Scholar Publishing Group

Water Pollution Prevention and Control Project

https://doi.org/10.38007/WPPCP.2020.010202
ISSN 2790-2730 Vol. 1, Issue 2: 11-19

Prediction of Water Quality Monitoring Indicators Based

on Random Forest
Erika Ottaviano*
Furtwangen University, 78120 Furtwangen, Germany
*
corresponding author

Keywords: Water Quality Monitoring, Indicator Monitoring, Water Quality Forecast,

Random Forest

Abstract: Water resources are not only the material basis to support the current social and
economic development, but also an important resource to maintain people’s daily life.
Therefore, the protection of water resources from being infringed not only requires the
cooperation of relevant researchers, but also is closely related to every resident in the
society. Everyone needs to pay more attention to it. In the current society, the water
environment in different regions is generally protected through multiple processes. Among
them, the first line of defense is the monitoring of water quality and the prediction model of
indicators. The monitoring and early warning model of water quality indicators is generally
based on the real-time monitoring of the content of various substances in the water
environment, so as to comprehensively evaluate the water environment and predict the
water quality in the future period according to the concentration of different substances.
The existing water quality indicator monitoring and prediction model not only is an
important technology to control water environmental pollution in the current society, but
also can reflect the quality status in the target waters and the future water quality
development trend in all aspects. More professional water quality related data can be
collected through the existing monitoring models of various indicators of water quality.
Based on the in-depth analysis of these data, the main pollutants in the target waters, the
current pollution situation and the possible future damage can be fully understood. On the
other hand, this water quality indicator monitoring model has also provided assistance for
professionals to develop more scientific and reasonable water pollution control plans. In
this paper, the existing water quality index monitoring and prediction model was updated
by using the stochastic forest algorithm model. The stochastic forest algorithm model
generally predicts the indicators and pollution status of the target water area in the future by
processing the relevant data of the water quality indicators of the target water area. Finally,
the performance of this new water quality monitoring index and prediction model has been
improved by about 26% on average.

Copyright: © 2020 by the authors. This is an Open Access article distributed under the Creative Commons Attribution License (CC BY 4.0), which
permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited
(https://creativecommons.org/licenses/by/4.0/).

11
 Water Pollution Prevention and Control Project

1. Introduction

Water quality monitoring is one of the key processes in the water resources protection process.
The formulation of water resources protection plan and treatment measures also need to analyze the
water quality monitoring data before they can be implemented. Although the existing monitoring
technology for multiple indicators of water quality has the advantages of high monitoring efficiency
and accurate water quality data collection, it can not meet more requirements in the current field of
water environmental protection. Therefore, more relevant researchers are needed to explore this, so
as to propose a better monitoring and prediction model for water quality indicators.
Some researchers in the field of water resources protection have explored the development of
water quality in recent years, and put forward a variety of water resources protection suggestions.
Solangi Ghulam Shabir explored the performance of water quality index, comprehensive pollution
index and geographical assessment tools in evaluating groundwater quality. Through the
comprehensive performance of the three models, he determined that the water quality index had a
good performance in the local groundwater quality assessment [1]. Alizadeh Mohamad Javad
explored the role of machine learning models in evaluating the water quality of a basin. Through the
analysis of machine learning and water quality evaluation model, he determined that it had a good
performance in water quality evaluation [2]. Zhi Wei explored the performance of deep learning
model in river water quality assessment. Through his research on the deep learning model, he
determined that the model could accurately evaluate water quality [3].
Camara Moriken explored the impact of land use rate on water quality in a certain area, and
determined that the large area of land use had a great impact on water quality [4]. Son Cao Truong
explored the role of water quality and pollution index in the comprehensive evaluation of water area
and determined its reliability [5]. Hamid Aadil explored some factors affecting water quality and
determined that changes in various factors in the environment could have a great impact on water
quality [6]. Muharemi Fitore explored a machine learning method for detecting water quality
anomalies and determined the reliability of this method [7]. However, these water resources
protection proposals can only slightly improve the existing water resources protection model, and
can not directly promote the substantial growth of water resources protection performance.
In addition, some researchers have explored the monitoring and prediction methods of water
resources, hoping to get a perfect monitoring and prediction model. Bisht Anil Kumar mainly
explored the water quality prediction model in a certain area and determined the feasibility of
artificial intelligence in the water quality prediction model [8]. Yang Huanhai explored the water
quality prediction model in multi-scale aquaculture and determined the reliability of this multi-scale
water quality prediction model [9]. Haghiabi Amir Hamzeh explored the role of a machine learning
method in the water quality prediction model, and verified the feasibility of this water quality
prediction model through experiments [10]. Barzegar Rahim explored the role of a hybrid algorithm
model in water quality prediction and determined the feasibility of this model [11]. Hassan Md
Mehedi explored the role of an algorithm model in the prediction of water quality index and
determined the reliability of this algorithm model [12]. However, most of these studies can not meet
the current social requirements for water resources monitoring and prediction. Therefore, more
in-depth research is needed.
The main purpose of this paper was to solve the shortcomings of the existing water quality
indicator monitoring model that can not analyze and process multiple water quality indicator data
and predict the water quality environment. Stochastic forest algorithm model was used to help the
existing water quality indicator monitoring model complete in-depth analysis of multiple types of
water quality data. The water resources and environment of the target waters were accurately
predicted to help the relevant staff formulate more reasonable water quality protection plans.

12
 Water Pollution Prevention and Control Project

2. Prediction of Water Quality Monitoring Indicators

Water quality monitoring has been an important basic work in current water quality management.
This work is mainly to provide basic data support for the protection and treatment of water areas
through real-time monitoring of the pollutants in the target water area and the factors that cause
water pollution [13]. However, the early water quality monitoring work is generally only the
detection of chemicals in the waters of rivers and lakes. However, with the further acceleration of
the process of social industrialization, the damage to water resources in the process of social
development is also increasing, which also urges the water quality detection mode to gradually
evolve into a professional water quality monitoring department and model to detect the water
pollution in most waters. Moreover, with the in-depth research of relevant researchers on the water
quality monitoring mode, the projects in the water quality monitoring have been constantly
improved, and the monitoring technology has also moved towards informatization and
standardization. The general flow of water quality indicator monitoring is shown in Figure 1.
Database
Online
Computing
monitoring
center
equipment

Control page

Data
cleaning Data
classification

Data
operation

Target Monitoring
waters unit

Figure 1. Schematic diagram of the general process of water quality index monitoring

In the current monitoring mode of water quality indicators, it is first necessary to clarify the
objectives to be monitored. Then, in different water areas, appropriate water quality indicator
detection schemes are used to detect multiple types of pollutants in the target water area, thus
providing scientific and effective data support for the protection and treatment of water areas. The
existing water quality indicator monitoring models generally include a variety of water quality
detection methods, including quality analysis, titration analysis, acid-base titration, and so on. These
methods also have different functions and detection effects in detecting different waters. Among
them, the quality analysis method is generally to separate various substances in the sampled water
resources, and then measure the content of various substances. The titration analysis method is to
conduct chemical reaction on the sampled water resources by chemical means, and then detect the

13
 Water Pollution Prevention and Control Project

substances in the water according to the reaction performance. The working flow of the algorithm in
the water quality indicator monitoring and prediction system is shown in Figure 2.
Sample
database

Data Data
cleaning association
analysis

Test
set

Data
inspection
Model
building

Training
effect
evaluation
Data
Dataset training
split

Water Result
quality evaluation
prediction

Figure 2. Schematic diagram of the algorithm workflow in the water quality index monitoring and
forecasting system

3. Random Forest Analysis

Most of the existing water quality indicator monitoring models use manual methods to sample
and detect water resources in the waters. With the continuous development of time, various
problems have gradually deepened, and can no longer meet the current requirements of water
quality monitoring. Therefore, a new automatic water quality indicator monitoring mode is needed
to replace this manual mode. This paper completes the construction of this automatic water quality
indicator monitoring mode by using random forest and a variety of intelligent sensor equipment.
The automatic water quality indicator monitoring mode uses correlation analysis, multiple
regression analysis and random forest algorithm model to optimize the operation mode in water
quality indicator monitoring. Among them, the random forest algorithm model is mainly a
supervised learning algorithm, which can solve multiple classification and regression problems by
constructing a set of decision trees [14]. With the continuous development of random forest
algorithm, it has been applied in many fields such as medicine, finance, e-commerce, etc. [15].
Benefiting from the powerful data relation mining ability of the current stochastic forest algorithm
model, it can quickly analyze and calculate various independent variables in the monitoring mode

14
 Water Pollution Prevention and Control Project

of water quality indicators, so as to obtain more accurate data of water quality indicators. This also
enables water quality to be monitored more quickly after being polluted, thus helping relevant
researchers to propose scientific water quality control plans. In addition, the stochastic forest
algorithm model can also make relatively accurate prediction of water pollution in the future
through the analysis of existing water quality data. The work flow of random forest in water quality
indicator monitoring mode is shown in Figure 3.
Establishment of
water quality data
indicators filtering

Parameter
Classification
model
building

model
evaluation Correlation
data training
analysis

Parametric Water
analysis quality
warning

classification
model

Figure 3. Schematic diagram of the workflow of random forest in the monitoring mode of water
quality indicators

4. Random Forest Algorithm

In recent years, with the rapid development of water quality indicator monitoring mode, not only
more water quality indicators can be monitored, but also the cost of water quality monitoring is
further reduced. However, the development speed of this water quality indicator monitoring mode
still cannot meet some requirements of current water resources protection. First of all, the amount of
data that needs to be processed in the current water resources protection has increased significantly,
and this surge in data volume has also increased the cost of processing data. Therefore, in order to
further improve the efficiency of data processing, data are analyzed more effectively, so as to
comprehensively reduce the cost of water quality indicator monitoring mode, which requires more
algorithm models to be integrated into the water quality indicator monitoring mode. This paper
mainly analyzes the correlation between the data of various indicators by using the stochastic forest
correlation algorithm model. This analysis can also measure the tightness of the relationship
between multiple variable data, so that a data operation model can be obtained by processing these

15
 Water Pollution Prevention and Control Project

data to predict the water quality more accurately.

First, the sampling algorithm is used to preliminarily analyze the data in the water quality
indicator monitoring model, so as to clean and classify the water quality indicators, as shown in
Formula (1).

(1)
Among them, n represents the type data amount of sample data. Then, two data averaging
methods are used to calculate the sample data, so as to determine the prediction of water quality
through the calculation process of the sample data, as shown in Formula (2).

(2)
Among them, represents the output value of the sample data, and T represents the vector
dimension. Then, the weighted average method is used to fine predict the quality of water quality,
as shown in Formula (3).

(3)
Among them, represents the weight value of the model. The addition of random forest
algorithm model not only further reduces the data operation cost of the existing water quality
monitoring model, but also greatly improves the efficiency of data operation. This also makes the
results of water quality monitoring more timely and plays a greater role in the protection of water
resources. On the other hand, the integration of random forest also enables the water quality
indicator monitoring model to predict the water quality pollution in the future through the analysis
of water quality indicator data.

5. Prediction Experiment of New Water Quality Monitoring Indicators

The high-quality development of social economy has not only promoted the reform of many
industries in the society, but also caused more pollution to the water resources and environment in
its region. This situation has also attracted the attention of environmental protection staff. In order
to protect the existing water resources so that they can be used sustainably, first of all, the water
resources in the region need to be better protected and treated. The first and most basic step of the
protection and treatment of water resources is the monitoring of water quality. Only through the
monitoring data of water quality can reasonable water resources protection and treatment strategies
be formulated. On the other hand, with the continuous progress of society, people have higher and
higher requirements for the function of water quality indicator monitoring mode. This high
requirement is further promoting the high-quality development of water quality indicator
monitoring mode and making due contributions to the protection and treatment of water resources.
In this paper, the performance of this new water quality index monitoring and prediction model
combined with random forest algorithm model was experimentally studied, and the superior
performance of this new water quality index monitoring and prediction model combined with
random forest was determined. At the same time, its monitoring ability in various water quality
indicators was analyzed.
First of all, the concentration of various substances in a section of water and their impact on the
water resources in this section of water were analyzed, as shown in Table 1.

16
 Water Pollution Prevention and Control Project

Table 1. Effect of concentration of various substances in waters on the condition of their waters
Concentration Water conditions
Turbidity 6.7 5.4
Chloride ion 7.2 6.2
Suspension 6.6 4.8
Electrical conductivity 7.1 6.5
The monitoring mode of water quality indicators is mainly a technology for real-time monitoring
of water bodies, which plays a major role in determining the categories and concentrations of
pollutants that may exist in a section of water. At the same time, in the monitoring process of water
quality indicators, the overall pollution status of the water body can also be evaluated through the
classification of pollutants and the change trend of water quality. The current monitoring mode of
water quality indicators can generally be divided into two modes: surface water quality monitoring
and groundwater monitoring. Although the two water quality indicator monitoring modes are
different for the monitored water bodies, their monitoring objectives are mostly the same. After
analyzing the concentration of four different substances in the water body and the overall situation
of the water body in Table 1, it can be determined that the turbidity and suspended solids
concentration of the water body have a great impact on the pollution status of the water body, which
also provides a reliable reference for the protection and treatment modes of different water bodies.
Then, the performance of the new water quality indicator monitoring and prediction model
proposed by the random forest and the existing water quality indicator monitoring model in many
aspects was analyzed, as shown in Figure 4.

a. Existing water quality indicators monitoring model performance diagram

b. New water quality indicators monitoring mode performance schematic

Figure 4. Schematic representation of the performance of the new water quality indicator
monitoring and prediction model and the existing water quality indicator monitoring model under
random forest

With the development of society, the monitoring of water quality needs to be carried out more

17
 Water Pollution Prevention and Control Project

and more frequently, so as to grasp the pollution status of the water area more timely, so as to
ensure the sustainability of the water area. Generally speaking, the water quality indicator
monitoring mode can fully describe the parameters of the target water quality. With the progress of
society, various substances in the water have also begun to increase rapidly. The performance of the
current water quality indicator monitoring mode is also beginning to fail to meet the requirements.
Through the analysis of the performance of the existing water quality indicator detection mode in
the four aspects of pollutant analysis, water pollution status, water pollution prediction and work
efficiency in Figure 4a, it is determined that the performance of the existing water quality indicator
detection mode in the prediction of water pollution cannot meet the expected goal, and this module
needs to be optimized in combination with other technologies. On the other hand, after the analysis
of the performance of the new water quality index detection and prediction model combined with
random forest in the same four aspects in Figure 4b, it is determined that the new water quality
index detection and prediction model combined with random forest has the highest improvement in
water pollution prediction, and other aspects have also improved. Finally, in a comprehensive way,
the performance of this new water quality index detection and prediction model has improved by
about 26% on average.

6. Conclusion

With the rapid development of various modern information technologies, the heavy industry and
other related industries in society have developed well. However, in this rapid development process,
these industries have caused great damage to the local ecological environment in many aspects.
This destruction of the ecological environment is also seriously endangering the quality of life of
residents in the region, and has a serious negative impact on the health of residents. Therefore, with
the development of technology and the improvement of people’s awareness of ecological
environmental protection, more and more researchers are studying the environmental protection
model. Among them, the water quality indicator monitoring model is a key technology in the water
resources environmental protection model. The monitoring of multiple indicators in the water
quality of different waters can not only help the environmental protection personnel in the region to
have a deeper understanding of the water pollution situation, but also analyze the main causes of
water pollution. In addition, the analysis of such water quality data can also help the relevant water
resources protection staff to build a more scientific and reasonable water resources protection model
and better control the current situation of water pollution. Through in-depth analysis of the
stochastic forest algorithm model, this paper determined that this algorithm model could be better
applied in the current water quality indicator monitoring model, and provided a better role for the
water quality prediction model. This water quality monitoring and prediction model based on
random forest can predict the pollution status and pollution sources of the target water area in a
certain extent through the analysis and calculation of various water quality data, thus helping
relevant staff to carry out prevention work in advance.

Funding

This article is not supported by any foundation.

Data Availability

Data sharing is not applicable to this article as no new data were created or analysed in this
study.

18
 Water Pollution Prevention and Control Project

Conflict of Interest

The author states that this article has no conflict of interest.

References

[1] Solangi Ghulam Shabir. Groundwater quality evaluation using the water quality index (WQI),
the synthetic pollution index (SPI), and geospatial tools: a case study of Sujawal district,
Pakistan. Human and Ecological Risk Assessment: An International Journal. (2020) 26(6):
1529-1549. https://doi.org/10.1080/10807039.2019.1588099
[2] Alizadeh Mohamad Javad. Effect of river flow on the quality of estuarine and coastal waters
using machine learning models. Engineering Applications of Computational Fluid Mechanics.
(2018) 12(1): 810-823. https://doi.org/10.1080/19942060.2018.1528480
[3] Zhi Wei. From hydrometeorology to river water quality: can a deep learning model predict
dissolved oxygen at the continental scale? Environmental science & technology. (2020) 55(4):
2357-2368. https://doi.org/10.1021/acs.est.0c06783
[4] Camara Moriken, Nor Rohaizah Jamil, Ahmad Fikri Bin Abdullah. Impact of land uses on
water quality in Malaysia: a review. Ecological Processes. (2019) 8(1): 1-10.
https://doi.org/10.1186/s13717-019-0164-x
[5] Son Cao Truong. Assessment of Cau River water quality assessment using a combination of
water quality and pollution indices. Journal of Water Supply: Research and Technology-Aqua.
(2020) 69(2): 160-172. https://doi.org/10.2166/aqua.2020.122
[6] Hamid Aadil, Sami Ullah Bhat, Arshid Jehangir. Local determinants influencing stream water
quality. Applied Water Science. (2020) 10(1): 1-16. https://doi.org/10.1007/s13201-019-1043-4
[7] Muharemi Fitore, Florin Leon. Machine learning approaches for anomaly detection of water
quality on a real-world data set. Journal of Information and Telecommunication. (2019) 3(3):
294-307. https://doi.org/10.1080/24751839.2019.1565653
[8] Bisht Anil Kumar. Artificial neural network based water quality forecasting model for ganga
river. International Journal of Engineering and Advanced Technology. (2019) 8(6): 2778-2785.
https://doi.org/10.35940/ijeat.F8841.088619
[9] HuanhaiYang, Shue Liu. A prediction model of aquaculture water quality based on multiscale
decomposition. Mathematical Biosciences and Engineering. (2020) 18(6): 7561-7579.
[10] Haghiabi Amir Hamzeh, Ali Heidar Nasrolahi, Abbas Parsaie. Water quality prediction using
machine learning methods. Water Quality Research Journal. (2018) 53(1): 3-13.
https://doi.org/10.2166/wqrj.2018.025
[11] Barzegar Rahim, Mohammad Taghi Aalami, Jan Adamowski. Short-term water quality
variable prediction using a hybrid CNN-LSTM deep learning model. Stochastic Environmental
Research and Risk Assessment. (2020) 34(2): 415-433.
https://doi.org/10.1007/s00477-020-01776-2
[12] Hassan Md Mehedi. Efficient prediction of water quality index (WQI) using machine learning
algorithms. Human-Centric Intelligent Systems. (2020) 1(3-4): 86-97.
https://doi.org/10.2991/hcis.k.211203.001
[13] Ahmed Umair. Water quality monitoring: from conventional to emerging technologies. Water
Supply. (2020) 20(1): 28-45. https://doi.org/10.2166/ws.2019.144
[14] Resende, Paulo Angelo Alves, Andre Costa Drummond. A survey of random forest based
methods for intrusion detection systems. ACM Computing Surveys (CSUR). (2018) 51(3): 1-36.
https://doi.org/10.1145/3178582
[15] ]Schonlau Matthias, Rosie Yuyan Zou. The random forest algorithm for statistical learning.
The Stata Journal. (2020) 20(1): 3-29. https://doi.org/10.1177/1536867X20909688
19