (standardized)
is provided.
• INSPIRE and ISO schemas are provided that are utilized to decompose the meta-
data into concepts of ontology using the NLP strategies. These are then stored
in RDF database and linkage between external ontologies and derived dataset
concepts are done.
• A mechanism and user interface for efficient searching of datasets that are regis-
tered by utilizing the semantics stored in the RDF store.

3.2.2 Results

The infrastructure that was implemented provided efficient mechanism for discov-
ering research data related to forestry and provided access to a wider audience. The
infrastructure is capable of handling the heterogeneous data of different variety that
Data Analysis and Prediction Using Big Data … 217

is required for integrated modeling in forestry such as the use case on climate adap-
tation that was discussed before. The major advantage is that the system is able to
integrate the scattered forestry data under one roof. More than 300 datasets from data
repositories of Europe are published by the system. An access point is also made
available for small companies and individuals for publishing their metadata. Further-
more, opportunities to publish reference datasets are provided to fresh modelers for
experimentation on integrated modeling.
The linkage of concepts of metadata with concepts of external ontology is com-
plicated owing to the shortage of amount and quality of the metadata. The automatic
selection of datasets that suit the needs of researchers like in technical domains
related to simulation and modeling is also performed by the system. The complex
nature of scientific datasets is not captured properly due to the absence of depth and
structure in metadata. Frequently utilized and important fields such as lineage do
not offer the required structure for addressing the data production processes which
are complex in nature. The structured addressing of data is prevented by the lack of
depth in the metadata structure. In the presented system, this issue was resolved by
linking the fragments of AGROVOC vocabulary that were isolated with comprehen-
sive subdomain with semantics that is specific. Apparently, this presents an elaborate
and customized mechanism and is not a generic solution.

4 Research Areas Associated with Analytics of Big Data

for Agriculture

This section describes the different open problems with solutions that exist for the
analysis of big data for agriculture.

4.1 Open Problems

Certain problems are created due to the application of techniques related to analysis
of big data in the agricultural domain. Some of the key problems are listed below
• Monopoly of agriculture data by only big agriculture organizations can take place
due to the expense involved in data storage and its distribution. This leads to a high
farmer dependency on large agricultural organizations as observed by the author
[41]. The dissemination of technology to smaller organizations may be restricted
due to the monopoly of big data by the larger organizations [42].
• The ownership issues due to the ambiguity created as to who will be the data owner
along with the related monetization issues may arise as suggested by Nandyala
and Kim [43]. The misuse of farming information by seed organizations and farms
may arise as observed by Shin and Choi [44]. Unnecessary speculation in markets
218 K. Aditya Shastry and H. A. Sanjay

related to commodity may be created by hedge funds due to the availability of

real-time agricultural data [21].
• Several issues related to the security, access, and accuracy of big data during its
analytics and collection may arise as elaborated by Nandyala and Kim [43] and
Sykuta [41].
• Furthermore, digital divide amid developing and developed countries is created
because of unbalanced technology access like software, computing power, band-
width of Internet, and lack of experts [12]. Also, the crucial benefits of big data may
be obtained only if the farmers are well educated with sufficient digital exposure
which is lacking in developing countries suggest [30–45].
• The developers of products have access to limited ground truth information [46].
But, the information on ground truth is essential for the product evaluation and
services operating under several settings related to weather or physical conditions.
Likewise, the large data visualization is quite complex.

4.2 Obstacles for Adopting Analysis Techniques for Big Data

in Agriculture

Some of the obstacles that come in the way of using techniques related to the big
data analysis are listed below:
Lack of human expertise and resources [45]
• Lack of reliable infrastructures for gathering and analyzing big data [20]. Infras-
tructures are required for the efficient storage of big data and management of
high-dimensional datasets.
• Lack of technical expertise and skills in the analysts for the accurate analysis of
big data [12, 22].
• The authors [43] and [16] pointed out that there is an absence of proper governance
and structure associated with big data.
• To provide fair share among the agriculture stakeholders attractive business models
are required as suggested in the works [35–48].
• Owing to the absence of definite semantics, the reuse and knowledge of big data
have become difficult for organizations and researchers.
The authors [1] observed in their study (Table 1) that focus is provided to large
volumes in case of big data such as the applications related to changes in climate and
weather, identification of land, decision making of farmers, finance and insurance,
and remote sensing. However, the other aspects of big data such as velocity, veracity,
valorization, and variety are just as significant as the volume of data. Ignoring these
features has led to the narrow and skewed perception regarding big data in agriculture.
Furthermore, the systems related to remote sensing for management of farms still
suffer from various practical difficulties like the gathering and distribution of images
within the specified time duration, errors related to sampling, and the absence of
Data Analysis and Prediction Using Big Data … 219

resolution data related to spatial. The challenges related to interpretation of images,

extraction of data, and the impact of conditions related to weather also prevail. Lastly,
the common obstacles observed by many agricultural researchers are the lack of the
data (or some part of it) along with its unreliable nature. The relevance of the data
with regard to time and variety also has come across as major challenges faced by
analytics of big data in agriculture.

4.3 Addressing Open Problems and Overcoming Barriers

From the perspective of society and political views, several agriculturists (especially
farmers) around the globe began to establish cooperative societies and online groups
in order to share their experiences about farming along with the knowledge of big data.
Proper policies regarding ownership of data need to be in place in order to safeguard
the copyrights of data owners and govern the user access to data. Furthermore,
policies to manage the security and management of data are required in order to
secure the big data and increase its impact by utilizing appropriate tools. Large
monetary investments are needed for building agriculture cloud infrastructures for
storing, visualizing, and analyzing the data related to agriculture. It is also required
for the analytics to be performed at a faster speed and larger scale. Agricultural data
should be available at reasonable rates or better be made open source for aiding
valuable research. User-friendly platforms may be created by utilizing aggregation,
reduction, and appropriate data analysis.
For data interoperability, technologies related to data semantics such as RDF,
linked data along with ontologies may be utilized. Certain open-source tools like
maps for identifying crop types and calendars, indices of vegetation, models to predict
yields, estimation of crop area, and prediction of seasonal weather may be useful
in the agriculture domain. To assist large-scale operations along with analysis of
varied data, it should be possible to merge these tools. As discussed before, big data
in agriculture needs to be made easily available for extensive research to happen.
With regard to this, several institutions and organizations along with agricultural
universities throughout the world have started providing free access to large-scale
agricultural data covering various domains of agriculture.

4.4 Possible Application Areas Related to the Big Data

Analysis in Agriculture

The probable areas of the application of analytical techniques in big data for resolving
several problems related to agriculture are discussed in this subsection. Some of the
potential application areas are listed below:
220 K. Aditya Shastry and H. A. Sanjay

1. To meet global standards for agriculture, effective platforms facilitating supply

chain stakeholders to have access to products of high-quality and processes
need to be developed [14, 45].
2. Better yield prediction tools along with forecasting accurate demands need to
be implemented in the future [22].
3. Accurate information on the usage of herbicides, fertilizers, and pesticides needs
to be provided to farmers for better crop growth [45].
4. To improve food tracking in supply chain equipments related to scanning in
plants, tracking of shipment, and monitoring the purchases of customers need
to be developed.
5. Foodborne diseases need to be prevented via the utilization of big data analysis
strategies in agriculture.
6. Efficient techniques for optimization are required since the agriculture products
tend to deteriorate after harvesting [48]. This is very much needed to reduce the
losses and enhance the quality of food products [17]. The popular optimization
methods like genetic algorithms [17], neural networks [49], and meta-heuristics
are being applied to processing of food.
7. Efficient remote sensing techniques are very much essential for mapping large-
scale lands to crops. This is required for the effective monitoring of cultivated
lands around the globe which can lead to better productivity as suggested by
Schuster [21], Becker-Reshef et al. [47], Barrett et al. [50], Waldhoff et al. [51].
8. Advanced scientific and simulation models for imitating the environmental
process need to be developed. These models are beneficial for policy mak-
ers in making agriculture-related decisions to achieve sustainability in physical
ecosystems [16, 27, 52].
9. To effectively capture the interaction between environment and plants, screening
techniques possessing high throughput are required. These techniques need to
provide higher accuracy and precision with the capability of analyzing the data
in a quantitative fashion [6, 53].
10. Automatic identification and removal of weeds [38], identification and pre-
vention of pests [54], and crop harvesting [51] may be done by developing
agricultural robots. These robots have the capability to progressively change
agriculture from its existing uncertainty to a more certain future.
11. Specific genetic engineering known as “genome editing” may be utilized to
modify the genome of crop or animal up to the level of each genetic letter as
observed by Hartung and Schiemann [37]. It is able to replicate the mutation
process on which the breeding of crops is reliant. Also, it does not cause the
generation of cross-genetic animals or plants [55]. The research on epigenetics
can be supplemented by this technology.
The above-mentioned applications are able to produce massive amounts of big data
that could be utilized by creators of policies in the future to balance the production
and consumption of agricultural products. Furthermore, making the agricultural data
open source would generate valuable research and development opportunities for
smart farming in the agriculture sector.
Data Analysis and Prediction Using Big Data … 221

5 Conclusions

A detailed discussion about the different techniques of big data analysis and its
applications in agriculture is provided in this chapter. Detailed results regarding
the techniques employed in various applications are also elaborated. The reader of
this chapter can be benefitted by the knowledge about the current projects going on
in the domain of agriculture by utilizing the techniques of big data analytics. The
application of appropriate characteristics of big data for diverse scenarios along with
the sources of big data and analytical techniques for big data are elaborated. The open
research problems along with the challenges faced for applying big data techniques
in agriculture are also discussed. As discussed, the openness and availability of
software and hardware, tools and techniques for analysis of big data, and increased
accessibility of open-source agricultural datasets can bring about radical changes in
how farming is done. Many start-up companies are coming up for performing analysis
of big data in agriculture. The development and research opportunities in agriculture
can be boosted by adopting open standards for datasets. The food shortage can be
effectively resolved by balancing the demand and supply of food through big data
analytical techniques. This in turn can protect the physical ecosystems and assist in
safeguarding our natural resources.

References

1. Kamilaris, A., Kartakoullis, A., Prenafeta-Boldú, F.X.: A review on the practice of big data
analysis in agriculture. Comput. Electron. Agric. 143, 23–37, ISSN 0168-1699 (2017). https://
doi.org/10.1016/j.compag.2017.09.037
2. Lokers, R., Knapen, R., Janssen, S., van Randen, Y., Jansen, J.: Analysis of big data technologies
for use in agro-environmental science. Environ. Modell. Softw. 84, 494–504, ISSN 1364-8152
(2016). https://doi.org/10.1016/j.envsoft.2016.07.017
3. Chi, M., Plaza, A., Benediktsson, J.A., Sun, Z., Shen, J., Zhu, Y.: Big data for remote sensing:
challenges and opportunities. Proc. IEEE 104(11), 2207–2219 (2016). https://doi.org/10.1109/
JPROC.2016.2598228
4. Rodriguez, D., de Voil, P., Rufino, M.C., Odendo M., van Wijk, M.T.: To mulch or to munch?
Big modelling of big data. Agric. Syst. 153, 32–42, ISSN 0308-521X (2017). https://doi.org/
10.1016/j.agsy.2017.01.010
5. Hashem, I.A.T., Yaqoob, I., Anuar, N.B., Mokhtar, S., Gani, A., Khan, S.U.: The rise of “big
data” on cloud computing: review and open research issues. Inf. Syst. 47, 98–115, ISSN 0306-
4379 (2015). https://doi.org/10.1016/j.is.2014.07.006
6. Karmas, A., Karantzalos, K.: Athanasiou, S.: Online analysis of remote sensing data for agri-
cultural applications. s.l., OSGeo’s European conference on free and open source software for
geospatial (2014)
7. Wolfert, S., Ge, L., Verdouw, C., Bogaardt, M.J.: Big data in smart farming—a review. Agric.
Syst. 153, 69–80, ISSN 0308-521X (2017). https://doi.org/10.1016/j.agsy.2017.01.023
8. Sonka, S.: Big data: fueling the next evolution of agricultural innovation. J. Innov. Manag. 4,
114–136 (2016). https://doi.org/10.24840/2183-0606_004.001_0008
9. Tripathi, S., Srinivas, V.V., Nanjundiah, R.S.: Downscaling of precipitation for climate change
scenarios: A support vector machine approach. J. Hydrol. 330(3–4), 621–640, ISSN 0022-1694
(2006). https://doi.org/10.1016/j.jhydrol.2006.04.030
222 K. Aditya Shastry and H. A. Sanjay

10. McQueen, R., Garner, S.R., Nevill-Manning, C.G., Witten, I.H.: Applying machine learning
to agricultural data. Compt. Electron. Agric 12(1) (1995)
11. GSMA.: mAgriProgramme. [Online] Available at http://www.gsma.com/
mobilefordevelopment/programmes/magri/programme-overview Accessed 2019 (2014)
12. Syngenta Foundation for Sustainable Agriculture.: FarmForce [Online]. Available at http://
www.farmforce.com/. Accessed 2018 (2016)
13. Global Envision.: Unleashing Ugandan farmers’ potential through mobile phones.
[Online] Available at https://www.mercycorps.org/research-resources/unleashing-ugandan-
farmers-potential-through-mobile-phones. Accessed 2017 (2006)
14. RIICE Partnership.: Remote sensing-based Information and insurance for crops in emerging
economies. [Online] Available at: http://www.riice.org/. Accessed 2019 (2014)
15. Wardlow, B.D., Egbert, S.L., Kastens, J.H.: Analysis of time-series MODIS 250 m vegetation
index data for crop classification in the US central great plains. Remote Sens. Environ. 108(3),
290–310 (2007)
16. Nativi, S., Mazzetti, P., Santoro, M., Papeschi, F., Craglia, M., Ochiai, O.: Big data challenges
in building the global earth observation system of systems. Environ. Modell. Softw. 68, 1–26,
ISSN 1364-8152 https://doi.org/10.1016/j.envsoft.2015.01.017 (2015)
17. FernándezPierna, J.A., Baeten, V., MichotteRenier, A., Cogdill, R.P., Dardenne, P.: Combina-
tion of support vector machines (SVM) and near-infrared (NIR) imaging spectroscopy for the
detection of meat and bone meal (MBM) in compound feeds. J. Chemom. 18(7–8), 341–349
(2004)
18. Gutiérrez, P.A., López-Granados, F., Peña-Barragán, J.M., Jurado-Expósito, M., Hervás-
Martínez, C.: Logistic regression product-unit neural networks for mapping Ridolfiasegetum
infestations in sunflower crop using multitemporal remote sensed data. Comput. Electr. Agric.
64(2), 293–306, ISSN 0168-1699, (2008). https://doi.org/10.1016/j.compag.2008.06.001
19. Chedad, A., Moshou, D., Aerts, J.M., Van Hirtum, A., Ramon, H., Berckmans, D.: AP—
animal production technology: recognition system for pig cough based on probabilistic neural
networks. J. Agric. Eng. Res. 79, 449–457 (2001). https://doi.org/10.1006/jaer.2001.0719
20. Akinboro, B.: Bringing mobile wallets to nigerian farmers. [Online] Available at http://www.
cgap.org/blog/bringing-mobile-wallets-nigerian-farmers. Accessed 2018 (2016)
21. Schuster, J.: Big data ethics and the digital age of agriculture. Am. Soc. Agric. Biol. Eng. 24(1),
20–21 (2017)
22. Frelat, R., Lopez-Ridaura, S., Giller, K.E., Herrero, M., Douxchamps, S., Djurfeldt, A.A.,
Erenstein, O., Henderson, B., Kassie, M., Paul, B.K., Rigolot, C., Ritzema, R.S., Rodriguez,
D., van Asten, P.J.A., van Wijk, M.T.: Proc. Nat. Acad. Sci. Jan 2016, 113(2), 458–463 https://
doi.org/10.1073/pnas.1518384112
23. Marcot, B.G., Holthausen, R.S., Raphael, M.G., Rowland, M.M., Wisdom, M.J.: Using
Bayesian belief networks to evaluate fish and wildlife population viability under land man-
agement alternatives from an environmental impact statement. Forest Ecol. Manag. 153(1–3),
29–42, ISSN 0378-1127 https://doi.org/10.1016/S0378-1127(01)00452-2 (2001)
24. Field to Market.: Fieldprint calculator. [Online] Available at https://www.fieldtomarket.org/
fieldprint-calculator/. Accessed 2018 (2015)
25. Tesfaye, K., Sonder, K., Cairns, J.E., Magorokosho, C., Tarekegn, A., Kassie, G.T., Getaneh,
F., Abdoulaye, T.A., Abate, T.B., Erenstein, O.:. Targeting drought-tolerant maize varieties in
Southern Africa: a geospatial crop modeling approach using big data (2016)
26. Fuchs, A., Wolff, H.: Drought and retribution: evidence from large scale rainfall index insurance
in Mexico. s.l., s.n., pp. 13–14 (2011)
27. Schnase, J.L., Duffy, D.Q., Tamkin, G.S., Nadeau, D., Thompson, J,H., Grieg, C.M., McInerney,
M.A., Webster, W.P.: MERRA analytic services: meeting the big data challenges of climate
science through cloud-enabled climate analytics-as-a-service. Comput. Environ. Urban Syst.
61, 198–211, ISSN 0198-9715 (2017). https://doi.org/10.1016/j.compenvurbsys.2013.12.003
28. Ma, C., Zhang, H.H., Wang, X.: Machine learning for big data analytics in plants. Trends Plant
Sci. 19(12), 798–808, ISSN 1360-1385 (2014). https://doi.org/10.1016/j.tplants.2014.08.004
Data Analysis and Prediction Using Big Data … 223

29. Mucherino, A., Papajorgji, P., Pardalos, P.: Data mining in agriculture. Springer Science and
Business Media 34 (2009)
30. Kempenaar, C., Lokhorst, C., Bleumer, E.J.B., Veerkamp, R.F., Been, T., van Evert, F.K.,
Boogaardt, M.J., Ge, L., Wolfert, J., Verdouw, C.N., van Bekkum, M., Feldbrugge, L., Jack,
P.C., Waaij, B.D., van Persie, M., Noorbergen, H.: Big data analysis for smart farming, s.l.:
Wageningen University and Research 655 (2016)
31. Armstrong, L., Diepeveen, D., Maddern, R.: The application of data mining techniques to
characterize agricultural soil profiles. Proc. AusDM’07 Proc. Sixth Austr. Conf. Data Min.
Anal. 70, 85–100 (2007)
32. Meyer, G.E., Neto, J.C., Jones, D.D., Hindman, T.W.: Intensified fuzzy clusters for classifying
plant, soil, and residue regions of interest from color images. Comput. Electr. Agric. 42(3),
161–180, ISSN 0168-1699 (2004). https://doi.org/10.1016/j.compag.2003.08.002
33. Ishimwe, R., Abutaleb, K., Ahmed, F.: Applications of thermal imaging in agriculture—a
review. Adv. Rem. Sens. 3, 128–140 (2014). https://doi.org/10.4236/ars.2014.33011
34. Lucas, M.T., Chhajed, D.: Applications of location analysis in agriculture: a survey. J. Oper.
Res. Soc. 55(6), 561–578 (2004)
35. De Leeuw, J., Vrieling, A., Shee, A., Atzberger, C., Hadgu, K.M., Biradar, C.M., Keah, H.,
Turvey, C.: The potential and uptake of remote sensing in insurance: a review. Rem. Sens.
6(11), 10888–10912 (2014)
36. Vibhute, A., Bodhe, S.K.: Application of image processing in agriculture: a survey. Int. J.
Comput. Appl. 52, 34–40 (2012). https://doi.org/10.5120/8176-1495
37. Hartung, F., Schiemann, J.: Precise plant breeding using new genome editing techniques: oppor-
tunities, safety and regulation in the EU. Plant J. 78(5), 742–752 (2014)
38. Blue River Technology.: [Online] Available at http://www.bluerivert.com/. Accessed 2019
(2011)
39. Kramer, K., Hengeveld, G.M., Schelhaas, M.J., van der Werf, D.C., de Winter, W.: Genetic
adaptive response: missing issue in climate change assessment studies (2013)
40. Nabuurs, G.J., Schelhaas, M.J., Pussinen, A.: Validation of the European forest information
scenario model (EFISCEN) and a projection of Finnish forests (2000)
41. Sykuta, M.: Big data in agriculture: property rights, privacy and competition in ag data services.
Int. Food Agribus. Manage. Rev. 19 (2016)
42. Carbonell, I.: The ethics of big data in big agriculture. Int. Policy Rev. 5(1), 1–13 (2016)
43. Nandyala, C.S., Kim, H.-K.: Big and meta data management for U-agriculture mobile services.
Int. J. Softw. Eng. Its Appl. (IJSEIA) 10(1), 257–270 (2016)
44. Shin, D.-H., Choi, M.J.: Ecological views of big data: perspectives and issues. Telematics
Inform. 32(2), 311–320 (2015)
45. Sawant, M., Urkude, R., Jawale, S.: Organized data and information for efficacious agriculture
using PRIDE model. Int. Food Agribus. Manag. Rev. 19(A) (2016)
46. Atzberger, C.: Advances in remote sensing of agriculture: Context description, existing opera-
tional monitoring systems and major information needs. Remote Sens 5(2), 949–981. aWhere
Inc., 2015. [Online] Available at http://www.awhere.com/. Accessed 2018 (2013)
47. Becker-Reshef, I., Justice, C., Sullivan, M., Vermote, E., Tucker, C., Anyamba, A., Small,
J., Pak, E., Masuoka, E., Schmaltz, J., Hansen, M., Pittman, K., Birkett, C., Williams, D.,
Curt, R., Doorn, B.: Monitoring global croplands with coarse resolution earth observations:
the global agriculture monitoring (GLAM) project. Remote Sens 2 (2010). https://doi.org/10.
3390/rs2061589
48. Wari, E., Zhu, W.: A survey on metaheuristics for optimization in food manufacturing industry.
Appl. Soft Comput. 46, 328–343 (2016)
49. Erenturk, S., Erenturk, K.: Comparison of genetic algorithm and neural network approaches
for the drying process of carrot. J. Food Eng. 78(3), 905–912 (2007)
50. Barrett, B., Nitze, I., Green, S., Cawkwell, F.: Assessment of multi-temporal, multi-sensor
radar and ancillary spatial data for grasslands monitoring in Ireland using machine learning
approaches. Remote Sens. Environ. 152(2), 109–124 (2014)
224 K. Aditya Shastry and H. A. Sanjay

51. Waldhoff, G., Curdt, C., Hoffmeister, D., Bareth, G.: Analysis of multitemporal and multisensor
remote sensing data for crop rotation mapping. Int. Arch. Photogramm. Remote Sens. Spat.
Inf. Sci. 25(1), 177–182 (2012)
52. Song, M.-L., Fisher, R., Wang, J.-L., Cui, L.-B.: Environmental performance evaluation with
big data: theories and methods. Ann. Oper. Res. 1–14 (2016)
53. Furbank, R.T., Teste, M.: Phenomics—technologies to relieve the phenotyping bottleneck.
Trends Plant Sci. 16(12), 635–644 (2011)
54. PEAT UG.: Plantix [Online] Available at http://plantix.net/. Accessed 2017 (2016)
55. González-Recio, O., Toro, M., Bach, A.: Past, present and future of epigenetics applied to
livestock breeding. Front. Genet. 6, 305 (2015)
Generic IoT Platform for Analytics
in Agriculture

Balasubramani Pradeep, R. Balasubramani , Jason Elroy Martis

and M. S. Sannidhan

Abstract Recently, it has become trend to use Internet of things for our day-to-day
life, wherein we use the setup of various devices for the large collection of data and
its transmission in various circumstances. It is actually made possible and success
due to the involvement of different sensors that are capable of collecting the type
and amount of data required. But many a times when using the setup of Internet of
Applications for any kind of developed applications or an application to be developed,
it is observed that the setup is built for a specified purpose and provides a one-
dimensional application which basically restricts the usage of the developed system
for some other purposes or does not provide a flexibility of extending the application
to add further modules. In regard to this, our research work aims in creating a universal
platform used especially in the field of agriculture for data transmission and reception
from various sensors and plotting the statistical graphs on a dashboard. Apart from
this as purpose of providing easy usage and extensibility smart devices can also be
installed which provides the plug and play capability of our system.

Keywords Internet of things · Smart sensors · Real-time data streaming ·

Agriculture

B. Pradeep
University of Texas at Dallas, Richardson, USA
e-mail: [email protected]
R. Balasubramani (B) · J. E. Martis · M. S. Sannidhan
NMAM Institute of Technology, Nitte-54110, Udupi, Karnataka, India
e-mail: [email protected]
J. E. Martis
e-mail: [email protected]
M. S. Sannidhan
e-mail: [email protected]

© Springer Nature Singapore Pte Ltd. 2020 225

P. K. Pattnaik et al. (eds.), Internet of Things and Analytics
for Agriculture, Volume 2, Studies in Big Data 67,
https://doi.org/10.1007/978-981-15-0663-5_11
226 B. Pradeep et al.

1 Introduction

The Internet of things (IoT) is a processing idea that combines the possibility of
standard physical items being virtually linked with the Internet and having the capa-
bility to organize and collaborate themselves to live gadgets. The IoT buzzword is
gargantuan on the principal grounds, especially in agriculture that can speak to itself
by carefully moving toward an option that is more prominent than being independent
from anyone else [1]. Never again does the query of data relate just to all clients,
however, it is linked with encompassing articles and knowledgeable information.
At present when numerous devices or gadgets combinatory as one, are known to
have a so called “surrounding knowledge”. Internet of things (IoT) does the work of
all the physical gadgets that we can envision in our everyday life for gathering and
communication of information through different stimuli and responses. This makes
life manageable and automated with minimal to no human interruption [2].
Currently, the field of agriculture is more hit upon as people do not get much access
to better and affordable technologies. It is also envisioned that if the trend continues
the field of agriculture will drop down to a point of no return. Countries whose GDP
depends much upon agriculture will be the most affected since 80% of the country
depends upon agriculture. People, nowadays, talk a lot about bringing the word
“smart” to the agriculture and this “Smart Agriculture” has certainly revolutionized
the domain. It could only be done with the help of sensing external stimuli and
communicating effectively to make a knowledgeable decision even at a microlevel
scale. Internet of things immediately comes to the rescue which aids in the creation of
these sensors and various interfaces. The major problem of Internet of things lies in
the crux of interface for communication. Nowadays, developers tweak and develop
application from scratch, which makes the life of software cycle much greater and too
cumbersome. Therefore, there exists an urgent need in creation of a generic platform
for detail communication of measurements from various sensors [3–6].
These sensors effectively communicate to the main controller which plots the
graphs for analytical reasoning. This article focuses on the process of communication
of data from the sensors effectively to the client for analytics and decision making. It
also includes processes to set up a “Smart Switch and plug in which a device can be
plugged in and controlled using a dashboard application after authentication. The data
transmission is also kept confidential and secure to prevent intrusion and preserve
clarity of informational integrity that is being exchanged between an authenticated
client and server [7].

1.1 Internet of Things

The Internet of things (IoT) is the expansion of internet connectivity into numerous
gadgets and everyday use items. The objects are embedded with electrical chips,
connectivity protocols, and other stimuli sensing hardware (such as sensors), and
Generic IoT Platform for Analytics in Agriculture 227

these lively gadgets can interact and interpret with other machines over the cloud
network, and they can be casually supervised and managed.
The definition of the acronym IoT has substantially improvised and has been
perfected due to the convergence of multiple technologies, instant data analytics,
artificial intelligence, commodity stimuli, and embedded system sensors. The age-old
technology of embedded systems, wireless sensor networks, control and automation
systems, and others cohesively help in the contribution to the Internet of things. In
the current market scenario, Internet of things technology is tangibly linked with
merchandises pertaining to the concept of the “smart”, gadgets and products (such
as lighting fixtures, thermal conductive stats, home systems security and movement
detection cameras, and other home appliances) are extensively connected, and can be
coordinated via wide range of extra venous devices associated with that ecosystem,
such as smartphones and intelligent voice activators and assistants [8].
The IoT concept has constantly being criticized closely, especially in regard to
secrecy and confidentiality and privacy concerns related to these devices and their
intention of pervasive presence. It is seen that Internet of things extends to very
magnanimous set of applications for embedded devices. This set of devices are
divided into consumer, industrial, and infrastructure spaces.

1.1.1 Consumer Applications

A very large portion of these smart devices is invented for the sole reason for consumer
sale. They include a very wide array of devices ranging from connected vehicles
to house management devices. Some examples are “SMART HOME”, “GOOGLE
HOME”, etc.

1.1.2 Industrial Applications

The IIoT refers to a wide array of connected sensors, sensitive instrumentation

devices, and other industrial devices inter-networked together with industrial appli-
ances that include manufacture processes and power generation. This internetworked
scenario allows for data assimilation, exchange, and analysis, potentially influencing
advancements in productivity and speed of service as well. The industrial Internet of
things is an evolving distribution-based controlling system which is abbreviated as
DCS. It gives a scope for a higher degree of influential control of cloud computing for
refinement of industrial processes [9, 10]. The IIoT is backed up by mechanisms such
as computer secrecy and security, cloud computing, edge computing devices, mobile
communication technologies, machine-to-machine interactions, three-dimensional
printing, advanced interaction robotics, cognitive computing, and human–computer
interfacing.
228 B. Pradeep et al.

1.1.3 Agriculture

There are innumerable applications in the field farming for Internet of things. They
include stimuli collection such as data on temperature, rains, humidity transitions,
wind conditions, pest control and management, and soil nature content. This infor-
mation is utilized for the purpose of automation on farming methodologies, analyze
and calculate well-informed decisions for the dynamic improvement in quality and
quantity and to minimize risk and waste, and reduce effort required to manage crops.
For example, farmers can now effectively supervise soil temperature and moisture
content of air from long distances and even apply calculated acquired data to perform
micro-precision fertilization programs. In August month of the year 2018, Toyota
introduced a dynamic partnership with Microsoft to create aquatic farming tools
using their web application suite for Internet of things technology. This helped in
generation of methodologies related to water management [11].

1.2 Internet of Things Architecture

See Fig. 1.
Following is the architecture of the Internet of things. The bottom layer proceeds
through a wide variety of environmental stimuli also known as “things” in the IoT
world. These things give a lot of information about the environment as well as help
in building a large dataset of linked informational context. Stage 1 form the primary
stage of the IoT architecture. It includes the sensors which sense the information
omitted by the “IoT things” and converts them into digital data for further processing.

Fig. 1 Architecture of internet of things

Generic IoT Platform for Analytics in Agriculture 229

Stage 2 includes data acquisitions and gateways which assimilate all the data gathered
from various sensors and converts them into appropriate metrics for measurement
and control. The data assimilation and control helps the gateways to visualize data
in plots, sheets, and graphs for further processing. Stage 3 converges to a term called
as “Edge”. The data acquisition from Stage 2 from various gateways is transmitted
to the edge infrastructure, which forms to be a backbone of the IoT framework.
Data to the edge nodes are transferred mostly via (Wi-fi, LoRa, or ZigBee Transfer)
wireless transfer for ease of access. Pre-analytical processing is done if required for
further stages. Stage 4 is the topmost stage. The main processes in the last stage
of IoT architecture converge in the cloud. It directly dives deep into the processing
perspective, having a short follow-up revision for feedback. The data is then brought
back to the world analyzed and well prepared for controlled and detailed assistance
and management [12–14].
Our entire work is divided into two major categories, namely the hardware and the
software section. The prior involves arranging the circuit setup for the data collection,
and we involve a small subset of sensors for the proposed work. The software section
creates a framework or an IoT platform for the user using multi-platform web inter-
facing. It provides much functionality to the user like user access and authentication,
device management and control stats.

2 Related Works

In their research article [15], Heo and Kim proposed an IoT framework that specially
focused on proving the security services. Article intended its work on security layer
to provide security management system. As a part of implementation, researchers
proposed a novel security architecture for an IoT system entitled M2M security
architecture.
Article [16] articulated by Ammar.et al. proposed an IoT framework that focused
on providing simplicity in implementing various IoT applications. In their imple-
mentation, they even focused on providing a strong security aspect for sensitive data
used in their application sets. As an outcome of their work, they proposed a novel
architecture for their framework which included all the essential elements for any
of the application development over an IoT along with a cross-compatible hardware
and essential security concepts.
In their article [17], authors have stressed the importance of security systems for
any of the IoT framework. Apart from this, they have even emphasized the need
of heterogeneous interoperability feature for any IoT framework. In this regard,
they have succeeded in identifying the loopholes involved in proposing a standard
security system. As a contribution to the research community, researchers in their
article provided a set of guidelines that needs to be adopted by any IoT systems in
terms of providing an efficient security aspect.
230 B. Pradeep et al.

Anindya Dey, Kim Stuart, Matthew E. Tolentino in their research article [18]
entitled “Characterizing the Impact of Topology on IoT Stream Processing” pro-
posed a real-time data collecting sensor using a real-time platform system for an IoT
named as “Namatad”. As an outcome of this article, researchers were successful in
identifying a topology for routing in an edge-based server with minimal latency and
maximum QoS. They also succeeded in implementing a machine learning model for
their work that can identify an alternative topology with respect to scaling and sensor
deployment.
In their article [19], authors have proposed the implementation of SoC virtual
platform for the IoT terminals. In this case, authors have focused on providing flexible
software development for an IoT platform through SoC virtual platform without
having to involve an actual hardware system that is required. This even provided the
advantage of using any developed software immediately on the board as they are
developed in a uniform environment. Result comparisons were successfully carried
out and verified using a Raspberry Pi board which proved the same result as they
were deployed on the board with a flexibility of development.
Article [20] authored by Dong-Hwan Park et al. proposed on the improvement of
interoperability between any of the IoT devices making it easy for the usage. The
focus was to provide ease of development in semantic IoT services. As a result,
they targeted on developing an open IoT service platform for an easy access, easy
connection and also more importantly an easy development. As an outcome, proposed
system offered various services like proving a sleep satisfaction, sensor observation
system in aggregation, and analysis of sleep pattern system. Ultimately proposed
system provided the advantage of both interoperability and scalability of services on
the IoT platform.
In their research article [21], paper discussed the proposal of implementing an
Android-based IoT platform that could predict the security threats occurring in future.
As a part of implementation, they focused on relying upon the permissions that
arrive before and after update of any software system. Through this, a system was
developed that could identify the security threats occurring based on the differences
in permissions.
In an article [22], authors have discussed about challenges and limitations that
are imposed when implementing a security aspect for an IoT system. According
to their proposed research, securing of an IoT by covering the aspects of storage
system, energy level and a protocol system of lightweight concept is more suitable.
The aspects covered in modifying an existing system as it provided limited level
of resources. Hence this research proposed a security protocol technique Physical
Unclonable Functions (PUF’s) considering the limitation in resources.
In their article [23] entitled “Light Weight Security Framework for IoT” proposed
by Thenmozhi et al. have concentrated on implementing a security system based on
lightweight cryptography. The technique aimed at proposing the lightweight cryp-
tography system due to the availability of limited physical resource configurations
like memory systems, processors, and also power performances. As an outcome of
their research article, they succeeded in implementing a lightweight security sys-
tem with a limited amount of data that ensure the data confidentiality. Experimental
Generic IoT Platform for Analytics in Agriculture 231

results have proved that the proposed system provided a better and more efficient
result when compared to the existing system in terms of security aspects. Apart from
this as a part of ensuring power dissipations, LIFI system was used in place of Wi-fi.
In a research article [24], researchers have MSSDL life cycle which basically
focused on increasing the reliability of software systems in terms of the security
aspects with respect to the bugs that usually encounter. The proposed system intended
in providing a great security system for the internal architecture of the Microsoft
systems. But the major drawback identified in the proposed system was its nature of
being heavyweight to the security aspects.
In their research article [25], authors have closely observed the defects encoun-
tered in the research work have proposed a technique called CLASP which functions
very similar to MSSDL in functionality but a better aspect of the proposed system is
that the system is much more lightweight than MSSDL.
Research article [26] proposed another better technique which is an optimized
version of CLASP. Authors in this research article focused on modifying CLASP to
provide a framework that helps to develop strategic security for application based on
security implementations for business risks.
In a research article [27] entitled “IoTSM: An End-to-end Security Model for IoT
Ecosystems “authored by Joseph Bugeja et al. analyzed various security implication
challenges with respect to IoT systems and to cover different aspects in tandem,
authors have proposed a novel IoT framework that can provide a strategic security
systems based on the organizations requirements. Proposed technique also focuses
on providing an able matured model for the IoT systems that have the capability of
expanding the security practices based on empirical model system.
In their research article [28], the authors have proposed a technique for small and
medium-scale IoT solutions based on the domain-level requirement. In the course,
they were successful in developing an IoT framework that works on the principle of
rapid deployment technique for an industry.
Abbass et al., in [29], have bridged the application of deep learning system to
access the security constraints for IoT intelligently. Basically, they have implemented
a CNN that can improve the ability of security risk evaluation on the basis of contin-
uous learning system. As a positive outcome of the research work, the proposal was
successful in accessing the security threats via deep learning model and proved ably
a better performance when compared to traditional techniques. As a part of negative
aspect, CNN was found to be a black box model and consumed horrendous amount
of time in data collection.

3 Methodology

We propose a generic IoT platform designed especially for agriculture helps us to get
the data from various setups and conversely control the devices from distant places
using the Internet as a medium. The readings of different sensors can be used to
control an air humidifier. In turn it helps in the conservation of resources, which in
232 B. Pradeep et al.

turn drives the development of new technologies. For example, a smart bulb can help
you save and monitor your energy consumption by showing you energy statistics.
Moreover, it will help us to regulate the intensity and heat at which your device
operates which will indeed help you save power as well as give nutrition to plants.
Traditional system designs have been stand alone in their approach providing a spe-
cific type of application to the end user based on the requirements posed. Hence,
the design approach followed the assembly of hardware components and the cor-
responding software interface to be highly specific. Our proposed system aimed at
being flexible in terms of design is also beneficial for budding IoT enthusiasts as it
would help them to gain an in-depth understanding of IoT and encourages to develop
a multi-application system which one can use in their day-to-day life and also worth
for the society in the field of agriculture (Fig. 2).
The above diagram shows an in-depth view of the proposed generic IoT platform.
The area marked in dot-dashed lines marks our area of interest. The framework works
in tandem with the IoT framework but separates it in part on with the authentication
and confidentiality. The data from the sensors are collectively accumulated and trans-
ferred to our platform. This causes the preliminary part in our platform. The device
is authenticated using authentication algorithms in order to verify the authenticity of
the data and is tagged with a subscription Id and stored in the data store. The filter
component is used to verify and strip off the unwanted residual subscription tag as
well as only notify the data which is necessary to the client. The encryption and web
interfacing wing encrypts the message using numerous encryption algorithms and
transfers it to web viewing html format for ease of access viewing to the client in
subscription. The data transfer component is a dynamic query processing for in term
registration and deletion of the device. It also helps in aiding the filtration process to
filter the quality of data given to the client.
The components will be explained in detail under further subsections.

Fig. 2 Architecture for the proposed generic IoT platform for agricultural system
Generic IoT Platform for Analytics in Agriculture 233

3.1 Device Authentication and Registration

This component is a very active component in our platform. Every sensor node
that communicates to our platform has to communicate to the client through it. It
ensures complete confidentiality and sees that no unwanted device can tap through
data to the client. Every sensor is attached an unique Id to it. Initially, we could have
used MAC-ID which is 48 bit in size, but nowadays every device has more than
one mac-id, which poses a problem. In order to overcome this, the platform gives a
unique registration code once the device communicates with the platform. The device
authentication is done by using MAC and SHA-1 algorithms. SHA-1 expanded as
Secure Hash Algorithm-1 as according to its name is an algorithm relied over a
hashing technique. The technique is used for cryptographic applications to ensure
secure transmission. Basically takes some plain text as its input entity and produces
an output which is a hash value of considered input. It takes an input message of
length n bits and produces an output message of length 120 bits. Generated output
is basically known as a message digest [30].

3.2 Subscription List

This component includes a large list of subscribed clients to whom the sensor data
readings are sent. The issue arises when multiple users subscribe to the same sensor
readings. These are achieved by a well-known protocol called Message Queuing
Telemetry Transport (MQTT). MQTT is basically a M2M protocol designed to be a
lightweight method of transport. It is also ideal for mobile applications because of
its small size, low power usage, minimized data packets, and efficient distribution of
information to one or many receivers [31–35]
The main advantages of this protocol are:
1. Better Error Reporting—Very precisely, a logical code is being imbibed
to required responsiveness for publications (PUBACK/PUBREC). A protocol
called MQTT actually came into existence with lots of applicability’s alike sen-
sors works on a pipeline model system—in case of a failure occurring in the
status of publication to be transmitted and delivered, in that awful case, sensor in
action is not going to consider any kind of action from its side. Nonetheless, the
applicability for the protocol under discussion is actually much expansive and an
application on any cell-phone may also have an eagerness to warn the end user of
the system if data required to be transmitted successfully fails to do so. It is very
unusual nowadays to have a so-called return codes on every acknowledgement
that the protocol system generates.
2. Shared Subscriptions—In case the rate of message transmission is considerably
immense, subscriptions which are considered for sharing can actually be adopted
for the sake of balancing the load, the messages across a number of receiving
clients.
234 B. Pradeep et al.

3. Message Properties—Being represented as a Metadata part of an header system

of a message that is meant for the transmission. Ones that are considered are
actually made use to develop other set of features in the list that is under consid-
eration along with that it has to even permit the end users to make use of those
properties that are being defined by the end user, e.g., to help in the encryption
of plaintext by actually intimating the receiving entity about the type of the key
to be applied for the purpose of decryption.
4. Message Expiry—One of the very effective methods that needs to be employed
in which case a preference to abandon the transmitted messages in case if they
fail to get delivered into the destination within a specific slice of time as defined
by the end user.
5. Session Expiry—In case a scenario occurs where a client involved in the com-
munication fails to form a connection within a period of time as defined by
the end user, in such cases evidently state (e.g., messages that are buffered and
subscriptions) can be disposed off without even cleaning up.
6. Topic Alias—One of the systems that actually permits strings in connection to
the topic in communicating messages to be replaced with an individual number,
cropping the amount of data bytes that actually needs to be transmitted in case
if a person in publishing iteratively makes use of the identical topics.
7. Will Delay—This system permits an actual communicating to be published if a
client that is involved in the communication system loses out of a connection for
a time that is higher than the one that is being defined by the end user. Permits
indications regarding the disruption of important applications in connection to
the client end without being beset by positives encountered out of false.
8. Allowed Function Discovery—In the beginning of forming a connection,
restrictions alike maximal size of a packet and the amount of (QoS > 0) messages
in transmission can be transmitted to notify the client end about its responsibility
in terms of communication.
Aforementioned list of subscriptions takes the responsibility of all sensor nodes
and the request/responses at the client end by preserving the required contents from
the data house and reading out the data during the point when it is actually required.

3.3 Filter

This component deals with the chipping off unwanted contents and verification of
the client subscription from the sensor reading and also verifies whether the proper
sensor readings are transferred to the user who has genuinely subscribed to it. It also
forms an intermediate layer between the data store and the web interface layer.
Generic IoT Platform for Analytics in Agriculture 235

3.4 Encryption and Web Interfacing

This layer completely deals with the security aspect of data transmission. It receives
incoming packets of data from the filter component and encrypts the packet for confi-
dentiality, to prevent masquerade and man in the middle attacks on data transmission.
Encryption is done which increases the performance and throughput of encryption
and decryption. Web viewing is used to transmit data to the device from which the
client wants to view. It can be any device and can have multiple displays. The job of
the web view ensures that any resolution display will not hinder the viewing details
of the user. It is done by using NodeJS, JQuery, and Bootstrap. The web interface is
very dynamic and fully fledged to introduce components to fulfill ease of access.
Logically the platform to be truly generic must be strictly classified into three
categories.
1. Sensor nodes: This logical subdivision only acquires data from external stimuli
and transmits them to the IoT Platform. The data transfer is done mainly via
wireless modes of communication. It ensures flexibility and ease of deployment.
The wireless modes of transfer are Wi-fi, LoRa, ZigBee.
2. IoT Platform: This is the main logical subdivision of the three-tiered M2M
architecture of our IoT platform. It includes the data analytics and control mech-
anisms for data acquisition and control. The data store used is implemented in
MongoDB, mainly concentrated on noSQL architecture.
3. Web View: The final sub division of the architecture is the web view. It contains
codes in JQuery, JavaScript, Bootstrap to utilize full functionality of the client
and to reduce the load on the server, bringing in the concept of Fat client. Only
data is pumped to the client and displayed beautifully on the browser. It also
manages multiple views to bring in flexibility.

4 Results and Discussions

We have tested our generic platform in the field of agriculture by sensing the most
common features that farmers depend upon like light ambience, temperature, and
humidity. An accelerometer is also attached to show if the sensor setup has fallen
or toppled over by extreme wind. These sensor’s data are progressively collected
by using a microcontroller like Arduino UNO and then passed on to the NodeMCU
12-E to transmit the data wirelessly to the IoT platform. The schematic of the circuit
is shown below (Fig. 3).
The results are obtained in the form of charts at our mobile end from the web
view as shown below. Note that the user has signaled that the platform should show
the web view in the form of plots (Fig. 4).
We have also added another set of IoT device called an actuator which will perform
a certain action for a certain event. This is used to switch the light on and off for
236 B. Pradeep et al.

Fig. 3 Schematic diagram for sensors showing the connection to the microcontroller

Fig. 4 Dashboard output on our IoT platform

regulation of light intensity of the plants. We call this device “Smart Switch”. The
schematics of the connection are shown below (Fig. 5).
We know that the NodeMCU cannot directly control AC devices; because they
are not configured to do that and the voltages used by AC devices are too high. So,
we uncouple the devices by using a device called a relay. The relay used here is a
solid state relay called S116s02. The NodeMCU acting as a microcontroller signals
the relay by passing a voltage of 5v through an npn transistor which in turn switches
Generic IoT Platform for Analytics in Agriculture 237

Fig. 5 Schematic diagram for “smart switch”

the light. We can also control the intensity of the light as shown in the schematic
below (Fig. 6).
The control of the Smart Switch is given in the dashboard as shown (Fig. 7).
Note that we have configured out platform to set the view of the Smart Switch. It
displays control for On/Off as well as adding features for timers and also measures
the energy consumed by the bulb. The user can directly manage the bulb from a
remote place. It can also be configured if it crosses any threshold kWh.

Fig. 6 Schematic diagram for “smart switch” light intensity regulation

238 B. Pradeep et al.

Fig. 7 Dashboard view for “smart switch”

As you know, the sensor and the user have to register before using our platform
as well as the sensors have to be authenticated. We have given a sign up page which
sends an email with authentication key of the sensor. The front view of the user
dashboard is shown below.
Analysis of the data can also be done from our platform as shown in Figs. 8 and 9,
respectively, depicted below.
Analysis of the data can also be done from our platform as shown in Fig. 10.
The figure depicted above shows how the temperature analysis is done from a
start data to end date. The data from the sensors from our date and time is filtered
accordingly from our platform database and then given to the web view as shown.

Fig. 8 Dashboard view for user

Generic IoT Platform for Analytics in Agriculture 239

Fig. 9 View for user devices registered and their user id’s

Fig. 10 Analysis view for temperature

5 Program Code and Flow

This section shows below the codes for all different components used in our setup.
This section is divided into two parts, i.e., the hardware part code for the sensor and
the software part of the IoT platform.
240 B. Pradeep et al.

5.1 Light Intensity Sensor and Data Transmission

The code shown below symbolizes how the nodeMCU transmits data and how the
Arduino collects the data from the sensor. Note that the sensor is analog in nature
and requires an analog-to-digital converter (ADC), already available in the Arduino
Board.

void intensity( ){

/*

Read the data from analog pin and map them from 0 to 255

*/

Inten=map(analogRead(A0),0,1024,0,255);

/*

map function maps A0 from 1024 values to 255 values only

*/

Serial.println(“Light Intensty is: “);

/*

Serially prints the light intensity with a message

*/
Serial.println(inten);

/*

Prints the light intensity value that is being mapped

*/
Generic IoT Platform for Analytics in Agriculture 241

Code that is used to write the data into the nodeMCU captured by Arduino

Void writedata( )
{

/*
Module to write the data
*/
Serial.write(temp);

/*
serially transmit the temperature from Arduino to
nodeMCU
*/
Serial.write(hum);
/*
serially transmit the humidity from Arduino to nodeMCU
*/

Serial.write(tinten);

/*
serially transmit the light intensity from Arduino to
nodeMCU
*/

Code for nodeMCU transmitting the data to the platform

while(Serial.avalable( ))
{
/*
check for serial availability from the
Arduino
*/
x=Serial.read();
/*
Convert the data to a specified format for
our proposed platform
*/

snprintf(msg.75,”%d”,x);
242 B. Pradeep et al.

if(i==0)
{

/* Encrypt the message using encrypt message

function along with authentication key
*/
encryptMessage(msg,”594a734/4569872156/89values”);
}

5.2 Smart Switch

The Smart Plug performs the following functionalities

i. ON/OFF devices (~100 W) from mobile app
ii. Regulation of light intensity from mobile app
iii. Timer functionality for turning the switch ON/OFF
iv. Energy consumption graph in kWh
v. Indicator or Smart Switch for the following conditions
• Connected
• Connecting
• Disconnected
• Reconfiguration of the NodeMCU in the Smart Switch.
The flowchart of the Smart Switch goes as follows (Fig. 11)

5.3 Data Store

The data store is implemented in Mongo DB and interacted using NodeJS script.
Schemas are created for device authentication and user identification. The schemas
are given below.

5.3.1 Schema for Device Authentication

Below is the code depicted for the design of the schema pertaining to the authenti-
cation required for an IoT device.
Generic IoT Platform for Analytics in Agriculture 243

Fig. 11 Flowchart for the “Smart Switch” working

244 B. Pradeep et al.

var authSchema = mongoose.Schema( {local : {

api_key : String, (a unique string to
identify each user(generated server-side)
email : String, (email id of the user:
verification email will be sent on this mail)
password : String, (password of the user: for
logging in after verification, hashed)
url : String, (url generated server-side to
be sent to the email for verification)
state : Boolean, (If verified: state = true -
> schema is deleted from collection, else
false)},{collection:'authSchema',safe:true});

5.3.2 Schema for User Details

Below is the code depicted for the purpose of storing the required user details in the
database.

local : {api_key : String,a unique string to

identify each user(generated server-side)
email : String, email id of the user:
verification email will be sent on this mail
password : String, password of the user: for
logging in after verification, hashed using
b-crypt node});

5.3.3 Schema for Notification Details

Below is the code depicted for the purpose of sending the notification to the device
that is registered with the proposed system.
Generic IoT Platform for Analytics in Agriculture 245

var notification = mongoose.Schema({time : {type :

Date, default: Date.now}, time of receiving
the notification from the device
message: String message conveyed by the
notification});

5.3.4 Schema for Sensor Details

Below we have depicted a code that is able to store the data collected from the sensor
nodes into our proposed platform data store.

var sensorSchema = mongoose.Schema({device_name :

String,// user inputted string to identify
the device
device_id : { type: String, unique: true },
server generated string to uniquely identify
a device
description : String, user inputted
description of the device
last_access : {type: Date, default:
Date.now}, time of creation of the device
(name is misleading)
device_type : String, user inputted : type of
device
timer : {type:String,default:"false"},
boolean, if timer is set or unset
timerDate : String, if timer is set, the date
to which it is set to
project_id : {type:String,default:""}, if the
device belongs to a project, this field
stores the id of the project it belongs to,
else stores nothing
246 B. Pradeep et al.

intensity : {type:String,default:"0"}, last

set intensity of the device
state : {type: String, default: "false"},
state is true if device is on, else false
activity : [notification],// array of all the
notifications sent by the given device to
the application.
values : [sensorDataSchema],// stores all the
time-value pairs used to plot the data sent
by the device
energy : [ ]// Array of numbers, stores the
energy values for the last 30 days.
// The indices of the array correspond to the
days of a month.
},{collection:'sensor',safe:true});

5.3.5 Schema for Sensor Data

The code that is put up is used for the web view of our proposed platform.

var sensorData = mongoose.Schema({

time : {type : Date, default:
Date.now},timestamp of the instant when the
data value to be plotted is received by the
application
value : Number, data value to be plotted,
sent by the device});

6 Conclusion and Future Work

Internet of things has become a mandatory requirement in our day-to-day lives, and
it is integrating into almost everything that we use every day. The only problem
that arises with this is the communication platform. For this, we propose a novel
solution that is completely generic on both ends, i.e., the objects as well as the users.
This solution provides a straightforward approach which tackles the communication
Generic IoT Platform for Analytics in Agriculture 247

platform scenario as well as introduced security and feasibility. We have also tested
this platform on our own built IoT devices such as Smart Switch and agricultural IoT
sensory devices. The major advantages of this are ease of connectivity for analytics
which even motivates an end user with very minimal knowledge on IoT frameworks.
In future, this work can be extended on all platforms that intensively depend on IoT
devices. We can further incorporate machine learning and deep learning approaches
for automatic analyses which make the system having minimal dependence on end
users.

References

1. Sannidhan, M.S., Martis, J.E. and Bhandary, A.: A cost effective approach for detecting elec-
tricity theft using raspberry pi board. In: 2017 International Conference on Current Trends in
Computer, Electrical, Electronics and Communication (CTCEEC) (2017)
2. Chanthakit, S., Rattanapoka, C.: MQTT based air quality monitoring system using node MCU
and node-RED. In: 2018 Seventh ICT International Student Project Conference (ICT-ISPC),
Nakhonpathom, pp 1–5 (2018)
3. Shenoy, J., Pingle, Y.: IOT in agriculture. In: 2016 3rd International Conference on Computing
for Sustainable Global Development (INDIACom), New Delhi, pp. 1456–1458 (2016)
4. Prathibha, S.R., Hongal, A., Jyothi, M.P.: IOT based monitoring system in smart agriculture. In:
2017 International Conference on Recent Advances in Electronics and Communication Tech-
nology (ICRAECT), Bangalore, pp. 81–84 (2017). https://doi.org/10.1109/icraect.2017.52
5. Jaiganesh, S., Gunaseelan, K., Ellappan, V.: IOT agriculture to improve food and farming tech-
nology. 2017 Conference on Emerging Devices and Smart Systems (ICEDSS), Tiruchengode,
p. 260–266 (2017). https://doi.org/10.1109/icedss.2017.8073690
6. Heble, S., Kumar, A., Prasad, K.V.V.D., Samirana, S., Rajalakshmi, P., Desai, U.B., Sabeena,
S.: A low power IoT network for smart agriculture IEEE 4th World Forum on Internet of Things
(WF-IoT). Singapore 2018, 609–614 (2018). https://doi.org/10.1109/WF-IoT.2018.8355152
7. Chaczko, Z., Braun, R.: Learning data engineering: creating IoT apps using the node-RED
and the RPI technologies. In: 2017 16th International Conference on Information Technology
Based Higher Education and Training (ITHET), Ohrid, pp. 1–8 (2017)
8. Rajalakshmi, A., Shahnasser, H.: Internet of things using node-red and alexa. In: 2017 17th
International Symposium on Communications and Information Technologies (ISCIT), Cairns,
QLD, pp. 1–4 (2017)
9. Abinaya, P., Geetha, C.: Dynamic detection of node replication attacks using X-RED in wireless
sensor networks. In: International Conference on Information Communication and Embedded
Systems (ICICES2014), Chennai, pp. 1–4 (2014)
10. Lekić, M., Gardašević, G.: IoT sensor integration to node-red platform. In: 2018 17th Interna-
tional Symposium INFOTEH-JAHORINA (INFOTEH), East Sarajevo, pp. 1–5 (2018)
11. Kodali, R.K., Anjum, A.: IoT based home automation using node-red. In: 2018 Second Inter-
national Conference on Green Computing and Internet of Things (ICGCIoT), pp. 386–390.
Bangalore, India (2018)
12. Zhong, C., Zhu, Z., Huang, R.: Study on the IOT architecture and gateway technology. 2015
14th International Symposium on Distributed Computing and Applications for Business Engi-
neering and Science (DCABES), Guiyang, pp. 196–199 (2015). https://doi.org/10.1109/dcabes.
2015.56
13. Saadeh, H., Almobaideen, W., Sabri, K.E.: Internet of things: a review to support IoT archi-
tecture’s design. In: 2017 2nd International Conference on the Applications of Information
Technology in Developing Renewable Energy Processes & Systems (IT-DREPS), Amman,
pp. 1–7 (2017). https://doi.org/10.1109/it-dreps.2017.8277803
248 B. Pradeep et al.

14. Navani, D., Jain, S., Nehra, M.S.: The Internet of things (IoT): a study of architectural ele-
ments. In: 2017 13th International Conference on Signal-Image Technology and Internet-Based
Systems (SITIS), Jaipur, pp. 473–478 (2017). https://doi.org/10.1109/sitis.2017.83
15. Heo, S., Kim, H.W.: An analysis of IoT security requirements and communications of the
Korean, pp. 16–22 (2017)
16. Ammar, M., Rusello, G., Crispo, B.: Internet of things: a survey on the security of IoT frame-
works. J. Inf. Secur. Appl. 38, 8–27 (2018)
17. Hwang, I., Kim, Y.-G.: Analysis of security standardization for the internet of things. In: 2017
International Conference on Platform Technology and Service (PlatCon) (2017)
18. Dey, A., Stuart, K., Tolentino, M. E.: Characterizing the impact of topology on IoT stream
processing. In: 2018 IEEE 4th World Forum on Internet of Things (WF-IoT) (2018)
19. Lee, H.-R., Lin, C.-H., Park, K.-H., Kim, W.-J., Cho, H.-J.: Development of SoC virtual plat-
form for IoT terminals based on OneM2M. 2017 International SoC Design Conference (ISOCC)
(2017)
20. Park, D.-H., Bang, H.-C., Pyo, C. S., Kang, S.-J.: Semantic open IoT service platform technol-
ogy. 2014 IEEE World Forum on Internet of Things (WF-IoT) (2014)
21. Park, I.-K., Kwak, J.: Implementation of permission management method for before and after
applications the update in android-based IoT platform environment (2017)
22. Idriss, T., Idriss, H., Bayoumi, M.: A PUF-based paradigm for IoT security. 2016 IEEE 3rd
World Forum on Internet of Things (WF-IoT) (2016)
23. Thenmozhi, S., Ranjitha, J.B., Anu, S., Soniya, A.J.: Light weight security framework for IoT.
In: 2019 5th International Conference on Advanced Computing and Communication Systems
(ICACCS) (2019)
24. Howard, M., Lipner, S.: The security development lifecycle: SDL: a process for developing
demonstrably more secure software, vol. 8. Microsoft Press (2006)
25. Graham, D.: Introduction to the CLASP process (2006) [Online]. Available https://goo.gl/
wducjb [Online]
26. OWASP: Software assurance maturity model. OWASP. [Online] (2018). Available https://goo.
gl/9cCA4h [Online]
27. Bugeja, J., Vogel, B., Jacobsson, A., Varshney, R.: IoTSM: an end-to-end security model for
IoT ecosystems. 2019 IEEE International Conference on Pervasive Computing and Commu-
nications Workshops (PerCom Workshops) (2019)
28. Vakaloudis, A., Oleary, C.: A framework for rapid integration of IoT Systems with industrial
environments. In: 2019 IEEE 5th World Forum on Internet of Things (WF-IoT) (2019)
29. Abbass, W., Bakraouy, Z., Baina, A., Bellafkih, M.: Classifying IoT security risks using deep
learning algorithms. In: 2018 6th International Conference on Wireless Networks and Mobile
Communications (WINCOM) (2018)
30. Chen, H.W., Lin, F.J.: Converging MQTT resources in ETSI standards based M2M platform.
In: 2014 IEEE International Conference on Internet of Things (iThings), and IEEE Green
Computing and Communications (GreenCom) and IEEE Cyber, Physical and Social Comp
31. Tantitharanukul, N., Osathanunkul, K., Hantrakul, K., Pramokchon, P., Khoenkaw, P.: MQTT-
topics management system for sharing of open data. 2017 International Conference on Digital
Arts, Media and Technology (ICDAMT), Chiang Mai, pp. 62–65 (2017)
32. Mishra, B.: TMCAS: An MQTT based collision avoidance system for railway networks. In:
2018 18th International Conference on Computational Science and Applications (ICCSA),
Melbourne, VIC, pp. 1–6 (2018)
33. Bryce, R., Shaw, T., Srivastava, G.: MQTT-G: a publish/subscribe protocol with geolocation.
In: 2018 41st International Conference on Telecommunications and Signal Processing (TSP),
Athens, pp. 1–4 (2018)
34. Sasaki, Y., Yokotani, T., Mukai, H.: Comparison with assured transfer of information mecha-
nisms in MQTT. 2018 International Japan-Africa Conference on Electronics, Communications
and Computations (JAC-ECC), Alexandria, Egypt, pp. 95–98 (2018)
35. Harsha, M.S., Bhavani, B.M., Kundhavai, K.R.: Analysis of vulnerabilities in MQTT security
using Shodan API and implementation of its countermeasures via authentication and ACLs.
In:2018 International Conference on Advances in Computing, Communications
Mission SDGs Through Food Waste
Management: Nature and Approaches

Arindam Chakrabarty and Uday Sankar Das

Abstract The global fraternity has been embarrassed to understand how to feed
the elephantine population of this planet. The greed of comfort, convenience, and
technology ecosystem has mesmerized our life in such a fashion that the human
society has been up-rooted from the nature. The wave of infrastructure development
results in heavy encroachment of fertile land as well as yielding to regressive fertility
of land. On the contrary, the society has not learned to optimize the utilization of
resources whatever forms it may be. These compel the United Nations to formulate
specific target-oriented Sustainable Development Goals that need to be achieved
by 2030. Food waste is still a menace of mankind. It may be of many forms and
dimensions. Food waste exists in every phase of supply chain. Most surprisingly, the
stigma of this menace reaches to our household also. The irony of life is that, on one
hand, we are habituated to accept that wastage in foodstuff as part of our livelihood,
and on the other hand, the United Nations reveals its concern for poverty, hunger, and
many other unaccomplished goals. The world is now on the verge of Fourth Industrial
Revolution. The IoT-based ecosystem has been emerged as an inseparable entity of
the modern societies. This paper has attempted to assess and account the loss of
economy for wastage of food items from global canvas to national perspective. This
study has also focused on how to use IoT platform so that the food wastage can be
reduced up to a considerable amount both in the supply chain and even in household
practices. This research work is based on secondary information like research papers,
reports, and results of other relevant studies. The paper has attempted to develop and
devise a conceptual and strategic model where the IoT ecosystem can be incorporated
to ensure real-time solutions and to curb on massive food wastage practices. If the
model is implemented and practiced with appropriate case specific modifications

A. Chakrabarty (B) · U. S. Das

Department of Management, Rajiv Gandhi University (Central University),
Rono Hills: Doimukh, Itanagar, Arunachal Pradesh 791112, India
e-mail: [email protected]; [email protected]
U. S. Das
e-mail: [email protected]

© Springer Nature Singapore Pte Ltd. 2020 249

P. K. Pattnaik et al. (eds.), Internet of Things and Analytics
for Agriculture, Volume 2, Studies in Big Data 67,
https://doi.org/10.1007/978-981-15-0663-5_12
250 A. Chakrabarty and U. S. Das

and customizations, it would make the economy more efficient and address different
perspectives and dimensions of UN Sustainable Development Goals (UNSDGs) to
a larger extent primarily in Indian Context.

Keywords Food waste management · Supply chain · Fourth Industrial Revolution

IoT ecosystem · UNSDGs · India

1 Introduction

The journey of human civilization started with the ignition of fire, invention of
wheels, and creations of indigenous tools and techniques through the passage of
various ages. Today, we are in the age of information automation and moving toward
Fourth Industrial Revolution. The success of our human civilization is based on key
primary needs, i.e., food, clothing, and shelter. However, this is the irony of life that
in spite of achieving various developmental indicators, the human society is still
unable to fulfill the basic needs like food which is reflected in various literatures,
research outcomes and from the experiences of reality. To cater the critical issues,
the United Nations move ahead from achieving Millennium Development Goals
(MDGs) to Sustainable Development Goals (SDGs) where poverty and hunger were
given primary thrust.

1.1 Food Waste Across the Globe

Around 1.3 billion tonnes of food produced gets wasted globally which account to
around one third of the food produced. A whopping US$680 billion and US$310
billion are lost in the industrialized countries and developing countries, respectively.
A 670 and 630 million tonnes of food are wasted for industrialized countries and
developing countries. Perishable items like fruits and vegetables account for the
highest losses post-harvest. Food loss and food waste per year are roughly account
to 20% of oilseeds, 35% for combined for meat, dairy and fish, 30% for cereals,
and 40–50% for root crops. Per capita food lost or wasted every year is between 95
and 115 kg/year in Europe and North America. While 6–11 kg a year is wasted in
countries of South-Eastern Asia and sub-Saharan Africa. A 40% loss of food items
occur at post-harvest or at processing level in developing nations, and same amount
is lost at retail and consumer levels in industrialized nations (Table 1; Fig. 1).
Mission SDGs through Food Waste Management: Nature and Approaches 251

Table 1 Per capita food losses and waste (kg/per year), at consumption and pre-consumptions
stages, in different regions [1]
Region Production to retailing Consumer
Europe 187 94
North America and Oceania 181 115
Industrialized Asia 165 70
Sub-Saharan Africa 159 7
North Africa, West and Central Asia 181 31
South and Southeast Asia 115 13
Latin America 198 24
Adapted from key finding of Food and Agriculture Organization of the United Nations

350
Consumer
300
ProducƟon to retailing
250
200
150
100
50
0

Fig. 1 Per capita food losses and waste (kg/per year), at consumption and pre-consumptions
stages, in different regions (adapted from key finding of Food and Agriculture Organization of
the United Nations) [2]

1.2 Food Waste in India

India has adorned the developing nation tag since the dawn of independence and is
at the frontiers of food loss and food waste. India is ranked at 103 in the latest Global
Hunger Index [3] report published in 2018 while a large section of the population still
live below poverty line (21.92% of the total population as per RBI data published on
Sep 16, 2015) [4] bureaucratic loopholes and intentional ignorance towards the issue
has kept the problem alive till date. FCI, a premier food distribution corporation and
a state owned utility is yet to fix the grain drain problem. It is reflected in numerous
news reports that the food loss by FCI was in the tune of 1.94 lakh MT food grain that
had been wasted between 2005–2013 [5]. Apart from this, India still lacks organized
252 A. Chakrabarty and U. S. Das

cold storage facilities or processing plants within proximity of farmers producing

perishable food items like fruits and vegetables.

1.3 Socio-economic Impact of Food Waste in the Context

of Sustainable Development Goal

The Global Hunger Index [3] of 2018 point a staggering 124 million people across the
globe suffer from acute hunger due to malnutrition, internal displacement, refugee
status and poverty. Four of the total 17 goals prescribed by the United Nations SDGs
focus on Poverty, Hunger, Good health and wellbeing, and responsible consumption
and production (Goal 1: No poverty; Goal 2: Zero hunger; Goal 3: Good health and
well-being; Goal 12: Responsible consumption and production). Food loss [6] is the
primary reasons why most of the developing world still suffers from malnutrition and
hunger which is caused by poor storage and processing in the post production phase
and developed nations are highest contributor towards food waste [6] due to fast pace
life, irresponsible and ignorant behavior towards the issue. While the former can be
solved by application of technology and infrastructure management the later can be
solved by improving upon the social physiological behaviors of the masses.

1.4 Introduction to Fourth Industrial Revolution

The term Fourth Industrial Revolution or 4IR [7] was coined by Professor Klaus
Schwab, founder and executive chairman of the World Economic Forum in his book
“The Fourth Industrial Revolution” based on the technologies of the artificial intelli-
gence, machine learning, quantum computing, 3D printing, and the IoT. Around the
year 1760, First Industrial Revolution started in Great Britain, which was powered by
the invention of steam engines. Almost a century later, Second Industrial revolution
started which was characterized by mass production (from craft-oriented production
to mass production) in various industries like steel, oil, and electric. Some of the
greatest inventions like internal combustion engine and light bulb came through in
the same period. The Third Industrial Revolution or “The Digital Revolution” started
somewhere around 1960s powered by the greatest inventions of the century, semi-
conductor chips, which gave rise to personal computing and eventuality made the
Internet a real possibility. Now, 4IR is knocking on the doorsteps of humanity which
will eventually change the way we eat, live, think, or nearly everything we do to
sustain our life.
Mission SDGs through Food Waste Management: Nature and Approaches 253

2 Literature Review

The unprecedented developments in the fields of digital, physical, biological tech-

nology are the three major drivers of the Fourth Industrial Revolution. There is an
attempt to understand the impact of these technologies on various global, industrial,
economic, and social developments [8].
One quarter of the food supplied for human consumption is wasted across the
food supply chain. High-income countries generate food waste at all levels including
household which is the highest. The study gathers data from 1062 Danish respondent
measuring the intention and attitude not to waste food. Food waste can be controlled
by making perceived behavioral control [9].
Sustainable Development Goals (SDGs) are successor of Millennium Develop-
ment Goals (MDGs). It proposes 17 goals with 169 targets with numerous indicators
[10].
Large-scale food waste in the global food supply chain has attracted attention
due to its environmental, social, and economic impacts. There has been an attempt
to understand the difference between food surplus, avoidable and unavoidable food
waste by various specialist, to manage the waste, and to identify the most appropriate
mechanism to create a sustainable supply chain management. There is also an attempt
to understand and distribute the food surplus to poor people or to convert it as an
animal feed [11].
The UN organization, Food and Agriculture Organization estimates a 32 percent
loss of food produce in 2009 based on weight while a 24 percent in terms of calorie.
Food wastage has negative impact on economy, and it represents a wasted invest-
ment, consumer expenses, and farmers’ income loss. While food loss refers to losses
incurred due to spills and spoils, food waste refers to losses due to infrastructure lim-
itations or post-consumer waste which is generally fit for consumption. Food waste
is generally a conscious decision to throw away the food [12].
Wireless sensor network (WSN) surrounds all living beings in the modern era and
influences day to day living. A communication between all these wireless-enabled
network creates the Internet of Things (IoT) to form a seamless environment to create
a common operating picture (COP). The IoT has evolved from the static web2 (social
networking web) to web3 (ubiquitous computing web) increasing the data demand
[13].
Continuous population growth will keep a continuous demand for food supply
for another 40 years approximately, while there will be a decrease in the capacity to
produce food due to overexploitation of land and other natural resources including a
threat from climate change. However, a more efficient ways of food production can
be explored [14].
In order to feed the nine billion human populations by 2050, a review of food
waste in the global supply chain is discussed. Data of post-harvest losses of grains
are outdated, and current global losses are unknown. The impacts of food waste in the
development of BRIC economies are also unknown while developing nations face
food wastage post-harvest due to its perishability developed countries contribute to
254 A. Chakrabarty and U. S. Das

high post-consumer food wastage which suggest a scope for behavioral change to
reduce wastage in affluent population [15].
There is an attempt to understand the link between inflation in food prices and
riots or food riots. Demonstrators of the riot pointed political repression injustice
and inequality which mobilized and bought together various political coalitions to
promote human dignity [16].
Electronic nose has been a trending technology for the last two decades largely due
to numerous applications built around the sensors. Recent changes in the computing
power have given the electronic nose a new possibility of various applications. It has
provided a numerous benefits in the fields of biomedical, agricultural, environmen-
tal, food, cosmetics, manufacturing, military, pharmaceutical, and various scientific
researches. Now, electronic noses can monitor all phases of industrial manufacturing
[17].
An introduction to radio frequency identification systems and their strength, weak-
nesses along with deployment challenges is discussed along with various extensions
that offer read/write memory and environmental sensing along with social issues
[18].
Electronic noses (e-noses) are sensors that can detect various volatile organic
compounds. A wide range of applications can be designed based on pattern recog-
nition with the help of artificial intelligence or neural network. The future trend of
the sensors is also explored [19].

3 Objectives of the Study

1. To explore varied forms of agriculture food loss at different stages up to the phase
of consumption.
2. To provide an IoT-based solution for reducing food waste to a considerable
amount primarily in closed system or household condition.

4 Research Methodology

This paper has attempted to understand, study food waste and its impact in the
global context, provide solution through existing technological framework (IoT and
electronic nose), and achieve related SDGs. This paper has been developed using
secondary information collected from various relevant sources and documents. The
paper has focused on how a technological solution would help achieve sustainable
development goals of the United Nations by 2030.
Mission SDGs through Food Waste Management: Nature and Approaches 255

5 Analysis and Interpretation

Analysis I
The various forms of agricultural food loss can be expressed by using the flow chart
diagram as mentioned below.
Loss of Agricultural Produce at Farming Stage
• Agriculture food loss starts at the very beginning when farmers produce crop
without adequate planning and sharing information among the producer’s com-
munity and other stakeholders, and as a result of that, food loss occurs with
overproduction of crops [20].
• Farmers harvest their crop prematurely for his personal consumption or to earn
money. This accounts both the economic and nutritional loss of food grains [20].
• Massive use of pesticides and fertilizers diminishes the nutritional value of the
food grains, and at times, it could be deterrent to human life.
Loss of Agricultural Produce at Transportation/Supply Chain
• During the movement of crops or food grains from field to warehouse or market,
there are several supply chain/transportations loopholes like lack of refrigera-
tion system, using FIFO technique or VED analysis (Valuable, Essential, and
Desirable) (Fig. 2).

Valuable
(Highly perishable and valuable Fruits & Vegetables)

Essential
(Moderately perishable and valuable Fruits & Vegetables)

Desirable
(Remotely perishable and valuable Fruits & Vegetables)

Fig. 2 Schematic diagram of application of VED analysis

256 A. Chakrabarty and U. S. Das

Market In case of distribution/

supply chain
Z
management, the
sequence of disbursement
Warehouse
of food grains must be
through FIFO (First-in
First-out) i.e. the X
Market
Y consignment must be
disbursed first followed
Warehouse by Y and Z if sequence of
crop loading is in the
order of X, Y, Z: unless
Market otherwise specified or
X special items / highly
perishable goods as the
Warehouse case may be.

Loading of raw food grains for different food produce

Unloading of food grains at different Market / Warehouse

Fig. 3 Schematic diagram: application of FIFO method

Loss of Agricultural Produce at Warehouse/Cold Storage

See Fig. 3.
Loss of Agricultural Produce at Market Place
• In course of selling of food grains in the market, it is difficult to predict actual
demand for each variety of food grains or crops. As a result of that, there is
substantial loss of food grains in the market places which could not be sold at
all or may be sold at a tendered expiry mode.
• In fact, accurate demand forecasting for every set of crops is next to impossible
due to uncertainty of market dynamics and impulse purchasing behavior of the
consumer.
• Lack of adequate refrigeration and cold storage system.
• Lack of Warehouse facilities.
• Due to excessive overproduction of crops, it results in higher storing cost and
stagnation of crops which yield loss of food value, nutrition, and economy
(Fig. 4).
Mission SDGs through Food Waste Management: Nature and Approaches 257

Agriculture Produce

Transportation
Market Food Loss

1. Premature harvesting

2. Loss due to lack of crop planning

and information sharing among
farmers
Warehouse / Cold
Storage 3. Lack of effective storage System

4. Perishability
Transportation

5. Infection due to Micro Organisms

Loss 6. Fermentation

Loss

Market
Transportation

House Hold

Fig. 4 Schematic diagram of food wastes at various stages

Loss of Agricultural Produce at Household

• Various ways of food waste take place at the household level both in uncooked
and cooked format.
• The loss of uncooked produce is primarily because of oversupply of vegetables
or food grains due to lack of ready information about the food reserves in the
household. From lower middle class and above category, the family depends on
household refrigerators for the purpose of preservation of food grains, vegeta-
bles, etc.
• The loss of cooked food occurs due to overestimation of perceived demand
of food for day to day domestic consumption. This loss can be managed by
experiential learning of the household and commitment of the members toward
sustainable use of resources.
258 A. Chakrabarty and U. S. Das

Fig. 5 IOT-based solution for reducing food waste primarily in closed system or household con-
dition

• In fact, the loss of uncooked produce at the household level may be solved with
appropriate smart in-house storage system and economic usage mechanism led
by IoT intervention.
Analysis II
See Fig. 5.
Indicative algorithm for the model
1. The confined storage system (e.g., household refrigerator) needs to be augmented
by multiple IoT devices that would act as E-Nose and E-iris as a means of sensory
devices.
2. The IoT, E-nose and E-Iris would be provide adequate database for image pro-
cessing for with pigment support (e-iris) and aromatic database (e-nose) on each
specific food grains and vegetables generally stored in the system.
3. The sensory device will receive appropriate stimuli both aromatic and pigment
and that would process using IoT, intelligence system (AI) by matching the preset
database and finally the system would generate its predictive analytics about the
nature of food produce stored in the system.
4. The processed results would be transferred at all the connected portals with the
IoT ecosystem on real-time basis so that the user can be aware of the quantity
and quality of each of the variants food products.
Mission SDGs through Food Waste Management: Nature and Approaches 259

5. Based on this input, the user can take most appropriate purchase decision for
further procurement of food grains, i.e., the set of variants to procure or not
to procure and at what quantity. This real-time information would enable the
household to prioritize which vegetables to cook immediately on priority basis
in order to minimize both the nutritional and economic losses.

6 Conclusion

Food waste management has multiple implications on society as well as its economy.
The world is highly apprehensive on the issues of food security, fertility of land, and
nutritional benefits for upbringing off the human civilization. The sustainable devel-
opment goals firmly advocate that poverty, hunger, and responsible production and
consumption are the few integral dimensions which need to be improved if the world
would like to achieve the essence of sustainability. Whatever the policy is formulated
at macro level, finally the success depends on what extent it has been implemented
and practiced at the very micro household level. This paper has emphasized on devel-
oping a sustainable solution to address the food loss issues at the household level
with the intervention of IoT led smart technology.
Limitation of the Study
This paper has designed and developed a strategic solution in order to combat
household syndrome. However, if the model is efficiently implemented, the degree
of minimization of food loss could be explored so that the impact of the model
could be studied, and necessary improvement on the structural morphology could be
incorporated.
Working definition:
e-nose: An electronic nose is used to identify odors by detecting the “fingerprint”
of a chemical compound using pattern recognition software [21].

References

1. Food and Agriculture Organization of United Nation: Data on food loss and waste (2019)
https://stats.unctad.org/Dgff2016/planet/goal12/data/Fig12_1.xls
2. Food and Agriculture Organization of United Nation: Key facts on food loss and waste (2019).
http://www.fao.org/save-food/resources/keyfindings/en/
3. Global Hunger Index (GHI): Global hunger index by severity and country wise ranking (2018).
https://www.globalhungerindex.org/results/
4. Reserve Bank of India: Number and percentage of population below poverty line (Sep, 2015).
https://www.rbi.org.in/scripts/PublicationsView.aspx?id=16603
5. The Hindu: FCI admits 1.94 lakh MT food grain wasted between 2005 and
13 (May, 2016). https://www.thehindu.com/news/national/fci-admits-194-lakh-mt-foodgrain-
wasted-between-200513/article5680994.ece
260 A. Chakrabarty and U. S. Das

6. Food and Agriculture Organization of United Nation: Technical platform on the measurement
and reduction of food loss and waste (2019). http://www.fao.org/platform-food-loss-waste/en/
7. World Economic Forum: The Fourth Industrial Revolution or 4IR (Jan, 2016). https://www.
weforum.org/agenda/2016/01/the-fourth-industrial-revolution-what-it-means-and-how-to-
respond/
8. Li, G., Hou, Y., Wu, A.: Fourth industrial revolution: technological drivers, impacts and coping
methods. Chin. Geogra. Sci. 27(4), 626–637 (2017)
9. Stancu, V., Haugaard, P., Lähteenmäki, L.: Determinants of consumer food waste behaviour:
two routes to food waste. Appetite 96, 7–17 (2016)
10. SDG, U.: Sustainable development goals (2015)
11. Papargyropoulou, E., Lozano, R., Steinberger, J.K., Wright, N., Bin Ujang, Z.: The food waste
hierarchy as a framework for the management of food surplus and food waste. J. Clean. Prod.
76, 106–115 (2014)
12. Lipinski, B., Hanson, C., Lomax, J., Kitinoja, L., Waite, R., Searchinger, T.: Reducing food
loss and waste. World Res. Inst. Working Paper, pp. 1–40 (2013)
13. Gubbi, J., Buyya, R., Marusic, S., Palaniswami, M.: Internet of things (IoT): a vision, archi-
tectural elements, and future directions. Future Gen. Comput. Syst. 29(7), 1645–1660 (2013)
14. Godfray, H.C.J., Beddington, J.R., Crute, I.R., Haddad, L., Lawrence, D., Muir, J.F., Toulmin,
C.: Food security: the challenge of feeding 9 billion people. Science 327(5967), 812–818 (2010)
15. Parfitt, J., Barthel, M., Macnaughton, S.: Food waste within food supply chains: quantification
and potential for change to 2050. Philos. Trans. Royal Soc. B: Biol. Sci. 365(1554), 3065–3081
(2010)
16. Bush, R.: Food riots: poverty, power and protest 1. J. Agrar. Change 10(1), 119–129 (2010)
17. Wilson, A., Baietto, M.: Applications and advances in electronic-nose technologies. Sensors
9(7), 5099–5148 (2009)
18. Want, R.: An introduction to RFID technology. IEEE Perv. Comput. 1, 25–33 (2006)
19. Scott, S.M., James, D., Ali, Z.: Data analysis for electronic nose systems. Microchim. Acta
156(3–4), 183–207 (2006)
20. Stuart, T.: Waste: uncovering the global food scandal. WW Norton & Company (2009)
21. IEEE Spectrum: Electronic nose or e-Nose (Oct, 2018). https://spectrum.ieee.org/the-human-
os/biomedical/devices/meet-the-enose-that-actually-sniffs
22. Tjoa, A.M., Tjoa, S.: The role of ICT to achieve the UN sustainable development goals (SDG).
In: IFIP World Inf. Technol. Forum (pp. 3–13). Springer, Cham (2016, September)
23. Schaller, E., Bosset, J.O., Escher, F.: ‘Electronic noses’ and their application to food. LWT-Food
Sci. Technol. 31(4), 305–316 (1998)
24. Economic Times: India wastes Rs. 244 crore worth of food a day: Report (Jan,
2018). https://economictimes.indiatimes.com/wealth/personal-finance-news/india-wastes-rs-
244-crore-worth-of-food-a-day-report/articleshow/62391489.cms?from=mdr
Frost Prediction in Highland Crops
Management Using IoT-Enabled System
and Multiple Regression

J. M. Mendez and D. D. Dasig

Abstract The magnitude of the effects of climate change in the agriculture-driven

growth, countries, and regions in the world has expurgated the crop production and
a decrease in agricultural productivity and eventual price surge. In 2014, it was
accounted that agriculture is one-third of the global gross domestic product which
was crucial to the country and global economic growth. With the recent problems
in the agricultural sector, agriculturists and engineers have explored the advent of
the cyber-physical system and automation age by using Precision Agriculture to
address these issues and problems particularly in frost prediction in highland crops
management. This chapter will discuss the concepts of Precision Agriculture and
applications of the Internet of things (IoT) in agriculture, the design and challenges
of an IoT-enabled system for highland crops management, and utilized multiple
regression as a frost prediction technique. The remote sensing device deployed in
the farm collects the frost climatic events including air velocity, temperature, and
humidity, and transmitted to the web server. The system provides early warnings
and frost forecast for the farmers using SMS. The system helped farmers as a useful
resource to conduct frost protection activities on the farm, thereby reducing the frost
harm to the crops.

Keywords Frost prediction · Internet of things · Multiple regression ·

Microclimate

J. M. Mendez
College of Computer Studies and Engineering, LORMA Colleges, Carlatan Campus,
San Fernando, La Union 2500, Philippines
e-mail: [email protected]
D. D. Dasig (B)
College of Science and Computer Studies, De La Salle University—Dasmarinas,
DBB-B, 4115 West Ave, Dasmariñas, Cavite, Philippines
e-mail: [email protected]; [email protected]

© Springer Nature Singapore Pte Ltd. 2020 261

P. K. Pattnaik et al. (eds.), Internet of Things and Analytics
for Agriculture, Volume 2, Studies in Big Data 67,
https://doi.org/10.1007/978-981-15-0663-5_13
262 J. M. Mendez and D. D. Dasig

1 Introduction

The agricultural sector has confronted issues and difficulties on how it could bolster
food production and feed the changing and increasing demand of the society. Agri-
cultural engineers, scientists, and technocrats have advocated on the development
of intelligent agricultural production to countervail food production concerns called
“smart” system. Agribusiness has been modified and changing its landscapes in a
few years. More and more farmers must adapt to new technologies to obtain products
with greater value and cost-effective, because of the current disruption and high com-
petition of the market and the high demands on the quality of products. Moreover,
linked to these determining factors, agriculture is suffering the damages caused by
climate change such as floods or constant temperature changes. In the same way as
the industry, agriculture is forced to modernize its work methodologies, to automate
its protocols, to reduce production costs, to increase production by improving the
quality of its products, and to fight against meteorological inclemency [1].
Atmosphere fluctuation and change can unfavorably influence human and natural
ecosystem as well as the ecological systems. Understanding the adaptation and miti-
gation capacities of agricultural producers is essential for improving climate-related
support to the sector. Farmers need be able to engage in a farm management and
assortments of approaches and maneuvers to accomplish and countervail climate-
related risks, as well as uncertainties from other sources [2]. Understanding how
climate affects your growing conditions is essential to successful gardening. Climate
is the extent of middling disparity in general climate configurations over time in a
very eclectic topographical region or zone.
In a farm field, climate strongly influences growing conditions: air and soil tem-
peratures, precipitation, and heat accumulated over the growing season. When air
temperatures fall below 0 °C, sensitive crops can be injured, with significant effects on
production. For example, in Benguet, there are more economic losses to frost damage
than to any other weather-related phenomenon. Therefore, impacts on affected farm-
ers and the local economy are often devastating. Although it is important, information
on how to protect crops from freezing is relatively limited. Consequently, there is a
need for a widely available, simplified source of information to help farmers address
this serious problem.
Technically, “frost” alludes to the arrangement of frost precious stones on surfaces,
either by solidifying of dew or a stage change from vapor to frost; notwithstanding,
the word is broadly utilized by people, in general, to depict a meteorological occasion
when yields and different plants experience freezing damage. Producers frequently
utilize the expressions “frost” and “freeze” reciprocally, with the vague definition
being “an air temperature not exactly or equivalent to 0 °C”. Instances of frost
definitions in the published works and literatures include: the event of a temperature
not exactly or equivalent to 0 °C estimated in a “Stevenson-screen” cover at a stature
somewhere in the range of 1.25 and 2.0 m; the event of an air temperature under 0 °C,
without characterizing the safe house type and tallness; when the surface temperature
Frost Prediction in Highland Crops Management … 263

dips under 0 °C; and the presence of a low air temperature that makes harm or passing
the plants, without reference to frost arrangement and formations.
In the investigation of [3], frost was defined and falling into two categories “ad-
vective” and “radiative.” Advective frosts are related to enormous scale attacks of
cold air with a well-blended, blustery atmosphere and a temperature that is regularly
below zero, notwithstanding during the daytime. Radiative frosts are related to cool-
ing because of energy loss through the radiant trade during clear, quiet evenings,
and with temperature reversals (e.g., temperature increments with heights). Now and
again, a mix of both advective and radiative conditions will happen. This might be
trailed by a few days of clear, quiet conditions that are helpful for radiation frosts.
Moreover, the researchers have posited conditions that are considered as “micro-scale
shift in weather conditions frost” or micro-scale advection frosts. These happen when
the area is presented to radiation-type frost conditions; however, nearby chilly air
waste prompts fast drops in temperature from a more minor perspective inside the
radiation frost territory.
Similarly, freezing and frosts bibliotheque definitions in lexicons and the writing
are variable and befuddling; be that as it may, on an overall premise, the term frost
protection is more ordinarily utilized than freeze protection. A “frost” is attributed to
as an event of an air temperature of 0 °C or lower, estimated at a stature of somewhere
in the range of 1.25–2.0 m above soil level, inside a suitable climate shelter. The water
inside plants could conceivably solidify during a frost episode, contingent upon a few
shirking factors (e.g., supercooling and convergence of frost-nucleating microscopic
organisms).
A “freezing” happens when extracellular water inside the plant freezes (e.g.,
changes from a fluid to frosts). This could conceivably prompt harm of the plant
tissue, contingent upon resistance factors (e.g., solute substance of the cells). A
frost occasion turns into a freeze occasion when extracellular frost frames within
the plants. Freeze damage happens when the plant tissue temperature falls beneath
a basic worth where there is an irreversible physiological condition that is helpful
for death or breakdown of the plant cells. This harming plant tissue temperature is
connected with air temperatures called “basic temperatures” estimated in standard
instrument covers. Below-zero air temperatures are brought about by decreases in
reasonable warmth substance of the air close to the surface, basically coming about
because of (1) a net vitality deficit through radiation from the surface to the sky (e.g.,
radiation frosts); (2) wind blowing in below-zero air to supplant hotter air (e.g., shift
in weather conditions frosts); or (3) a mix of the two procedures.
A “shift in weather conditions frosts” or “freeze” happens when the air temperature
falls beneath 32 °F (0 °C) during the night, yet the breeze speed is sufficiently high
(normally more prominent than 5 mph) with the goal that it blends the air and there
is practically no reversal arrangement. In a “solidify,” the temperature often remains
beneath 32 °F (0 °C) notwithstanding during the day, and it can proceed for a few
days straight. Then again, “radiation frosts” or essentially a “frosts” is regular in the
Province of Benguet, and these happen when there are quiet breezes and clear skies,
a reversal structures, and the temperature fall underneath 32 °F (0 °C). As a rule, the
264 J. M. Mendez and D. D. Dasig

temperature will transcend 32 °F (0 °C) during the day following a frost occasion;
be that as it may, frost occasions can happen a few evenings in succession.
Both freeze and frost occasions are related to low stickiness. Now and again,
the atmosphere conditions begin as a “solidify” on the principal night and change
to a “frosts” on ensuing evenings. Most insurance strategies are compelling against
“frosts” occasions yet not against “solidify” occasions. Since the harm is brought
about by frosts development inside the plant tissue bringing about cell drying out and
cell divider harm, the terms frosts and freezes harm are exchangeable. Frost occasions
can be partitioned into two classes: radiation frosts and advective frosts. The previous
will, in general, happen at the meteorological attributes of clear skies, no wind, and
low dew-point temperature. The last typically happens under the meteorological
states of overcast skies, moderate-to-solid breezes, no temperature reversal, and low
moist. Frost occasions are progressively basic in the Province of Benguet, so the
analyst utilized the term frosts harm when depicting the damaging impact of frosts
arrangement on vegetables.
Frosts hazard is a significant unavoidable truth for agribusiness. Frigid tempera-
tures limit the length of the developing season and are responsible for decreases in
yield and nature of harvests. Delicate harvests can endure genuine harm at tempera-
tures above solidifying also. In either case, the base temperature is basic. Of all the
climate elements influencing farming, the base temperature is among the most factors
here and there. Contiguous fields and nursery destinations can have altogether dif-
ferent frosts-free periods and are along these lines fit various employments. Hazard
can be diminished via cautious administration of the land and successful response
to figures. Pre-summer and late-summer frosts more often than not include least
temperatures only marginally beneath solidifying or freezing; it is in these circum-
stances that preventive measures can be best. By effectively turning away a solitary
solidifying circumstance, the developing season might be reached out by weeks.
The area of Benguet in the self-ruling locale of Cordillera speaks to the most part
rain-fed upland agroecological zone dedicated to high-esteem crop generation and
exchange. Benguet’s rocky landscape and immense reliance on occasional precipi-
tation make the rural partners in the territory especially touchy to regular climatic
oddities. The region would suitably fill in as study site when investigating the esti-
mation of atmosphere-related data especially for delicate on-farmers and off-farmers
decisions [4].
Situated in the northern part of the Philippines, alongside the Cordillera’s moun-
tain ranges, and a huge and largely agribusiness-based Benguet area, poverty is
prevalent in the region. Benguet is no more unusual to penury. In 2000, the National
Statistics Coordination Board [5] announced that somewhat over 30% of the com-
monplace populace lives in poverty. Land corruption and degradation, peripheral
and marginal profitability and inadequate or lack of foundation and infrastructure
are a portion of the issues being looked by the province. These serious interrelated
issues are basic in numerous highland and upland areas and territories as a result
of unfavorable natural conditions and an absence of resources that these regions are
named “Less Favored.” It is comparable given these reasons that Benguet can be
Frost Prediction in Highland Crops Management … 265

named a less favored zone. A few man-made and natural resources impede regions
and locales economic development.
From January 30 to February 2, 2019, Atok, Benguet, hit a 9 °C temperature
that made a cover of frost secured a portion of this present town’s homesteads,
yet neighborhood authorities eased apprehensions, and that this would affect the
supply and costs of vegetables in the Province of Benguet [6]. At the point when
Atok encountered a temperature of 7°, farmers mixed to the patches of farmland
still grasped by frost. Sitios (sub-towns) Lower and Upper Engladad in Barangay
(village), Paoay in Atok endured pockets of frost, which additionally influenced the
Kibungan towns of Nilicob, Proper Madaymen, Taliboy-oc, Cagam-is, and Masala,
and the town of Cada in Mankayan town, as indicated by agriculture information
discharged by the Philippine Information Agency’s office in the Cordillera.
A portion of the farmers depended on indigenous learning with regard to frost
forecasting. It was accepted that an unmistakable sky with no wind development
toward the evening is an indication that a frost scene will happen around the evening
or early morning the next day. At the point when frost occurs, farmers assume that
there is nothing that should be possible about it.
This study was conducted to develop and deploy a frost prediction in highland
crops management using IoT-enabled system and multiple regression. The system
will also monitor the farm field temperature, humidity, and dew as the most important
climatic parameters. The system was deployed and tested on the select municipalities
in the Province of Benguet, Philippines as illustrated in Fig. 1.

2 Background of the Study

Agriculture has been the conventional and still the real wellspring of employment
for the individuals in the province. At least 54% of the labor force is engaged in
vegetable/cut flower farming. Vegetable products from the province supply 80% of
vegetable needs of the metropolis. Vegetables grown are cabbage, Chinese cabbage,
broccoli, cauliflower, potatoes, carrots, and other temperate vegetables. Cut flower
production is also gaining attention because of its promising profitability. These
are roses, chrysanthemums, lilies, carnation, Shasta daisies, and others. The “Rose
Garden of the Philippines” is situated in the area especially Barangay Bahong, La
Trinidad. Tough and slanting territory and valleys describe the province.
The province is a leading producer of vegetables; it also has potential for grow-
ing oranges, pears, and other temperate fruits. Also, just recently, Benguet has been
dubbed as the term “Strawberry Country.” In Benguet, vegetable farmers brought
their produce either directly or through the assembler wholesalers, commission
agents, or wholesalers to the La Trinidad Trading Post and Hangar Market. The
traders at the trading post initially got a large proportion of the farmers’ produce.
Subsequent trading activities, however, occurred where vegetables sold at the trading
post went to the Hangar market. Some of these vegetables were sold in the domestic
market through local wholesaler-retailers and retailers while the major proportion
266 J. M. Mendez and D. D. Dasig

Fig. 1 Geographical map of the Province of Benguet, Philippines

of about 73 and 17% were brought out of the region through the large assembler
wholesalers and wholesalers [7].
Data from the office of the Provincial Agriculture in Benguet revealed that in
terms of average production in metric tons from 2010 to 2016, potatoes (18,666 mt)
are the top commodity produced followed by cabbage (14,790 mt) and chayote at
9472 mt. Carrot is the top four with 7219 mt, and completing the top five is Chinese
cabbage with 4222 mt’ production from 2002 to 2015. Even looking at the five-year
average production from 2011 to 2015, we can still infer that the same vegetables
are in the top five in Benguet. There is about 19.5 thousand metric tons for potato,
chayote (17.5 mt), cabbage (16.9 mt), carrots (8.8 mt), and completing the top five
is Chinese cabbage at 4447 mt. In terms of average production from 2011 to 2015,
Broccoli is included in the top 10 produced vegetables in Benguet (Fig. 2).
Frost was considered as a regular and accepted phenomenon in certain municipal-
ities of the province, until recently when its occurrence and perceived impact gained
more interest from social media, traditional news outlets, and the general public. As
of now, there is no available frost forecast or advisory in the province. In the study [1],
it characterized the occurrence of frost in some areas in Benguet and explained that
Frost Prediction in Highland Crops Management … 267

Fig. 2 Crops produced by the Province of Benguet, Philippines

the areas with high risk have a high frequency of the cloudless nights with temper-
atures falling below within a minimum temperature threshold. This data applicable
as cloud cover is one of the central points for radiative frost occurrences. During the
passage of a cold surge, highly elevated mountainous areas with clear skies like Atok
and Kibungan are more prone to frost. Apart from the meteorological conditions, the
geology assumes an imperative job in the frost arrangement and frost formations.
Elevation and morning potential insolation (MPI) have the most influence on frost
risk. The seasonal variability of the solar radiation in the morning contributes to the
soil energy balance, local diurnal wind development, and risk of radiative frost [1].
Plants are not damaged by freezing temperature but by frost formation inside
plant tissues. The frost crystals, which are huge concerning plant cells, structure in
the space between cells. The frost crystals develop by coaxing the water out of the
cells prompting parchedness and dehydration of the cell [8, 9]. Along these lines,
when the frost liquefies, the cell divider is harmed. In this manner, anything that
decreases the opportunity of frost development inside the plant tissue maintains a
strategic distance from freeze injury [10, 11].
In addition, when a basic change in one of the estimations happens, at that point
the farmers and the agriculture specialists were recognized employing SMS. Extreme
climate conditions have been recognized as one of the main reasons for the decline
of agrarian efficiency and agricultural productivity in numerous nations around the
world. In the agriculture industry, advanced decision support through Internet of
things (IoT) technologies are gaining attentions as it enables precision farming. Smart
agriculture based on microclimate data analysis and advanced decision support sys-
tems helps improve productivity, avoid unnecessary costs related to harvesting, and
use proper pesticide and fertilizers to enable precision farming [12, 13].
268 J. M. Mendez and D. D. Dasig

Temperature is one of the essential elements affecting plant development and

geological appropriation toward compositional quality [14]. Similarly, frosts will
have an adverse effects on crop production and even environmentally, and econom-
ically. The high humidity, frost as well as hail have adverse effects to crops yield
and quality [15]. The minimum temperature which damage is caused varies with
plant species and with growth stage for a given plant. Precision agriculture (PA) uti-
lizes the information and communication technologies and engineering application
in order to manage crop production and the lifecycle of crop management. Part of the
Internet of things (IoT) technology the ambiental parameters needed (e.g., temper-
ature, moistness, dew) can be estimated utilizing sensors. Wireless sensor networks
(WSN) consist of random or planned spatially distributed nodes, also called nodes,
each of which is equipped with sensors, memory for storage, a microprocessor for
computing their decisions, batteries for energy supply and can communicate wire-
lessly with other nodes in a short-range. Today, it is possible to build multifunctional
sensor device, which are smaller, cheaper, that can facilitate better communication,
and operationally, a low-power device, in contrast with the traditional data loggers.
Notwithstanding the noteworthy advances in farming innovation in the course
of the most recent couple of decades, agricultural sector and production remain
dependent on weather and climate. It is a clear reality that climatic variability will
play an even greater role than in the past, as sufficient food supplies will not be
available to feed the world population adequately at its present rate of increase,
decelerating advances in technologies, natural resources and security with up-to-
date information on crop conditions [16–18]. These variations can be addressed and
treated using data-driven decisions and programs.
Apart from the different techniques developed to fight against frost, there still
exists an unresolved issue: to foresee when the frost will happen and on the off chance
that it is relied upon to happen, to determine the exact period that is more effective to
run the anti-frost technique. This is a problematic concern if the administration will
deliberate about that in this similar portion of terrestrial land, the temperatures may
be different. Climatic changes observed and felt by the community, generally, agree
with observed changes by PAGASA noted as still “normal” although the temperature
and rainfall pattern is pronounced as some of the manifestations of climate change.
These communities are also replete with mechanisms both at the farm and house-
hold levels to negotiate changes in the climate. Such mechanisms, however, are not
necessarily sustainable.
According to [19, 20], the majority of the farmer in Benguet cultivates medium-
scale (48.2%) and small-scale farms (42.1%). Small-scale farms predominate in
Kibungan (58.8%) and Kapangan (74.3%), while medium-scale farms are in Atok
(71.9%) and Bugias (56.3%). These results validate NSCB 2007 data where a com-
bined area of 6743 ha from the total 29,983 ha of Benguet farms is less than a hectare
[5]. Of these farms, nearly three-fourths (70.1%) are owned by the farmers, while
28.5% are either renting (17.8%) or caretaking (10.7%). Most (54.3%) of the farmers
inherited their landholdings.
Frost Prediction in Highland Crops Management … 269

Before the modern way of weather forecasting, the indigenous people (IPs) of
Benguet have their system of forecasting weather mainly based on the natural phe-
nomenon (biotic and abiotic) on their environment. Under this system are periodical
changes on the behavior of local animals most especially birds, drop and rise of
temperature, rainfall, and other weather elements. Birds are scientifically proven
to have demonstrated to be able to foresee or predict that there will be a shift in
weather/climate through detecting changes in the barometric (air) pressure [21, 22].
For example, nearby people in the region or province, they are guided by the behav-
ior of local birds. In that capacity, they named every month with the neighborhood
vernaculars as agricultural calendar.
In the past, the IPs of Benguet plans their agriculture activities on planting and
harvesting which help strengthened their community and household food security.
Usually, land preparation and clearing of the garden for the first cropping are con-
ducted from January to February. The yield would be planted in March and April and
would be harvested in June or July. In the past, March and April are rainy months;
thus, farmers of Loo and Paoay could plant in these months. But with climate change,
gardens with no irrigation/water source are not farmed except for farmers with water
pump. The usual first cropping crops would be carrots and cabbage. In Bayabas and
Taloy Sur, the month of March would be the start of “pinagkakaingin” (swidden
farming).
The stormy season as a rule starts in May. Be that as it may, it has been observed
that the standard cycle for the dry and rainy season has been disturbed, thereby
changing of planting and harvesting schedules. Farmers now would take the risk of
planting knowing that the timing of rainy and dry seasons has been distorted. The next
cropping will start from October to December, and the month of August is known
to have the strongest typhoons; therefore, gardens are fallowed in this month; Lan
preparations start in September and planting in frost October or November. In Atok,
cropping starts before the expected time of frost formation. Most frost occasions
happen during clear and quiet evenings.

3 Methodology

In this section, the methods, tools, instrument, and data gathering techniques are
presented. The researchers employed the Design Science Research Methodology
(DSR). This methodology guided the researchers to propose, develop, and deploy
an IoT-based technology for highland crop management as a new development in
precision farming to improve the farmers’ livelihood techniques and their social
aspect through modifications on the processes and technologies [23, 24].
270 J. M. Mendez and D. D. Dasig

Table 1 Distribution of respondents

Respondents Total population Sample size
Farmers 853 156
Land owners 226 41
Local government officials (PAGASA, DA) 20 4
Total 1099 201

3.1 Research Locale and Sampling Technique

The study was conducted for the Province of Benguet of Cordillera Administrative
Region (CAR) in the Philippines. Benguet is one of the highland provinces in the
Philippines, and also known as the Salad Bowl of the Philippines being the huge
producers of upland vegetables [25–27]. During the pilot testing, tests were conducted
in the crops and vegetable farms in Atok, Benguet, 16° 35 N 120° 42 E with 7400 ft
elevation and most frequently affected by frost and climactic variations. Atok has a
mountainous land area, with many steep cliffs. Accordingly, the Philippine Highway
System’s highest point is Atok’s Barangay Paoay at 2255 m (7400 ft) above sea level
[25, 26].
A systematic sampling was utilized as a more efficient sampling technique to
simplify the process of selecting a sample or to ensure ideal dispersion of sample
units throughout the population [28–30]. Lynch formula was used to determine the
research respondents or participants;

SS = Z 2 ∗ ( p) ∗ (1 − p)
(1.1)
C2
where Z = Z value, p = percentage picking a choice, expressed as decimal and C =
confidence interval.
The population of the study is one thousand ninety-nine (1099) and was shown
as illustrated in Table 1 according to their group and common characteristics as
farmers—853, landowners—226, and local government officials (PAGASA, DA)—
20.

3.2 Project Development Methodology

The researchers utilized a Modified Iterative and Incremental Project development

methodology as illustrated in Fig. 3. This methodology aimed at mitigating pitfalls in
the projects due to unrealistic goals and planning [31–33]. The requirements include
the functional and non-functional requirements. The whole project was broken into
Frost Prediction in Highland Crops Management … 271

Fig. 3 Modified iterative and incremental methodology

smaller chunks with respective milestone and project stakeholders, project cost-
ing, and timeline and quality metrics. The requirements were analyzed and catego-
rized into user requirements, process requirements, software requirements, hardware
requirements, and technical requirements [32, 34].
During the analysis phase, two areas were treated, the data analysis and require-
ments analysis. During the system design phase, the block diagrams, schematic
diagrams, database design, web design, mobile app design, and among others were
designed and synchronized with other requirements. During the system develop-
ment phase, the researchers embarked into codifying the designs, bread boarding,
and implementing the schematic diagrams. Each design is presented to the users to
encourage user engagement in the SD phases, and to accommodate changes to the
requirements [31, 32, 34].
Testing of system subcomponents, components, functional requirements, and
other specification requirements was also conducted to identify bugs, glitches, and
other system issues to be remediated early on. Anti-ANDAP was subjected to respon-
dents and system users’ evaluation using the software quality metrics ISO 9126. After
a series of successful tests, and user acceptance test with the project sponsor, anti-
ANDAP was deployed in Atok, Benguet.

4 Results and Discussion

This section will detail the existing farming practices to protect vegetables from
frost, parameters considered in determining frost occurrence. It will also discuss the
272 J. M. Mendez and D. D. Dasig

architecture and design of the IoT-based Highland Crop Management, and the frost
prediction technique using multiple regression.

4.1 Existing Farming Practices to Protect Vegetables

from Frost

Based on the data gathered from the active observation, informal and semi-structured
interview, and secondary data analysis, the researchers concurred with the following
results. Due to its high elevation, the Province of Benguet has a mild climate with
about 10 °C compared to the temperature at Baguio City Proper. From January 30
to February 2, 2019, Atok, Benguet, hit a 7 °C temperature that made a blanket of
frost covered some of this town’s farms, but local officials dispelled qualms that
this frost would have an adverse effects and impact on the resource and price of
vegetables in the Province of Benguet. When Atok experienced a temperature of 7°,
farmers scrambled to the patches of farmland still embraced by frost. Sitios Lower
and Upper Engladad in Barangay Paoay in Atok suffered pockets of frost, according
to agriculture data released by the Philippine Information Agency’s off frost in the
Cordillera.
According to the farmers, frost was considered as a regular and accepted phe-
nomenon in Atok, until recently when its occurrence and perceived impact gained
more interest from social media, traditional news outlets, and the general public. As
of now, there is no available frost forecasting or advisory in the province as stated
by Edward Haight owner of the Sakura Park in Barangay Atok; also there are a few
sorts of research that were directed in their homestead; however, a portion of the
exploration was not focused on the open field.
Frost is most likely to happen during calm, clear nights, especially if the humidity
is low. Under these conditions, there are no clouds to reflect heat to the earth, and no
wind to mix the descending cold air with ascending warm air. Temperature changes
all the more immediately when mugginess is low, which are the reason temperatures
drop so quickly after sunset in desert areas. Most of the time if the farmers feel that
frost will occur, they usually get “dayami” or straw grass from the lowland areas and
they fog their farm using the straw grass, but the risk of it was the transportation of
straw grass to Atok and frost usually occur at dawn; most of the time was 2:00 a.m.
to 6:00 a.m. The farmers cannot go to their homestead during those seasons of the
day. This data is important as overcast spread is one of the central points for radiative
frost occurrences. During the passage of a cold surge, highly elevated mountainous
areas with clear skies like Atok and Kibungan are more prone to frost according to
Mr. Haight.
Also, to understand the cold weather, the farmers considered and observed if the
sun warms the world’s surface during the day and the warmth is emanated again
into the climate during the night. The chilliest temperatures happen about sunrise.
Mists around evening time assimilate and reflect warmth to the earth. Quiet, starry
Frost Prediction in Highland Crops Management … 273

evenings represent the most serious peril of frost since there is no wind to blend the
rising warm air with the diving cold air and no mists to transmit warmth back to the
dirt.
Humidity eases back temperature changes noticeable all around. Cold air settles
descending, streaming like water, to the absolute bottom. Tourist rises. Cold breeze
mixes temperature misfortune.
The dew point is the temperature at which the air discharges moisture. It is an
element of temperature and relative stickiness. At the point when the dew point
is above frosty, it leaves the proof as haze noticeable all around or dew on the
ground. During the evening, vegetable field (from radiation misfortune) to the dew
point, product buildup, or dew and after that may keep on cooling to the point of
solidification or lower. The dew at that point solidifies into frost.
Farmers of high-value crops that are vulnerable to frost consistently check climate
figures for the daily dew point. The dew point is considered the “storm cellar” or
low temperature for the night because the buildup of humidity or arrangement and
formation of frosts eases back or freezes the pace of temperature drop. On the off
chance that the dew point is above frigid, most plants are probably not going to be
harmed by frost. In any case, subsolidifying dew-point flags the need to think about
defensive measures. As illustrated in Fig. 4, the low humidity and humid air and the
moist soil can create frost to their vegetable fields.
Accordingly, low temperature (e.g., chilling and solidifying) damage can hap-
pen in all plants; however, the components and kinds of harm differ impressively.
Numerous organic products, vegetable, and decorative yields of tropical starting point
experience physiological harm when exposed to temperatures underneath about +
12.5 °C, henceforth well above frigid gum-based paint. Nonetheless, harm over 0 °C
is chilling damage as opposed to solidify damage. Freeze damage happens in all
plants because of frost development. Harvest plants that create in tropical atmo-
spheres, frequently experience serious frost harm when presented to a temperature
marginally underneath zero like Atok, Benguet. A few vegetables, which began in
a temperate atmosphere, yet can be harmed at temperatures close to 5 °C and some
subtropical organic products trees that can withstand temperatures to 2–3 °C. Species

Fig. 4 Occurrence of frost

274 J. M. Mendez and D. D. Dasig

or assortments show diverse frost harm at the equivalent temperature and phenolog-
ical arrange, contingent upon precursor climate conditions, and their adjustment to
cold temperatures before a frost night is designated “solidifying.”
In the meeting with CHARM farmers (Cordillera Highland Agricultural Resource
Management), an association of farmer and landowners of Benguet, it was exempli-
fied that during cold periods, plants will, in general, freeze against freeze damage,
and they lose the freezing state after a warm spell. Freezing is most presumably
identified with an in-wrinkle in solute substance of the plant tissue or diminishes in
ice-nucleation dynamic (INA) microorganism fixations during cold periods or a mix.
Frost harm to harvests has been an issue for people since the primary yields were
developed. Regardless of whether all parts of yield creation are very much overseen,
one night of frosty temperatures can prompt complete crop misfortune.
In Atok, Benguet, harming occasions are ordinarily advective, frost occasions
with quiet conditions and more grounded reversals are progressively run of the mill.
The harming occasions ordinarily begin with shift in weather conditions of virus air
pursued by a couple of evenings of radiation frost. In calm atmospheres, frost periods
are shorter in term and happen more every now and again than in different towns of
Benguet.
To an enormous degree, the potential for frost harm relies upon neighborhood
conditions. Subsequently, it is hard to show a topographical assessment of potential
harm.
The normal length of the frost-free period, which keeps going from the event of
the last below zero temperature in the spring to the first in the pre-winter, is here and
used to topographically portray the potential for harm.
Shielding plants from frost harm more often than not is practices by catching the
warmth that is emanating starting from the earliest stage clear plastic, by moving the
plant to a protected area, or by straightforwardly giving warmth by means of lights.
Each measure gives around 2–3 degrees of protection. Keep plants well-watered to
upset the drying up of frosts (Fig. 5).
Similarly, during frost at night, if the temperature is low enough, extracellular
water freezes, drawing water out of the plant cells. If cell desiccation is limited, the

Fig. 5 Traditional way of protecting the crops from frost

Frost Prediction in Highland Crops Management … 275

water will move back into the cells as temperatures rise the next morning without
causing damage. During this process, the leaves will take on a black water-soaked
appearance, but they regain a normal appearance as warming takes place. If desic-
cation is severe, cell wall damage causes cell death. The lower the temperature, the
longer the exposure, and the faster the temperature drops, the greater the damage to
the plant.
If an unexpected light frost catches unprepared, the farmers simply drape a plastic
tarp, blanket, or newspapers over plants before nightfall to trap the heat. It is better
to drape the cover over a frame or wire cages so that it does not touch the plant
leaves. Leaves touching the covers are likely to be frost damaged. People tend to
leave coverings on for weeks, but the coverings should be removed during the day to
allow heat in and then recovered at night. Other farmers turn on sprinklers to decrease
the temperature by drizzling the farm field.
Based on the interview responses, and active observations, the farmers relied on
indigenous knowledge in frost forecasting. It was believed that a clear sky with no
wind movement in the afternoon is a sign that a frost episode will occur at night or
early morning. When frost happens, the farmers simply drape a plastic tarp, blanket,
or newspapers over plants before nightfall to trap the heat. Some of the farmers also
use light bulbs to heat the plants, and others turn on the sprinkler as manual frost
protection.

4.2 Parameters Considered in Determining Frost Occurrence

Internet of things (IoT) is a set of device and software elements, sensors that allow
association, accumulation, representation and information sharing, which through
their properties improve vitality productivity and reduce human effort while also
bringing economic benefits [35–37]. Nowadays, different organizations and govern-
ment departments are taking interest in implementing the technology for agriculture
parameters measurement. Manual collection of parameters for desired factors can be
erratic, not continuous, and produce variations from measurements results in diffi-
culty in controlling agriculture environment. IoT and remote sensor hub can decrease
the time and endeavors required for checking the horticulture condition [35, 36, 38].
The IoT likewise maintains a strategic distance from the loss of horticulture param-
eter database and spares it into a capacity gadget or cloud for long life [36, 37, 39].
Also, it will allow monitoring in critical place without the need to put personnel
in hazardous situations. A monitoring system should have quicker response time,
better quality control of the crop, and lower labor cost [38–40]. In this study, several
parameters such as soil moisture, precipitation, dew, the temperature are considered
as variables needed in detecting frost occurrence; this parameter can monitor on a
real-time basis using IoT [35, 36, 41].
Sensors are set at various areas to gather the information to anticipate and predict
the behavior of a particular area in a Barangay in Atok, Benguet. The nodes are
monitored remotely using the Internet, and the data gathered from the sensors are
276 J. M. Mendez and D. D. Dasig

put in the cloud and to extend the evaluated pattern on the web browser. The data
collected concerning the normal or specified ranges of particular parameters. The
embedded system is an integration of sensor device, wireless communication which
enables the user to remotely access the various parameters and store the data in the
cloud [35, 37, 39, 41].

4.3 The IoT-Based Highland Crop Management

4.3.1 IoT Architecture the System Block Diagram

The IoT-based Highland Crop Management utilized Internet-based technical archi-

tecture “IoT architecture” as illustrated in Fig. 6. In this architecture, the anti-ANDAP
system was developed and deployed using the three layers including the IoT device
layer, IoT gateway layer, and the IoT platform layer.
The researchers concur that the IoT architecture was able to facilitate the wireless
sensor network (WSN) gateway so that ANDAP can perform its functions such as
measuring, processing, and analyzing environmental indicators related to frost [42].
The IoT-based Highland Crop Management System is illustrated in the block dia-
gram in Fig. 7. The sensor unit includes the temperature, humidity, dew, precipitation,
and soil sensors. In the IoT architecture, the sensor unit is the IoT device layer. It is
also called the sensor node in a wireless sensor network context. The sensor node is
connected to the Arduino Mega 328P microcontroller and power by a solar panel.
The GSM/GPS module with Wi-Fi shield is the IoT gateway layer. This layer serves
as the bridge between the IoT device layer and IoT platform layer. It uses the Wi-Fi
Access Protocols IEEE 802.11 standard, while the IoT platform layer includes the
Online Crop Management System and the mobile application for anti-ANDAP, web,
and other cloud computing services.
In this system, three are three (3) sensor nodes deployed in the locale (Atok,
Benguet). The sensor nodes as connected together as a wireless sensor network.
The WSN is connected to a gateway node (GSM/GPS and Wi-Fi shields), which
allows the IoT device to send sensory data from the environment and other ambiental
information need for frost forecasting to the platform layer, such as the Online Crop
Management System illustrated in Fig. 8.

Fig. 6 IoT architecture

Frost Prediction in Highland Crops Management … 277

Fig. 7 IoT-based highland crop management block diagram

Fig. 8 Graphical user interface of online highland crop management system

4.3.2 The System—Online Crop Management System

Using the online application and mobile application, farmers can check to monitor
the frost prediction parameters themselves even at home.
Due to scarce in electricity supply in the locality, farmers need to monitor their
distant sprinklers during irrigation at night. The system interface has a feature-rich
pump starter and sensors status monitoring. Farmers can also switch ON and OFF
278 J. M. Mendez and D. D. Dasig

Fig. 9 Graphical user interface of environmental and ambiental condition monitoring for frost
prediction

the sprinklers when deemed necessary. Figures 8 and 9 show the graphical user
interface of Online Highland Crop Management System. The GUI also shows the
data analytics that are generated based on the climatic variations and sensor nodes
data sent to the web-based and mobile applications.

4.4 Frost Prediction Using Multiple Regression

Regression constitutes a supervised learning model, which aims to provide the pre-
diction of output varies according to the input variables, which are known. Most
known algorithms include linear regression and logistic regression, as well as step-
wise regression. In determining the reliability of the frost prediction using the Internet
of things, the researcher used multiple linear regression (MLR) as predictive analyt-
ics on forecasting frost occurrence in Atok, Benguet, because this technique is much
simpler than other techniques and can be more useful for operational activities to
prevent frost damages to agricultural crops.
Regression attempts to determine the strength of the relationship between one
dependent variable usually denoted by Y and a series of other changing variables
known as independent variables. In multiple linear regression (MLR), there are more
than two variables among which one is the dependent variable and all others are the
independent variables, and the equation looks like this:

Yi = β0 + β1 xi1 + β2 xi2 + β3 xi3 · · · β p xi p (2)

Frost Prediction in Highland Crops Management … 279

where Y i is the i-th value of the dependent variable, β 0 is the intercept, β 1 is the i-th
angular coefficient, and x i is the i-th vector of observations (predictors).
For this study, there are six (6) predictors that believed have significant impact
on frost detection and are used in applying multiple regression such as average
temperature (°C), maximum temperature (°C), minimum temperature (°C), average
humidity (%) (measured by 8:30 am and afternoon humidity measured by 5:30 p.m.),
dew (mm), precipitation (%). For cross-validation, local temperature data taken from
Atok, Benguet, was taken as dataset. The dataset from PAGASA for 3 years was used
for the experiment during the month and year from January 2017 to May 2019 which
has a total number (k) of 29 observations taken. To measure the quality of the MLR
equation, the predicted temperature amount is compared with the actual temperature
gathered by the nodes on the field.
Before the MLR was undertaken, the correlation of the predictors using correlation
entity was conducted with the results illustrated in Table 2. The results imply that
minimum temperature was highly correlated with average temperature and maximum
temperature with minimum temperature. Thus, precipitation is not correlated with
maximum temperature, and humidity is not correlated with precipitation because of
the negative values shown on their correlation.
Table 3 describes the summary output of the regression statistics taken from

Table 2 Correlation of predictors

Predictor Ave Min Max Dew Precip. Hum. (%)
(mm)
Ave C 1
Min C 0.993917954 1
Max C 0.876113185 0.864295666 1
Dew (mm) 0.533599548 0.535485727 0.417316696 1
Precip. −0.65237895 −0.65832911 −0.47075748 −0.7103738 1
(mm)
Hum. (%) 0.4437614 0.468292939 0.357942558 0.06579885 −0.7723735 1
ave C average temperature in °C; min C minimum temperature in °C; max C maximum temperature in °C;
dew dew in mm; precip. (mm) precipitation in mm; hum. (%) humidity (%)

Table 3 Regression statistics

Summary output
Regression statistics
Multiple R 0.751513272
R square 0.564772198
Adjusted R square 0.446073707
Standard error 6.337164685
Observations 29
280 J. M. Mendez and D. D. Dasig

Table 4 Multiple regression table

Predictors Coefficient correlation Standard error p-value
Intercept −67.3909377 65.55815905 0.015
Average temperature (°C) 2.97551657 5.29959839 0.080
Minimum temperature (°C) −4.725115525 4.948244714 0.050
Maximum temperature (°C) 1.67216714 1.243787101 0.092
Dew (mm) 4.735493886 0.989893894 0.061
Precipitation (%) 7.163066034 3.479092271 0.051
Humidity (%) −0.203586847 0.35390595 0.070

the dataset which has R square of 0.56; hence, it shows that the regression square
has a 56% low variability of statistic and 0.75 multiple R signifies a 75% good
multiple regression. The adjusted regression square was computed with 0.446 output
to compare between predictors, and standard error was 6.337 quite high.
Further, all predictors considered in the multiple regression have significant
impacts in the fitting of the intercept as illustrated in Table 4. Maximum temperature,
average temperatures, and dew and precipitation have positive effects on tempera-
ture, while the effects of minimum temperature and humidity are negative. On the
p-value interpretation, all of the predictors are significant since they are all below
the 0.15 standard error for p-value.
Applying multiple regression approaches on that dataset finds out predictable
equation between temperature and other predictors. So MLR was computed given
below.

Yi = −67.3909377 + 2.97551657 ∗ X 1 − 4.725115525 ∗ X 2

+ 1.67216714 ∗ X 3 + 4.735493886 ∗ X 4
+ 7.163066034 ∗ X 5 − 0.203586847 ∗ X 6 (3)

where
X1 average temperature
X2 minimum temperature
X3 maximum temperature
X4 dew
X5 precipitation and
X6 humidity.
Based on the dataset, example numbers are X 1 = 17, X 2 = 13, X 3 = 19, X 4 =
0.3052, X 5 = 9.8, and X 6 = 72; the value of the multiple regression would be Y i =
10.522, which signifies that there are 10.52% frost that will occur with that specific
numbers of predictors.
Based on the residual output of the dataset, the predicted value was shown and
plotted as shown in Table 5.
Frost Prediction in Highland Crops Management … 281

Table 5 Residual output

Observation Predicted t
1 10.522585
2 10.35941
3 8.8045597
4 10.220742
5 12.662168
6 13.622304
7 15.304227
8 11.030896
9 5.9691042
10 11.31716
11 9.1312452
12 9.6276945
13 10.957603
14 11.361045
15 11.178337
16 11.543338
17 15.226635
18 17.720764
19 14.660084
20 11.913541
21 15.0971
22 27.351176
23 13.473796
24 28.093354
25 17.358863
26 28.106629
27 22.230981
28 21.220673

It signifies that during the 24th month which falls last December 2018, there is
a high tendency that frost occurs and during 26th month which falls on February
also a high tendency that frost will occur. Results show that frost did occur during
December 2018 and February 2018.
The researchers plot graphs as illustrated in Figs. 10 and 11 that depict the relation-
ship between the actual value of temperature data and predicted value of temperature
and the actual value of the average temperature and the relative humidity using mul-
tiple linear regression equation, and from the graph, it is observed that MLR method
282 J. M. Mendez and D. D. Dasig

Fig. 10 Comparison of the average temperature and relative humidity

Fig. 11 Comparison of the actual temperature and the forecasted temperature

for prediction of temperature achieves closer values between actual and predicted
temperature values.
Also, the residual plots of the predictors during the testing were taken as shown
in Fig. 12.
The line fit plot of the residual is compared with the predicted value based on the
residual output of the dataset. It shows that all of the predictors are in-line with the
real-time values as with the predicted values as shown in Fig. 13.
Another figure describes the comparison of the real-time dew and the forecasted
dew based on the multiple linear regressions taken from the dataset as illustrated in
Fig. 14.
The comparison between the real-time precipitation and the forecasted is also
shown in Fig. 15.
Figure 16 shows the comparison of the humidity based on the dataset taken as
during the testing.
Frost Prediction in Highland Crops Management … 283

Fig. 12 Residual plots of the predictors

The comparison between the real-time minimum temperature and the forecasted
minimum temperature signifies that the line graph was correlated with each other
also illustrated in Fig. 17.
On the basis of the results, the multiple correlation and regression could be suc-
cessfully used to analyze the relationships between the temperature, humidity, dew,
and precipitation. Frost could be easily predicted using the multiple linear regression
function. The graphs above prove the efficiency of the regression. With larger dataset,
better and more precision can be achieved.
284 J. M. Mendez and D. D. Dasig

Fig. 13 Residual line fit

Fig. 14 Comparison of the real-time dew and forecasted dew

Frost Prediction in Highland Crops Management … 285

Fig. 15 Comparison of the real-time precipitation and the forecasted precipitation

Fig. 16 Comparison of real-time humidity and forecasted humidity

Fig. 17 Comparison of the real-time minimum temperature and forecasted temperature

286 J. M. Mendez and D. D. Dasig

5 Conclusions and Recommendation

This chapter provided the rudimentary and concepts of Precision Agriculture, and
applications of the Internet of Things (IoT) in agriculture, the design and challenges
of an IoT-enabled system for highland crops management, and utilized Multiple
Regression as a frost prediction technique. The study was conducted for the Province
of Benguet of Cordillera Administrative Region (CAR) in the Philippines. Benguet
is one of the highland provinces in the Philippines, and also known as the Salad Bowl
of the Philippines being the huge producers of upland vegetables [25–27]. During the
pilot testing, tests were conducted in the crops and vegetable farms in Atok, Benguet,
16° 35 N 120° 42 E with 7400 ft elevation and most frequently affected by frost
and climactic variations. Atok has a mountainous land area, with many steep cliffs.
Accordingly, the Philippine Highway System’s highest point is Atok’s Barangay
Paoay at 2255 m (7400 ft) above sea level [25, 26].
The researchers employed the Design Science Research Methodology (DSR).
This methodology guided the researchers to propose, develop, and deploy an IoT-
based technology for highland crop management as a new development in precision
farming to improve the farmers’ livelihood techniques and their social aspect through
modifications on the processes and technologies; as well as employed Modified
Iterative and Incremental Project development methodology. Based on the interview
responses, and active observations, the farmers relied on indigenous knowledge in
frost forecasting.
To help the vegetable farmers in frost forecasting and mitigate its adverse effects
to their farms, the IoT-based Highland Crop Management was developed using the
Internet-based technical architecture “IoT architecture.” Based on the results, the
multiple correlation and regression could be successfully used to analyze the rela-
tionships between the temperature, humidity, dew, and precipitation. Frost could
be easily predicted using the multiple linear regression function. The graphs above
prove the efficiency of the regression. With larger dataset, better and more precision
can be achieved.
Further studies could be done on IoT application in agriculture, such as the use of
blockchain technology in the lifecycle of agricultural production system, financing
and product marketing, applying other machine learning techniques to help predict
climatic variations, and other bioinformatics and biomedical techniques for plant and
crop management.

References

1. Marasigan, A.R.: Characterization of frost events in Benguet, Philippines. Published Master’s

Thesis for the degree in meteorology. Institute of Environmental Science and Meteorology,
University of the Philippines, Diliman (2017)
2. Hansen, J., Marx, S., Weber, E.: The role of climate perceptions, expectations, and forecasts
in farmer decision making. The Argentine Pampas and South Florida. International Research
Institute for Climate Prediction (IRI) the Earth Institute, Columbia University (2014)
Frost Prediction in Highland Crops Management … 287

3. Kalma, J.D., Laughlin, G.P., Caprio, J.M., Hamer, P.J.C.: The Bioclimatology of Frost: Its
Occurrence, Impact and Prediction. Springer, Berlin (2012)
4. National Statistics Offfrost: Cut flower industry: an update. J. Philippine Statist. 59(2) Sec-
ond Quarter 2008. Manila, Philippines. Retrieved from https://psa.gov.ph/sites/default/files/
2ndQ2008.pdf, Accessed 28 Aug 2018
5. National Statistical Coordination Board, October 2011. Philippine Statistical Yearbook (2011).
Retrieved from https://psa.gov.ph/sites/default/files/2011_PSY_0.pdf, 28 Aug 2018
6. Fagyan, A.: 8 villages in Benguet susceptible to climate change. SunStar Baguio. 07
June 2017. Retrieved from http://www.sunstar.com.ph/baguio/opinion/2017/06/07/8-villages-
benguetsusceptible-climate-change-546212, 4 July 2018
7. Piadozo, M.E.S.: Efficiency of Benguet vegetable prfrost linkages. J. Int. Soc. Asian Agric.
Sci 19(2), 104–119 (2013)
8. Hoffmann, N.E., Bischof, J.C.: The cryobiology of cryosurgical injury. Urology 60(2), 40–49
(2002)
9. Petzold, G., Aguilera, J.M.: Frost morphology: fundamentals and technological applications
in foods. Food Biophys. 4(4), 378–396 (2009)
10. Snyder, R.L., Melo-Abreu, J.D.: Frost protection: fundamentals, practfrost and economics.
Frost Prot. Fundam. Practfrost Econ. 1, 1–240 (2005)
11. Battany, M.C.: Vineyard frost protection with upward-blowing wind machines. Agric. For.
Meteorol. 157, 39–48 (2012)
12. Moon, A., Kim, J., Zhang, J., Liu, H., Son, S.W.: Understanding the impact of lossy compres-
sions on IoT smart farm analytics. In: 2017 IEEE International Conference on Big Data (Big
Data), pp. 4602–4611. IEEE (2017, December)
13. Vuran, M.C., Salam, A., Wong, R., Irmak, S.: Internet of underground things in precision
agriculture: architecture and technology aspects. Ad Hoc Netw. 81, 160–173 (2018)
14. Lee, S.K., Kader, A.A.: Preharvest and postharvest factors influencing vitamin C content of
horticultural crops. Postharvest Biol. Technol. 20(3), 207–220 (2000)
15. Rosenzweig, C., Iglesias, A., Yang, X.B., Epstein, P.R., Chivian, E.: Climate change and
extreme weather events; implications for food production, plant diseases, and pests. Global
Change Human Health 2(2), 90–104 (2001)
16. Mermut, A.R., Eswaran, H.: Some major developments in soil science since the mid-1960s.
Geoderma 100(3–4), 403–426 (2001)
17. Joshi, P.K., Gulati, A., Birthal, P.S., Tewari, L.: Agriculture diversification in South Asia:
patterns, determinants and policy implications. Econ. Polit. Week. 2457–2467 (2004)
18. Piao, S., Ciais, P., Huang, Y., Shen, Z., Peng, S., Li, J., Friedlingstein, P.: The impacts of climate
change on water resources and agriculture in China. Nature 467(7311), 43 (2010)
19. Napaldet, J.T., Buot Jr., I.E.: Floral diversity assessment of Balili river as potential phytoreme-
diators. Wetlands Biodiv. 7, 17–28 (2017)
20. Napaldet Jones, T.: Diversity of aquatic macrophytes in Balili River, La Trinidad, Benguet,
Philippines as potential phytoremediators. Biodiver. J. Biol. Div. 20(4), 1048–1054 (2019)
21. Jenouvrier, S.: Impacts of climate change on avian populations. Glob. Change Biol. 19(7),
2036–2057 (2013)
22. McMichael, A.J., Campbell-Lendrum, D.H., Corvalán, C.F., Ebi, K.L., Githeko, A., Scheraga,
J.D., Woodward, A.: Climate change and human health: risks and responses. World Health
Organization (2003)
23. Peffers, K., Tuunanen, T., Rothenberger, M.A., Chatterjee, S.: A design science research
methodology for information systems research. J. Manag. Inf. Syst. 24(3), 45–77 (2007)
24. Baskerville, R., Pries-Heje, J., Venable, J.: Soft design science methodology. In: Proceedings
of the 4th International Conference on Design Science Research in Information Systems and
Technology (p. 9). ACM (2009, May)
25. San Luis, G.D., Balangcod, T.D., Abucay Jr., J.B., Wong, F.M., Balangcod, K.D., Afifi, N.I.G.,
Apostol, O.G.: Phytochemical and Antimicrobial Screening of Indigenous Species That Have
Potential for Revegetation of Landslides in Atok. Benguet, Philippines (2014)
288 J. M. Mendez and D. D. Dasig

26. Barcelo, J.M., Barcelo, R.C., Alvarez, A.: Ochratoxin A, fungal contamination and antioxidant
property of defective Arabica coffee in Benguet, Philippines. Emirates J. Food Agri. 10–17
(2017)
27. Simongo, D.K., Gonzales, I.C., Sagalla, E.J.D.: Evaluation of potato entries for yield and fry
quality grown in different elevations of Benguet, Philippines. J. ISSAAS Int. Soc. Southeast
Asian Agric. Sci. 17(2), 117–127 (2011)
28. Stehman, S.: Comparison of systematic and random sampling for estimating the accuracy
of maps generated from remotely sensed data. PE RS-Photogramm. Eng. Rem. Sens. 58(9),
1343–1350 (1992)
29. Deeks, J.J., Macaskill, P., Irwig, L.: The performance of tests of publication bias and other
sample size effects in systematic reviews of diagnostic test accuracy was assessed. J. Clin.
Epidemiol. 58(9), 882–893 (2005)
30. Onwuegbuzie, A.J., Leech, N.L.: Sampling designs in qualitative research: making the sampling
process more public. Qual. Report 12(2), 238–254 (2007)
31. Larman, C., Basili, V.R.: Iterative and incremental developments: a brief history. Computer
36(6), 47–56 (2003)
32. Chomicki, J.: Iterative modification and incremental evaluation of preference queries. In: Inter-
national Symposium on Foundations of Information and Knowledge Systems, pp. 63–82.
Springer, Berlin (2006, February)
33. Daniel, D.D.: An evaluative study on project management tools for managing engineering
projects. Ascendens Asia J. Multidisc. Res. Abstr. 1(1) (2017)
34. Nyssen, C.: An efficient and accurate iterative method, allowing large incremental steps, to
solve elasto-plastic problems. In: Computational Methods in Nonlinear Structural and Solid
Mechanics, pp. 63–71. Pergamon (1981)
35. Ashton, K.: That ‘internet of things’ thing. RFID J 22(7), 97–114 (2009)
36. Atzori, L., Iera, A., Morabito, G.: The internet of things: a survey. Comput. Netw. 54(15),
2787–2805 (2010)
37. Dasig, D.D.: A study on the sectors of economy servfrostd by pre-industry system developers
among companies in Metro Manila: A tool for business reengineering. arXiv preprint arXiv:
1409.7277 (2014)
38. Gubbi, J., Buyya, R., Marusic, S., Palaniswami, M.: Internet of things (IoT): a vision, archi-
tectural elements, and future directions. Fut. Gen. Comput. Syst. 29(7), 1645–1660 (2013)
39. Dasig Jr., D.: A frontier in organizational and business process innovation in servfrost manage-
ment through lean six sigma Kaizen project implementation. J. Adm. Bus. Stud. 3(6), 263–283
(2017)
40. Dasig, D.D., Benosa, M.C., Pahayahay, A.B., Asejo, N.R., Tadeo, C., Ellaine, M., Sabado,
D.C.: Globalization of technology capability in these technological constellations. Int. J. Bus.
Econ. Affairs, 106–115
41. Dasig, D.: User experience of embedded system students on Arduino and field programmable
gate array (FPGA). In: Proceedings of the Second International Conference on Advances in
Applied Science and Environmental Engineering-ASEE, pp. 124–128 (2014, February)
42. Gubbi, J., Buyya, R., Marusic, S., Palaniswami, M.: Internet of Things (IoT): a vision, archi-
tectural elements, and future directions. Future Gen. Comput. Syst. 29(7), 1645–1660 (2013)