Article

Current and Potential Land Use/Land Cover (LULC) Scenarios in

Dry Lands Using a CA-Markov Simulation Model and the
Classification and Regression Tree (CART) Method:
A Cloud-Based Google Earth Engine (GEE) Approach
Elsayed A. Abdelsamie 1 , Abdel-rahman A. Mustafa 2 , Abdelbaset S. El-Sorogy 3 , Hanafey F. Maswada 4 ,
Sattam A. Almadani 3 , Mohamed S. Shokr 5, * , Ahmed I. El-Desoky 6 and Jose Emilio Meroño de Larriva 7

1 National Authority for Remote Sensing and Space Sciences, Cairo 1564, Egypt;
[email protected]
2 Soil and Water Department, Faculty of Agriculture, Sohag University, Sohag 82524, Egypt;
[email protected]
3 Geology and Geophysics Department, College of Science, King Saud University, P.O. Box 2455, Riyadh 11451,
Saudi Arabia; [email protected] (A.S.E.-S.); [email protected] (S.A.A.)
4 Agricultural Botany Department, Faculty of Agriculture, Tanta University, Tanta 31527, Egypt;
[email protected]
5 Soil and Water Department, Faculty of Agriculture, Tanta University, Tanta 31527, Egypt
6 Department of Soils and Water, Faculty of Agriculture, Al-Azhar University, Assiut 71524, Egypt;
[email protected]
7 Department of Graphic Engineering and Geomatics, Campus de Rabanales, University of Cordoba,
14071 Cordoba, Spain; [email protected]
* Correspondence: [email protected]

Citation: Abdelsamie, E.A.; Mustafa,

Abstract: Rapid population growth accelerates changes in land use and land cover (LULC), strain-
A.-r.A.; El-Sorogy, A.S.; Maswada, ing natural resource availability. Monitoring LULC changes is essential for managing resources
H.F.; Almadani, S.A.; Shokr, M.S.; and assessing climate change impacts. This study focused on extracting LULC data from 1993 to
El-Desoky, A.I.; Meroño de Larriva, 2024 using the classification and regression tree (CART) method on the Google Earth Engine (GEE)
J.E. Current and Potential Land platform in Qena Governorate, Egypt. Moreover, the cellular automata (CA) Markov model was
Use/Land Cover (LULC) Scenarios in used to anticipate the future changes in LULC for the research area in 2040 and 2050. Three multi-
Dry Lands Using a CA-Markov spectral satellite images—Landsat thematic mapper (TM), enhanced thematic mapper (ETM+), and
Simulation Model and the operational land imager (OLI)—were analyzed and verified using the GEE code editor. The CART
Classification and Regression Tree
classifier, integrated into GEE, identified four major LULC categories: urban areas, water bodies,
(CART) Method: A Cloud-Based
cultivated soils, and bare areas. From 1993 to 2008, urban areas expanded by 57 km2 , while bare and
Google Earth Engine (GEE) Approach.
cultivated soils decreased by 12.4 km2 and 42.7 km2 , respectively. Between 2008 and 2024, water
Sustainability 2024, 16, 11130. https://
doi.org/10.3390/su162411130
bodies increased by 24.4 km2 , urban areas gained 24.2 km2 , and cultivated and bare soils declined by
22.2 km2 and 26.4 km2 , respectively. The CA-Markov model’s thematic maps highlighted the spatial
Academic Editor: Elisa Grieco
distribution of forecasted LULC changes for 2040 and 2050. The results indicated that the urban areas,
Received: 19 November 2024 agricultural land, and water bodies will all increase. However, as anticipated, the areas of bare lands
Revised: 10 December 2024 shrank during the years under study. These findings provide valuable insights for decision makers,
Accepted: 17 December 2024 aiding in improved land-use management, strategic planning for land reclamation, and sustainable
Published: 19 December 2024 agricultural production programs.

Keywords: CART; drylands; GEE; LULC; CA-Markov

Copyright: © 2024 by the authors.

Licensee MDPI, Basel, Switzerland.
This article is an open access article
distributed under the terms and
1. Introduction
conditions of the Creative Commons Some fertile soils in the Nile Valley and Delta have lost their fertility and production
Attribution (CC BY) license (https:// due to deterioration [1]. Food security, which has emerged as the main concern especially
creativecommons.org/licenses/by/ in developing country [2], must thus be achieved through the expansion of reclaimed
4.0/). soils and the sustainable development of existing agricultural fields [3]. One of the main

Sustainability 2024, 16, 11130. https://doi.org/10.3390/su162411130 https://www.mdpi.com/journal/sustainability

Sustainability 2024, 16, 11130 2 of 19

issues with global environmental change and sustainable development is the change in
LULC. Rapid changes in LULC have resulted from economic activity and rapid global
population growth, which have caused urban agglomeration and subsequent construction
land development [4–6]. Poor farming practices and changes in land-use patterns are some
of the main causes of soil degradation, nutrient cycling disturbance, and negative impacts
on ecosystem services and biodiversity [7].
Because Egypt is an arid country, long-term management and land-use planning
of natural resources depend on the identification and projection of LULC changes. The
physical makeup and attributes of the land elements on the surface of the earth can be char-
acterized as including LULC. LULC is a multifaceted process with causes and consequences
occurring at several social and geographic levels. A basic relationship exists between the
dynamic analysis of LULC changes over time and space and large-scale and long-term
monitoring methodologies [8,9]. While there is still a comparable demand and supply for
land, the need for land is growing. Policies pertaining to LULC must be modified in order
to satisfy this need. Researchers, governments, and international organizations have always
been interested in mapping LULC. In this context, planning and management of natural
resources, research and assessment of environmental quality, and regular monitoring of
existing ecosystems are all dependent on the achievement of accurate LULC products at
different spatial scales (local, national, and global) [10]. Effective land management tech-
niques and policymaking are made possible by these processes, which aid in understanding
the dynamic nature of landscapes [11–13]. Environmental change is acknowledged to be
driven by LULC change at all temporal and geographical dimensions. Along with climate
change, biodiversity loss, and air, soil, and water pollution, these changes ought to be
the top concerns for human populations [14]. Studying how LULC changes over time is
therefore vital and important. Thus, it has become increasingly important for academics
and policymakers worldwide to monitor and mitigate the detrimental effects of LULC
while maintaining the production of necessary resources.
Field surveys and manual visual interpretation served as the primary foundation for
early LULC product mapping. Due to its inefficiency and high expense, this approach
finds it challenging to meet the requirements for fast and routine LULC mapping and
monitoring [15]. The adaptability of remote sensing, GIS, and LULC change models makes
them an effective tool for mapping and monitoring LULC changes. In general, the short-
comings of traditional approaches are that they do not account for environmental, social,
or demographic circumstances. Satellite imagery has long been one of the main sources of
information for the information extraction, analysis, and monitoring of changes in LULC
dynamics. Changes on the earth’s surface can be identified thanks to the availability of
remotely sensed data, which offer a range of temporal, spectral, and spatial resolutions [16].
More and more, precise and trustworthy LULC maps require gathering and analyzing
large amounts of satellite data. Due to its great computational complexity, this work
requires massive storage capacity, strong processing, and adaptability to utilize several
strategies [17]. Furthermore, as classification techniques have advanced, LULC mapping
has made extensive use of machine learning algorithms, such as random forest (RF), gradi-
ent boosting decision tree (GBDT), and support vector machine (SVM) [18,19]. According
to Jin et al. [20], these have created new opportunities for more thorough LULC mapping.
They do, however, also provide significant obstacles to the effectiveness of remote sensing
(RS) data processing and collecting [19–21]. As evidenced by the Google Earth Engine
(GEE), which not only makes full-band and high-intensity image computing feasible but
also enables the integration of multisource and multi-scale global remote sensing data,
scientists are working to grow and develop satellite remote sensing cloud storage and cloud
computing platforms to meet these challenges [22]. One of Google Lab’s products, the
GEE, employs cloud computing to provide high-performance supercomputer services for
raster and vector datasets [17]. In recent years, a big-data cloud-based network computing
resource platform called GEE has surfaced and shown itself to be quite competitive. Ac-
cording to Tamimiminia et al. [23], it offers enormous geographic datasets with previously
Sustainability 2024, 16, 11130 3 of 19

unheard-of online computational and visual analytical capabilities. Researchers can pre-
process or download multi-temporal image data that satisfy filtering requirements using
the GEE platform [24]. LULC classification and analysis can then be performed online using
a variety of machine learning methods. Because of this, the GEE platform is being used
more frequently to produce large-scale, effective LULC products and change-monitoring
studies [25]. GEE provides free access to satellite imagery from several sources, including
Sentinel, Moderate Resolution Imaging Spectroradiometer (MODIS), and Landsat. The
GEE Explorer and GEE Code Editor are the two web-based platforms by which users can
utilize its benefits. With GEE Code Editor, users can create more analyses and presentations
by using JavaScript or Python code programming, while with GEE Explorer, users can
only examine a restricted collection of satellite imagery [26]. GEE is potentially useful for
creating continuous long-term LULC maps because it contains a sizable volume of various
long-term raw remote sensing observations and dynamic classification algorithms, such
as RF models [27,28] and support vector machines (SVMs) [29,30]. The GEE and related
Google Earth platforms have been used as pilot sites for a number of studies, including
mapping vegetation indices, global urban land clipping, surface water monitoring, land
surface temperature, and LULC mapping using Landsat-8 images [31,32].
Many researchers frequently employ the Markov Chain model (MCM), one of the
models used to simulate how LULC develops over time. According to Khwarahm et al. [33]
the MCM quantifies conversion states and the rate of transfer between various land-use
categories. Although the MCM is capable of estimating land use, it lacks spatial explicitness
since it is unable to pinpoint the precise location of changes, which is crucial information
for understanding the effects of forecasted changes. The primary aim of this study was
to develop a CART model on the GEE platform to analyze changes in LULC from 1993
to 2024. The study’s main contribution is the development of a map of land cover in
Qena Governorate, Egypt, covering the years 1993–2008 and 2008–2024. Additionally, the
study analyzed the spatiotemporal changes in LULC from 1993 to 2024 and used spatial
modelling CA-Markov to forecast land-use maps for 2040 and 2050.

2. Materials and Methods

2.1. Description of the Research Area
Qena is the third governorate from the southern border of Egypt, following Aswan
and Luxor. It is situated between latitudes 25◦ 13′′ and 26◦ 15′′ N and longitudes 32◦ 00′′ and
32◦ 55′′ E. As seen in Figure 1, it is bordered to the south by the Luxor Governorate, which
broke away from Qena in 2010, and to the north by the Sohag Governorate, the New Valley
Governorate to the west, and the Red Sea Governorate to the east. The total area of the
Qena Governorate is 9565 km2 (http://qena.gov.eg/Default.aspx accessed on 5 December
2024). The main geographical feature of the Qena Governorate is the 162 km long Nile River,
which flows through the valley from south to north and divides the governorate’s territory
into east and west (http://qena.gov.eg/Default.aspx accessed on 5 December 2024). The
Qena Governorate has a total population of 3,224,573, with 80.3% living in rural areas and
19.7% in cities. Qena is one of the governorates with the highest population densities. Qena
Governorate has a predominantly agricultural economy. The province’s main crops include
sesame, alfalfa, palm trees, sugar cane, bananas, wheat, maize, vegetables (tomatoes), and
a few aromatic and medicinal plants. Sugar cane is the most important crop, followed by
wheat in the winter and maize in the summer [34]. The study area is located in an arid
region of North Africa, with hot summers and mild winters with little to no rainfall
(https://www.eeaa.gov.eg/Uploads/Reports/Files/20221128214734177.pdf accessed on
5 December 2024). The average annual rainfall is about 0.7 mm; rain can fall in April,
November, and December. The daily mean temperature rises significantly throughout the
summer months. The daily average temperature is 25.9 ◦ C, with lows of 17 ◦ C and highs
of 34.1 ◦ C. The average annual evaporation rate is 19.5 mm/day, and relative humidity
is 22% lower in the summer than in the winter [35,36]. Table 1 shows the four major
 the summer months. The daily average temperature is 25.9 °C, with lows of 17 °C and
highs of 34.1 °C. The average annual evaporation rate is 19.5 mm/day, and relative
humidity
Sustainability is 22% lower in the summer than in the winter [35,36]. Table 1 shows 4the
2024, 16, 11130 of 19 four

major land-use types that comprise the research area: cultivated soils, bare soils, urban
regions, and aquatic bodies[34–36]
land-use types that comprise the research area: cultivated soils, bare soils, urban regions,
and aquatic bodies [34–36].

Figure 1. Location of the study area.

Figure 1. Location of the study area.
Table 1. Description of different land-uses of the study area.

Table 1. Description of different land-uses of the study area.

LULC Classes Description

s Urban areas Description

Urban and rural residential, services, commercial, industrial, and roads.
Urban and rural residential, services,
Water bodies
commercial, industrial, and roads.
Nile River, canals, and drainage patterns.
Nile River, canals, and drainage patterns.
Cultivated soils Old cultivated soils in the Nile Valley and recently cultivated soil.
ls Old cultivated soils in the Nile Valley and recently cultivated soil.
Bare soils Soils of eastern and western parts of Nile Valley.
Soils of eastern and western parts of Nile Valley.
2.2. The Data Used
2.2. The Data Used For LULC mapping for the years 1993–2024, three multispectral satellite images from
Landsat Thematic
For LULC mapping for theMapper
yearsTM (15/2/1993–15/12/1993),
1993–2024, Enhanced Thematic
three multispectral Mapper
satellite images
ETM+(15/2/2008–15/12/2008), and Operation Land Imager OLI (15/2/2024–15/5/2024)
from Landsat Thematic Mapper
were used; TM (15/2/1993–15/12/1993),
these images Enhanced
were then compiled and methodically Thematic
processed on the Mapper
GEE
platform [37]. Additionally, this study employed cloud cover
ETM+(15/2/2008–15/12/2008), and Operation Land Imager OLI (15/2/2024–15/5/2024) and air pollution to filter
the data that were accessible through the downloading data gateway to those with less
were used; these images werecover.
than 5% cloud then
Thecompiled andwas
data availability methodically
confirmed usingprocessed on the
the Google Earth CodeGEE
Editor platform throughout the acquisition and processing of the satellite
platform [37]. Additionally, this study employed cloud cover and air pollution to filter images utilizing
the GEE platform and the algorithm that was built. Figure 2 presents the flowchart of the
the data that were approach
accessible through
framework the
used in thedownloading data gateway to those with less
current investigation.
 than 5% cloud cover. The data availability was confirmed using the Google Earth Code
Editor platform throughout the acquisition and processing of the satellite images
utilizing the GEE platform and the algorithm that was built. Figure 2 presents the
flowchart of the approach framework used in the current investigation.
Sustainability 2024, 16, 11130 5 of 19

Figure 2. Flowchart of the methodology framework followed in the current study.

Figure 2. Flowchart of the methodology framework followed in the current study.
2.3. Classification and Regression Tree (CART) Classifier
2.3. Classification
CART wasand Regression
initially Tree
created by(CART)
BreimanClassifier
et al. [38] and was based on regression analysis
andCARTsimple wasif–else decision-making
initially created by Breimanlogic. Through
et al. [38]the anddivision
was basedof training data into
on regression
consecutive
analysis subgroups
and simple (64,decision-making
if–else 53, 251, and 72 samples for urbanthe
logic. Through areas, water of
division bodies, cultivated
training data
lands, and bar soil, respectively), CART constructed a decision tree.
into consecutive subgroups (64, 53, 251, and 72 samples for urban areas, water bodies, Recursive data division
based onlands,
cultivated a predetermined
and bar soil,threshold was the
respectively), result
CART of this phenomenon,
constructed which
a decision tree. persisted
Recursive
until CART achieved non-dividable leaf nodes on the tree. Afterwards, these leaf nodes
data division based on a predetermined threshold was the result of this phenomenon,
on the decision tree were employed as output for forecast, and the decision tree was then
which persisted until CART achieved non-dividable leaf nodes on the tree. Afterwards,
utilized in the test phase to evaluate the performance of the algorithm. Because CART is
these leaf nodes
extremely on the
sensitive decisionintree
to changes the were employed
training dataset, asevenoutput for forecast,
little changes and
in the the
training
decision tree was then utilized in the test phase to evaluate the performance
dataset can produce entirely different results. The main drawback of CART is its excessive of the
algorithm.
reliance on Because
training CART is extremely
datasets. Even so,sensitive
CART isto changes in the training
a straightforward dataset,
model that even
has shown
its durability
little changes inover the time.
trainingBecause
datasetof can
its capacity
producetoentirely
managedifferent
intricateresults.
decisionTherules
mainand
nonlinear interactions, the CART algorithm was selected as
drawback of CART is its excessive reliance on training datasets. Even so, CART is athe classifier [39]. Preparing
training data and
straightforward model utilizing
that hastheshown
chosen itsfeatures
durabilityareover
the time.
first steps
Becausein training the CART
of its capacity to
model. A tree-based decision structure is produced by the algorithm’s recursive dataset
manage intricate decision rules and nonlinear interactions, the CART algorithm was
splitting according to informative attributes [40]. It chooses the most discriminating trait
selected as the classifier [39]. Preparing training data and utilizing the chosen features are
and builds child nodes based on that choice. To optimize the splits, the CART algorithm
the first steps
assesses in training
splitting criteriathe CART
such model. A tree-based
as information gain or Ginidecision
impurity structure is produced
[41]. After by
being trained,
thethe CART model can accurately categorize land cover during the winter by applying It
algorithm’s recursive dataset splitting according to informative attributes [40]. the
chooses
decision therules
most discriminating
it has learned to new traitdata.
andThisbuilds child
study nodesCART
applied basedclassification
on that choice.
usingTothe
optimize the splits, the CART algorithm assesses
“classifier.smileCart” method that is already integrated into GEE.splitting criteria such as information
gain or Gini impurity [41]. After being trained, the CART model can accurately
2.4. Land-Use Change Analysis by CA-Markov Modeling
The Markov analysis approach is a straightforward statistical tool that relies on
a transition probability matrix based on neighborhood effects via a spatially impacted
algorithm [42]. The Markov model has gained popularity in recent years as a means of
fitting complex spatial patterns using practical rules in the urban development process [43].
Sustainability 2024, 16, 11130 6 of 19

Despite its simplicity, the Markov model has a significant restriction regarding the spatial
distribution of data related to land uses. Although each land use’s transition probability
matrix might be precise, it is unknown where the occurrences will occur in space [44].
Therefore, by incorporating a spatial dimension into the model through the application
of cellular automata (CA) filter, the CA-Markov chain model was created to improve the
accuracy of determining the change of different categories of land use. As a result, CA
possesses the potential to change its state based on its neighbors and prior state [45]. There-
fore, by first generating a dataset of land uses at a specific time and then projecting the
changing probabilities of this data for a future time, a hybrid CA-Markov model can be
used to detect at a dynamic degree of estimation the spatial and temporal changes of the
various land uses in the Qena Governorate [46]. Nonetheless, the process of detecting
LULC changes involves forecasting how the pixels in the satellite image would change
over time, switching from one land-use type at time (t1) to another at time (t2). For that
objective, the Equations (1)–(3) created by Ma et al. [45] were employed.
 
P11 P12 P1m
P = (pij) =  P21 P22 P2m  (1)
Pm1 Pm2 Pmm

m
∑ j=1 Pij = 1, i, j = 1, 2, . . . , m (2)

St+1 = Pij × St (3)

Here, St represents the current land-use status at time t, St+1 refers to the land-use
state at the next time point (t + 1), and Pij is the transitional probability of shifting from one
specific land-use type m to another. The value of Pij ranges between 0 and 1. CA-Markov
chain analysis was conducted using IDRISI 7.02 software to determine the transition and
probability matrices for specific land-use areas during the periods 1993–2008 and 2008–2024.

2.5. Land Use/Land Cover Change Forecasting

A cellular automaton (CA) with two variables was employed in the simulation pro-
cedure for the future land-use forecast. These variables are the local variable, which is
typically given interactions, and the discrete variable in time and space [47]. Cell, cell space,
rule, neighbor, and time make up this model. Additionally, this model can identify the
neighbors; the closer a cell is to its neighbors, the higher the weight factor. As a result,
the states of nearby grid cells may be forecasted using the weight factor and transition
probability [48]. Equation (4) was utilized to ascertain the change in land use.

LU(i,t2) − LUAi 1
Si = × × 100% (4)
LU(i,t1) t2 − t1

where Si is land-use change, LU(i,t1) is land-use change at an earlier time, LU(i,t2) is land-use
changed area at a later time, and LUAi is area with no change.
The CA-Markov model in the IDRISI 7.02 software was used to estimate the outcomes
of LULC alterations in 2040 and 2050. As stated by Tali et al. [49], the LULC images of the
Qena Governorate during the two time periods of 1993 as earlier land use (t1) and 2008 as
later land use (t2) were used in the current investigation.
Using suitable images, the standard contiguity filter (5 × 5 pixels) was applied to
determine the neighborhoods of each land-use class’s cells (Equation (6)). A 5 × 5 matrix
space was used to encompass each cell’s center in order to produce a good influence. This
spatial filter was applied in close proximity to the pre-existing category and excluded
arbitrary land-use changes [50]. Additionally, pixels closest to the current land-use category
are more suitable than those farther away [51]. In the current work, 2040 and 2050 LULC
maps were forecasted using spatially explicit, contiguous weights created from neighboring
Sustainability 2024, 16, 11130 7 of 19

pixels. The contiguity filter displayed below is the one utilized for the LULC change analysis
(Equation (5)) [3].
 
00100
0 1 1 1 0
 
Contiguity filter 5 × 5 = 
1 1 1 1 1
 (5)
0 1 1 1 0
00100
There are four primary stages to the Markov model’s approach for LULC analysis,
LULC changes, and forecast. These are as follows:
1. The Markov chains model was used to compute the transition probabilities matrices
for the years 1993, 2008, and 2024;
2. From 1993 to 2024, a set of conditional probability data for various land uses was
obtained by applying these transition matrices;
3. Using the CA-Markov spatial operator in IDRISI 7.02 software, which is based on
Markov chain analysis, the transition probabilities matrices of 1993, 2008, and 2024 as
well as conditional probability data were combined to simulate the land-use maps of
2040, 2024, and 2024 as a base map.
A new land-use map was created by overlaying all of the data from the previous steps.

2.6. The Validation Stage of CA-Markov Model Results

The CA- Markov model was validated using the validation module in the IDRISI
7.02 software. The projected outcomes were then contrasted with the observed dataset of
LULC changes between 1993 and 2008. This was evaluated by utilizing the kappa index
of agreement to estimate the level of agreement between the reference map from 1993
and the simulated map from 2008. The following Equations (6)–(9) represent the kappa
indices: kappa for location (k location), kappa for no information (k no), kappa for quantity
(k quantity), and kappa for standard (k standard) [52].

Kno = (M(m)N(n) /(P(p) − N(n)) (6)

Klocation = (M(m)N(m) /(P(m) − N(m)) (7)

Kquantity = (M(m)H(m) /(k(m) − H(m)) (8)

Kstandard = (M(m)N(n) /(P(p) − N(m)) (9)

The degree of conformity between the reference map and the unaltered comparison
map is denoted by M(m).The agreement between a map with a distribution of m among the
different categories in each cell and the reference map is denoted by N (m). The agreements
arising from chance are denoted by N (n). H (m) is the agreement between a modified
comparison map and the reference map. The reference map’s agreement with a modified
comparison map is denoted by (P (m); the reference map’s agreement with a map that
offers perfect information on both quantity and location is denoted by P (p). The agreement
between a modified comparison map and the reference map is expressed as K (m) [53].

2.7. Accuracy Assessment

Any classifier’s accuracy is influenced by the quantity of input data, the research
area, and the satellite sensor. Different LULC accuracies for various classifiers in various
study areas have been identified in a number of studies [54–56]. Accuracy assessment
can be used to examine the efficiency of any classifier following LULC classification.
The kappa coefficient and total accuracy are the most often used metrics for accuracy
evaluation. The kappa coefficient evaluates the ability of a classifier in perspective of
Sustainability 2024, 16, 11130 8 of 19

the anticipated values. The range of values for the kappa coefficient is 0 to 1, where 0
represents no ability, and 1 represents the classifier’s perfect ability. On the other hand,
overall accuracy can be defined as the ratio of correctly categorized pixels to all pixels
classified. An embedded confusion matrix technique is present in GEE. The GEE-embedded
confusion matrix approach was used in this study to validate and assess the classification
accuracy. The following Equations (10)–(13) can be used to calculate the kappa coefficient
and assess overall accuracy [57].

Cp
Overall Accuracy = ( ) × 100 (10)
Tp

where Cp indicates correctly classified pixels, and Tp indicates the total pixels.

N ∑icr=1 x jk − ∑icr=1 ( x j + − x + j
Kappa coefficient = (11)
N 2 − ∑ri=1 ( xj + × x + j)

Here, N represents the total number of observations; cr denotes the total count of rows
and columns; xjk is the number of observations found in row j and column k; xj+ + is the
marginal total for row j; and x + j is the marginal total for column j. Additionally, the user’s
accuracy for each class is calculated by dividing the total number of pixels classified in that
class by the number of correctly classified pixels in the class. The user’s accuracy formula
is given as follows:

propertly classi f ied pixels o f class

User′ s Accuracy = × 100 (12)
total classi f ied pixels o f class(the row total )

The potential for a pixel to exist in its corresponding projected class is the essence of
user accuracy. Accordingly, the ratio of accurately identified pixels to all reference pixels
in a class is used to determine the producer’s accuracy. The formula for determining
producer’s accuracy is as follows:

propertly classi f ied pixels o f class

Producer′ s Accuracy = (13)
total classi f ied pixels o f class(the column total )

2.8. Generation of Thematic Maps

The geospatial analysis wizard provided in the ArcGIS Pro 3.2 software was used to
create theme maps of several LULC classes for the studied periods (1993, 2008, and 2024)
and forecasted periods (2040 and 2050).

3. Results and Discussion

3.1. Accuracy Assessment
The accuracy assessment of analyzing the LULC change of different land-use classes
in Qena Governorate during both time periods (1993–2008) and (2008–2024) indicated that
forecasted and observed data are perfectly matched, and a high accuracy of estimation was
achieved. The kappa values for these variations range between 0 and 1 (0% and 100%); the
closer the value to 100%, the better the accuracy of the agreement [53]. The probability of
changing land use from a category for another was also calculated for each land-use class
in the studied area using Markov chain modeling of different Landsat satellite images for
different years.
The acquired results, as indicated in Table 2, demonstrate that for the various study
periods of 1993, 2008, and 2024, the classification accuracy satisfied the requirement that
the sensor data had at least 80% accuracy [54]. The LULC classification’s overall accuracy
varied between 97.29% and 99.33% (Table 2). These percentages indicate high precision,
according to Anderson et al. [55,56]; therefore, the LULC maps can be used reliably for
forecasting and studying LULC change dynamics.
 according to Anderson et al. [55,56]; therefore, the LULC maps can be used re
forecasting and studying LULC change dynamics.

Sustainability 2024, 16, 11130 Table 2. Evaluation of each classified image’s accuracy from 1993, 2008, and 92024.
of 19

Land Uses/Classes Landsat TM 1993 Landsat ETM+ 2008 Landsat OLI 202
PC 2.
Table (%) UC
Evaluation (%)classified
of each PCimage’s
(%) accuracy from
UC 1993,
(%) 2008, and 2024.
PC (%) UC (%
Urban areas 99.94 96.45 99.44 96.27 96.54 97.7
Landsat TM 1993 Landsat ETM+ 2008 Landsat OLI 2024
Land Uses/Classes
Water bodies 96.43 100 96.61 100 98.52 100
PC (%) UC (%) PC (%) UC (%) PC (%) UC (%)
Cultivated lands 100 99.95 99.81 99.93 99.89 99.3
Urban areas 99.94 96.45 99.44 96.27 96.54 97.74
Bare soils
Water bodies 100 96.43 99.96 100 99.18 96.61 99.74
100 98.52 90.92 100 85.0
Cultivated(%)
Overall accuracy lands 100
99.33 99.95 99.81 99.23 99.93 99.89 99.3897.29
Bare soils 100 99.96 99.18 99.74 90.92 85.06
KappaOverall
coefficient
accuracy (%)
0.99 99.33 0.98
99.23 97.29
0.96
Kappa coefficient 0.99 accuracy; UC = user’s
Note: PC = producer’s 0.98accuracy. 0.96
Note: PC = producer’s accuracy; UC = user’s accuracy.

Figures 3–5 display the classification maps that were created using t
Figures 3–5 display the classification maps that were created using the Qena Gover-
Governorate’s
norate’s historical historical images.
images. Different Different
LULC LULC
classes were classes were
categorized using categorized
using different and
and distinguishing
distinguishing colors
colors based based
on the on on
legends thethese
legends
maps. on these maps.

Sustainability 2025, 17, x FOR PEER REVIEW

Figure The
Figure3. 3. classified
The imageimage
classified of the study area
of the in 1993.
study area in 1993.

Figure4.4.The
Figure Theclassified imageimage
classified of the study
of thearea in 2008.
study area in 2008.
 Sustainability 2024, 16, 11130 10 of 19
Figure 4. The classified image of the study area in 2008.

Figure5. 5.
Figure TheThe classified
classified imageimage of thearea
of the study study area
in 2024. in 2024.
3.2. Change Analysis of LULC Maps
3.2. Many
Change Analysis
studies have of LULCLULC
utilized Mapschanges in the past periods and forecasted the
possible
Manychanges in future
studies situations
have usingLULC
utilized remote sensing
changes (RS)inandthegeographic information
past periods and foreca
systems (GIS) together with the CA-Markov model [3,58–61]. The LULC areas’ annual
possible
change changes
in different in future
classes situations
of the study area was using
calculatedremote
in square sensing (RS)andand ge
kilometers
information
percentage in thesystems
studied time(GIS) together
periods. withspread
Urban areas the CA-Markov modeland
across both cultivated [3,58–61].
bare Th
soils as population grew. Between 1993 and 2008, there was a net growth of 57.00 km2 in
areas’ annual change in different classes of the study area was calculated i
the urban areas (Figure 6). On the other hand, the net loss for the bare and cultivated soils
kilometers
was andkm
12.4 and 42.7 percentage in Actually,
2 , respectively. the studied
duringtime periods.
this period Urban areas
(1993–2008), spread acr
many bare
soils were exposed
cultivated and to the reclamation
bare process [34]. grew.
soils as population As a result, there were
Between 1993 more
andcultivated
2008, there w
areas and less bare soil. In terms of water bodies, certain islands entirely vanished within
growth of 57.00 km2 in the urban areas (Figure 6). On the other hand, the net lo
Sustainability 2025, 17, x FOR PEER REVIEW
the research region, while other islands sprouted up elsewhere. Depositional processes also
bare and
caused newcultivated soils was
islands and sandbars 12.4 as
to form and 42.7
well kmfloodplains
as new 2, respectively. Actually, during th
to arise on the convex
sides of the river flow. Therefore, from 1993 to 2008,
(1993–2008), many bare soils were exposed to the reclamation the Nile River lost 1.9 km2process
in total. [34]. As
there were more cultivated areas and less bare soil. In terms of water bodies
islands entirely vanished within the research region, while other islands spro
elsewhere. Depositional processes also caused new islands and sandbars to form
as new floodplains to arise on the convex sides of the river flow. Therefore, from
2008, the Nile River lost 1.9 km2 in total.

Figure
Figure 6. The changes
6. The in km2 in
changes of each
km2LULC category
of each LULCfor 1993–2008.
category for 1993–2008.

In the research region, the area covered by water bodies increased b

km2 over the second period (2008–2024). This could be as a result of nu
 being built and renovated. Significant advancements have been made in th
the Decent Life program, which was started in 2019 with the goal of pr
development, according to the Ministry of Irrigation and Water Resourc
Sustainability 2024, 16, 11130
ministry’s announcement on 7 May 2024, a substantial 3230 km 11 ofofcanals
19 ha
rehabilitation. It is important to note that the efforts to line the canal are bei
in 18 governorates,
In the research region,withthe areaspecific
covered by attention
water bodies paid to areas
increased like 24.4
by almost km2 Luxo
Menya,
over the second period (2008–2024). This could be as a result of numerous canals be-
Asyut, Qena, Beni Suef, Sohag, and Beheira. By ensuring fair access to up
ing built and renovated. Significant advancements have been made in the first phase
management
of infrastructure,
the Decent Life program, which wasthis planned
started in 2019 with distribution promotes
the goal of promoting ruralrural p
development, according to the Ministry of Irrigation and Water Resources. As per the
crop growth. Reclamation efforts have accelerated during this time, but
ministry’s announcement on 7 May 2024, a substantial 3230 km of canals have undergone
Qena Governorate
rehabilitation. is one
It is important to noteofthat
thethe most heavily
efforts to line thepopulated governorates
canal are being undertaken [62]
in 18 governorates,
CAPMAS [62],withthespecific attention of
population paidthe
to areas
Qena like Governorate
Menya, Luxor, Menoufeya,
increased to
Asyut, Qena, Beni Suef, Sohag, and Beheira. By ensuring fair access to upgraded water
3,224,573 in
management 2023, with
infrastructure, this80.3%
plannedof the population
distribution promotes ruralresiding in and
prosperity ruralcropregions
growth.
urban Reclamation efforts have accelerated
areas. Consequently, the netduring this time,
loss for but unfortunately,
cultivated and bareQena soils was
Governorate is one of the most heavily populated governorates [62]. According to CAP-
km2,[62],
MAS respectively.
the population of One of the
the Qena primary
Governorate issues
increased endangering
to approximately the study
3,224,573
in 2023, with 80.3%
extremely fertileof terrain
the population
is urbanresiding in ruralThe
sprawl. regions
mostandproductive
19.7% in urban areas.
land in the r
Consequently, the net loss for cultivated and bare soils was 22.2 and 26.4 km2 , respectively.
hasofbeen
One significantly
the primary lost to urban
issues endangering the studydevelopments (Figure
area’s scarce, extremely 7).terrain
fertile Built-up
is are
expanding
urban sprawl. The overmost agricultural
productive land soils.
in theThe area
research added
region to urban
has been settlements
significantly lost w
to urban developments (Figure 7). Built-up areas are rapidly expanding over agricultural
km2. The variations in LULC classes in the study area between 2008 and 2
soils. The area added to urban settlements was about 24.2 km2 . The variations in LULC
in (Figure
classes 7). area between 2008 and 2024 are shown in (Figure 7).
in the study

Figure
Figure 7. The changes
7. The in km2 in
changes of each
km2LULC category
of each LULCfor 2008–2024.
category for 2008–2024.
3.3. Markov Chain Model Analysis
The acquired data, which are displayed in Tables 3 and 4, illustrate how the transition
probabilities changed over time from 1993 to 2024. Gains are obtained by deducting the
persistence (diagonal entries) from each group’s total column, while losses are obtained
by deducting the persistence from each group’s total row [63]. For example, from 1993 to
2008, there was a 52.28% chance that urban regions would remain urban areas; in contrast,
there was a 22.44% probability that cultivated soils would eventually become urban areas,
and so on for other LULC categories. In 2008, there was a 42.54% and a 76.42% probability,
respectively, that the bare soils and water bodies would stay the same.
Sustainability 2024, 16, 11130 12 of 19

Table 3. Transition probability matrix based on Qena’s 1993–2008 land-use maps.

Probability of Changing from 1993 to 2008 Subtotals

Land Uses/Classes
Urban Areas Water Bodies Cultivated Soils Bare Soils Total Loss
Urban areas 0.5228 0.0550 0.413 0.0092 1 0.4772
Water bodies 0.0469 0.7642 0.1889 0.0000 1 0.2358
Cultivated soils 0.2244 0.0396 0.7352 0.0008 1 0.2648
Bare soils 0.4050 0.0014 0.1682 0.4254 1 0.5746
Total 1.1991 0.8602 1.5053 0.4354 4
Gain 0.6763 0.0960 0.7701 0.0100

Table 4. Transition probability matrix based on Qena’s land-use maps from 2008 to 2024.

Probability of Changing from 2008 to 2024 Subtotals

Land Uses/Classes
Urban Areas Water Bodies Cultivated Soils Bare Soils Total Loss
Urban areas 0.839 0.0000 0.161 0.0000 1 0.161
Water bodies 0.1578 0.838 0.0042 0.0000 1 0.162
Cultivated soils 0.1448 0.0097 0.8454 0.0001 1 0.1546
Bare soils 0.1716 0.0000 0.0000 0.8284 1 0.1716
Total 1.3132 0.8477 1.0106 0.8285 4
Gain 0.4742 0.0097 0.1652 0.0001

3.4. Land-Use Change Dynamics

Table 5 provides more insight into the LULC class changes of the Qena Governorate
area by calculating the annual change for the two analyzed time periods (1993–2008 and
2008–2024). It is clear from the data that were acquired that, for the two time periods, the
cultivated lands declined yearly by 2.85 km2 and 1.48 km2 , respectively. The second period
in which the reclamation processes of bare soils were observed had the lowest annual
reduction. As a result, during the two time periods, the amount of exposed soils decreased
by 0.83 km2 and 1.76 km2 , respectively, annually. Over the two investigated periods, the
annual growth in urban areas was 3.80 km2 and 1.61 km2 , respectively. In terms of the water
bodies, they shrank by 0.13 km2 annually between 1993 and 2008 but slowly increased by
1.63 km2 between 2008 and 2024 as a consequence of the study area’s canals and waterways
being rehabilitated and lined.

Table 5. LULC classes in the Qena Governorate area change annually.

Average Annual Changes km2

Land Uses/Classes
1993 to 2008 2008 to 2024
Urban areas 3.80 1.61
Water bodies −0.13 1.63
Cultivated lands −2.85 −1.48
Bare soils −0.83 −1.76

The results of the LULC change in the various LULC classes in the Qena Governorate
across the time periods under study are shown in (Table 6). The temporal distribution of
the LULC change in Qena Governorate classes across time, expressed in square kilometers,
is shown in (Figure 8). Between 1993 and 2024, urban areas—which include houses,
roadways, industrial regions, and other human activities—grew quickly. With population
expansion and the broader evolution of the construction of roads for industry and new
urban settlements, urban areas have grown significantly, from roughly 200.20 km2 (16.44%)
in 1993 to roughly 257.20 km2 (21.12%) in 2008 and 281.40 km2 (23.11%) in 2024. In addition,
the urban areas increased clearly [34] (Table 6). People migrated to the low-lying desert
 With population expansion and the broader evolution of the construction of roads fo
industry and new urban settlements, urban areas have grown significantly, from roughly
200.20 km2 (16.44%) in 1993 to roughly 257.20 km2 (21.12%) in 2008 and 281.40 km
(23.11%) in 2024. In addition, the urban areas increased clearly [34] (Table 6). Peopl
Sustainability 2024, 16, 11130 13 of 19
migrated to the low-lying desert fringe (planned reclaimed lands) areas that wer
overrun for the limestone plateau from the restricted Nile Valley. The area covered by
water(planned
fringe bodies in 1993 was
reclaimed roughly
lands) 66.10
areas that km2
were with 5.43%,
overrun graduallyplateau
for the limestone declining
fromto the64.20 km
(i.e., 5.27%
restricted NileofValley.
the entire study
The area area)byinwater
covered 2008bodies
and growing to around
in 1993 was 88.60km2
roughly 66.10 kmwith
2 in 2024 as a
5.43%, gradually declining to 64.20 km 2 (i.e., 5.27% of the entire study area) in 2008 and
result of the canals and waterways in the study area being restored and lined. Over the
growing to around 88.60 km2 in 2024 as a result of the canals and waterways in the study
past thirty-one years, the research area’s cultivated land has shrunk. Between 1993 and
area being restored and lined. Over the past thirty-one years, the research area’s cultivated
2024,has
land it underwent
shrunk. Betweena minor
1993alteration,
and 2024, it going from 865.30
underwent a minortoalteration,
800.40 km 2, respectively. In
going from
1993, 86.30 km
865.30 to 800.40 km , respectively. In 1993, 86.30 km (7.09%) of the Wadi El Nile area soils,
2 2(7.09%) of the Wadi El Nile 2 area was covered by bare was which
covered
includesbythebare soils, which
highlands includes
that thethe
encircle highlands
lowlandsthat[34].
encircle
Thistheclass
lowlands [34]. to
dropped This73.87 km
class dropped to 73.87 km 2 (6.07%) in 2008 and 47.46 (3.90%) in 2024 because of recently
(6.07%) in 2008 and 47.46 (3.90%) in 2024 because of recently proposed land reclamation
proposed land reclamation [34] (Table 6).
[34] (Table 6).
Table 6. Land use/land cover distribution across time, expressed in km2 by year.
Table 6. Land use/land cover distribution across time, expressed in km2 by year.
1993 2008 Changed Areas 2024 Changed Areas
1993
Areas 2008
Areas Changed
1993–2008 AreasAreas 2024 2008–2024
Changed Areas
Land Uses/Classes
Land Uses/Classes Areas Areas 1993–2008 Areas 2008–2024
km2 % km2 % km2 km2 % km2
km 2 % km 2 % km 2 km 2 % km2
Urban areas 200.20 16.44 257.20 21.12 57.0
Urban areas 200.20 16.44 257.20 21.12 57.0 281.40 23.11
281.40 23.11 24.2
24.2
Water bodies 66.10 5.43 64.20 5.27 −1.9 88.60 7.27 24.4
Water bodies 66.10 5.43 64.20 5.27 −1.9 88.60 7.27 24.4
Cultivated soils 865.30 71.05 822.60 67.54 −42.7 800.40 65.72 −22.2
Cultivated soils 865.30 71.05 822.60 67.54 −42.7 800.40 65.72 −22.2
Bare soils 86.30 7.09 73.87 6.07 −12.4 47.46 3.90 −26.4
Bare soils 86.30 7.09 73.87 6.07 −12.4 47.46 3.90 −26.4

2 between
Figure8.8.LULC
Figure LULC change
change distribution
distribution overover
time time
in kmin km2 between 1993,and
1993, 2008, 2008, and 2024.
2024.

3.5. LULC Changes Prediction of 2040 and 2050

3.5. LULC Changes Prediction of 2040 and 2050
The CA-Markov model’s forecast for LULC change across all land-use classes in the
Qena The CA-Markov
Governorate were model’s forecast
verified using datafor
fromLULC
2008 change
and 2024.across all9land-use
Figures classes in th
and 10 present
the thematic maps of the spatial distribution of the anticipated LULC
Qena Governorate were verified using data from 2008 and 2024. Figures 9 and changes in 2040
10 presen
and 2050. Table 7 displays the acquired statistics on the anticipated LULC changes in the
the thematic maps of the spatial distribution of the anticipated LULC changes in 2040 and
various classes of the Qena Governorate area throughout the two forecasted time periods
2050. Table
(2040 and 7 displays
2050). the acquired
In 2040, urban statistics
areas will grow on km
by 44.90 the2 ,anticipated
and by 2050, LULC changes
they will grow in th
various
to 331.60 classes of the Qena
km2 (27.23%). Governorate
This anticipated risearea throughout
in urban regions the
maytwo forecastedtotime
be connected the period
current population-growth-driven expansion of built-up areas [64–67] (Table 7). Moreover,
in 2050, the proportion of water bodies in the study region will rise steadily to 27.90% of its
total area. Cultivated lands will drop by 47 km2 in 2040 and then reach 731.30 km2 in 2050
(Table 7). Urban development will ultimately result in fewer bare soils [3].
 current population-growth-driven expansion of built-up areas [64–67] (Table 7)
Moreover, in 2050, the proportion of water bodies in the study region will rise steadily to
27.90% of its total area. Cultivated lands will drop by 47 km2 in 2040 and then reach
Sustainability 2024, 16, 11130 731.30 km2 in 2050 (Table 7). Urban development will ultimately result in fewer bare soil
14 of 19
[3].

Figure 9. Spatial distribution map of forecasted LULC changes during 2040.

Table 7. Changes in km2 that are anticipated for LULC classes in 2040 and 205

Changed Areas
Land Uses/Classes Projected Area Projected Area
(km2)
2040 % 2024 to 2040 2050 %
Urban areas 326.3 26.79 44.9 331.60 27.23
Water bodies 107.4 8.82 18.8 135.30 11.11
Cultivated lands 753.4 61.86 −47.0 731.30 60.05
Bare soils 30.8 2.53 −16.7 19.70 1.62
Total 1217.9Spatialdistribution
Figure9.9.Spatial
Figure
100.00
distribution
mapmap of forecasted
of forecasted LULCLULC changes
changes
1217.90
during during
2040. 2040.
100.00

Table 7. Changes in km2 that are anticipated for LULC classes in 2040 and 2050.

Changed Areas Changed Areas

Land Uses/Classes Projected Area Projected Area
(km2) (km2)
2040 % 2024 to 2040 2050 % 2040 to 2050
Urban areas 326.3 26.79 44.9 331.60 27.23 5.32
Water bodies 107.4 8.82 18.8 135.30 11.11 27.90
Cultivated lands 753.4 61.86 −47.0 731.30 60.05 −22.10
Bare soils 30.8 2.53 −16.7 19.70 1.62 −11.08
Total 1217.9 100.00 1217.90 100.00

Figure 10. Spatial distribution map of forecasted LULC changes during 2050.
Figure 10. Spatial distribution map of forecasted LULC changes during 2050.
Table 7. Changes in km2 that are anticipated for LULC classes in 2040 and 2050.
3.6. Model Validation
Projected Area Changed Areas (km2 ) Projected Area Changed Areas (km2 )
Land Uses/Classes
2040 The% IDRISI software’s
2024 to 2040 validate
2050 tool was
% used to generate
2040 to 2050 the kap
Urban areas 326.3 26.79 k no, k location,
standard, 44.9 and k331.60
location27.23 5.32 confirm the
layer to further
Water bodies 107.4 8.82 18.8 135.30 11.11 27.90
Cultivated lands Figure 10.61.86
753.4 Spatial distribution map of forecasted
−47.0 LULC changes
731.30 60.05 during 2050.
−22.10
Bare soils 30.8 2.53 −16.7 19.70 1.62 −11.08
Total 3.6. Model
1217.9 Validation
100.00 1217.90 100.00
The IDRISI software’s validate tool was used to generate the kappa statistics of the k
standard, k no, k location, and k location layer to further confirm the tests. Kappa can b
Sustainability 2024, 16, 11130 15 of 19
Sustainability 2025, 17, x FOR PEER REVIEW 15 of 19

3.6. Model Validation

usedThe IDRISI software’s
to describe varying levelsvalidate tool was used
of consistency. to be
It can generate the into
classified kappafivestatistics
groups:of the
slight
k(0.0–0.20),
standard, k no, k location, and k location layer to further confirm the
fair (0.21–0.40), moderate (0.41–0.60), substantial (0.61–0.80), and almosttests. Kappa can be
used to describe varying levels of consistency. It can be classified into five groups: slight
perfect (0.81–1). Based on the comparison between the categorized LULC map for 2008
(0.0–0.20), fair (0.21–0.40), moderate (0.41–0.60), substantial (0.61–0.80), and almost perfect
and 2024 and the forecasted LULC for 2008 and 2024, the k coefficient for the quantity
(0.81–1). Based on the comparison between the categorized LULC map for 2008 and 2024
and the
and location was determined
forecasted LULC for 2008 for the
andpurpose
2024, theofk validating
coefficient the model.
for the The kand
quantity standard
location is
0.9321,
was k no is 0.9640,
determined for thek purpose
location is of0.9956, and the
validating k location
model.layer
The kisstandard
0.9956, according
is 0.9321,to k the
no
statistics of k indicators for 2008. Comparably, the 2024 k indicator statistics
is 0.9640, k location is 0.9956, and k location layer is 0.9956, according to the statistics of indicate that
kthe
indicators
k standard for is
2008. Comparably,
0.9246, the k no isthe 2024 k
0.9599, indicator
the statistics
k location indicate
is 0.9504, thatkthe
and the k standard
location layer
is
is 0.9246,
equal tothe k no(Table
0.9504 is 0.9599, thethe
8). As k location
outcomes, is the
0.9504, and the k location
quantification layer is mistakes
and localization equal to
0.9504 (Table 8).reduced.
were greatly As the outcomes,
Both precisethe quantification
location andand localization
quantity mistakesare
specification wereperfectly
greatly
reduced. Both precise location and quantity specification are perfectly attainable with the
attainable with the simulated model; in addition, the projected and real maps had small
simulated model; in addition, the projected and real maps had small quantification and
quantification and location errors [47,68,69]. Based on the good agreement between the
location errors [47,68,69]. Based on the good agreement between the anticipated and real
LULC map, and
anticipated Tablereal LULC
8 and map,
Figure 11Table
verify8that
and the
Figure 11 verify
accuracy that theofaccuracy
evaluation the bothevaluation
classified
of theofboth
data 2008classified
and 2024data of 2008and
is suitable andacceptable
2024 is suitable
for anyand acceptableThis
application. for any
showsapplication.
that the
This shows
number and that the number
location of the LULCand location
changes of canthe
beLULC
exactlychanges
specified can
bybe exactly
using specified by
the CA-Markov
using the
model CA-Markov
[5,68,70,71]. Thesemodel [5,68,70,71].
results completelyThese
concur results
with completely concur with those of
those of [3,60,72].
[3,60,72].
Table 8. K indicators for 2008 and 2024.
Table 8. K indicators for 2008 and 2024.
K Indicators 2008 2024
K Indicators
k standard 0.9321 2008 2024
0.9246
kk
nostandard 0.9640 0.9321 0.9246
0.9599
k no
k location 0.9956
0.9640 0.9599
0.9504
k location 0.9956 0.9504
k location layer 0.9956 0.9504
k location layer 0.9956 0.9504

Figure11.
Figure 11. The
The comparison
comparison between
between simulated
simulatedand
andreal
realin km22 of
inkm of 2008
2008 and
and 2024.
2024.

3.7.
3.7. Limitations
Limitations and
and Prospects
Prospects
Since the training dataset was gathered for a particular geographic location, altering the
Since the training dataset was gathered for a particular geographic location, altering
study area may have an impact on the model’s performance, even though a comprehensive
the study
training areawith
dataset mayadequate
have an impact(covering
diversity on the model’s performance,
all land types) even though
and encompassing differ-a
comprehensive training dataset with adequate diversity (covering all land
ent pixel-level and object-level features for training classifiers was collected. However, types) andif
encompassing
training samplesdifferent pixel-level
are provided andnew
from the object-level features
target regions, thefor training can
framework classifiers was
be success-
collected.
fully appliedHowever, if training
in additional studysamples are provided
areas. There are a numberfromofthe new target
difficulties whenregions,
workingthe
with satellitecan
framework imagery, such as the
be successfully existence
applied of biasesstudy
in additional and insufficient
areas. Thereinformation.
are a number Forof
difficulties when working with satellite imagery, such as the existence of biases and
insufficient information. For example, in the target study area, cloud coverage may
impede the use of important data for LULC mapping. Furthermore, depending on just
Sustainability 2024, 16, 11130 16 of 19

example, in the target study area, cloud coverage may impede the use of important data for
LULC mapping. Furthermore, depending on just one distant sensing source may present
difficulties due to complex classes within the LULC system. In order to overcome the
limits posed by insufficient satellite imagery and intricate class distinctions, it is advised
that additional data sources be incorporated and integrated in order to greatly increase
the quality and dependability of LULC mapping. This study ran into a number of issues
when utilizing GEE for classification despite its strong points, which include a broad range
of features for spatial data processing, analysis, and interpretation. These restrictions
include the platform’s integration of different features, data processing volume, and com-
puting time constraints. To guarantee the completeness of the training data, the model
training procedure necessitates a large number of sample points when the study region is
large. Using deep learning techniques and comparing their results is advised to generate
dynamic LULC maps.

4. Conclusions
Using the classification and regression tree (CART) method on the Google Earth Engine
(GEE) platform in the Qena Governorate, Egypt, this study concentrated on extracting
LULC data from 1993 to 2024. Additionally, the CA-Markov model was utilized to predict
future changes in LULC for the research area in 2040 and 2050. Urban areas grew by
57 km2 between 1993 and 2008, while cultivated and bare soils shrank by 42.7 and 12.4 km2 ,
respectively. There was also a net loss of 1.9 km2 in the Nile River. Water bodies expanded
by 24.4 km2 , urban areas expanded by 24.2 km2 , and bare and farmed soils decreased by
26.2 and 26.4 km2 , respectively, between 2008 and 2024. Thematic maps of CA-Markov
model showed the spatial distribution of anticipated expected increases in LULC for 2040
and 2050. According to the findings, there will be an increase in aquatic bodies, agricultural
land, and urban regions. But, as expected, during the research years, the extent of bare
lands decreased.
Planning decisions increasingly demand an awareness of the spatiotemporal dynamics
of land use/land cover changes (LULCCs) as well as their current quantification and
forecasting. Conducting a thorough analysis on LULC using the GEE platform and the
CA-Markov chain model is an effective way for tracking and assessing LULC in the Qena
Governorate over three distinct time periods, i.e.,1993, 2008, and 2024. Furthermore, the
model can be beneficial for forecasting the LULC for 2040 and 2050. Land-cover dynamics
have a substantial impact on land management, urban planning, risk assessment, ecosystem
services assessment, and disaster response, emphasizing the importance of monitoring
these dynamics. Across all model stages, an accuracy of over 90% was attained. Thus,
in this investigation, CA-Markov modeling yielded results that are both dependable and
accurate, which is promising. Additionally, this study demonstrated that using GEE and
the CART model to analyze remote sensing data produced insightful information about
patterns and changes in land cover.

Author Contributions: Conceptualization, E.A.A., A.-r.A.M., H.F.M., M.S.S. and J.E.M.d.L.; Method-
ology, E.A.A., A.-r.A.M., M.S.S. and J.E.M.d.L.; Software, E.A.A., A.-r.A.M., M.S.S. and J.E.M.d.L.;
Validation, E.A.A., A.-r.A.M. and M.S.S.; Formal analysis, E.A.A., A.-r.A.M. and M.S.S.; Investiga-
tion, E.A.A. and A.-r.A.M.; Resources, A.-r.A.M.; Data curation, A.-r.A.M.; Writing—original draft,
A.-r.A.M., A.S.E.-S., S.A.A., M.S.S., A.I.E.-D. and J.E.M.d.L.; Writing—review & editing, H.F.M., M.S.S.
and A.I.E.-D.; Visualization, A.-r.A.M.; Supervision, A.S.E.-S., H.F.M., S.A.A., A.I.E.-D. and J.E.M.d.L.;
Project administration, A.S.E.-S., S.A.A. and J.E.M.d.L. All authors have read and agreed to the
published version of the manuscript.
Funding: Researchers Supporting Project number (RSPD2024R1044), King Saud University, Riyadh,
Saudi Arabia.
Institutional Review Board Statement: Not applicable.
Informed Consent Statement: Not applicable.
Sustainability 2024, 16, 11130 17 of 19

Data Availability Statement: Data are contained within the article.

Acknowledgments: The authors thank Researchers Supporting Project number (RSPD2024R1044),
King Saud University, Riyadh, Saudi Arabia.
Conflicts of Interest: The authors declare no conflicts of interest.

References
1. Biro, K.; Pradhan, B.; Buchroithner, M.; Makeschin, F. Land use/land cover change analysis and its impact on soil properties in
the northern part of Gadarif region, Sudan. Land Degrad. Dev. 2013, 24, 90–102. [CrossRef]
2. El Behairy, R.A.; El Arwash, H.M.; El Baroudy, A.A.; Ibrahim, M.M.; Mohamed, E.S.; Rebouh, N.Y.; Shokr, M.S. Artificial
intelligence integrated GIS for land suitability assessment of wheat crop growth in arid zones to sustain food security. Agronomy
2023, 13, 1281. [CrossRef]
3. Selmy, S.A.; Kucher, D.E.; Mozgeris, G.; Moursy, A.R.; Jimenez-Ballesta, R.; Kucher, O.D.; Fadl, M.E.; Mustafa, A.-r.A. Detecting,
analyzing, and predicting land use/land cover (LULC) changes in arid regions using landsat images, CA-Markov hybrid model,
and GIS techniques. Remote Sens. 2023, 15, 5522. [CrossRef]
4. Guan, D.; Li, H.; Inohae, T.; Su, W.; Nagaie, T.; Hokao, K.J. Modeling urban land use change by the integration of cellular
automaton and Markov model. Ecol. Model. 2011, 222, 3761–3772. [CrossRef]
5. Halmy, M.W.A.; Gessler, P.E.; Hicke, J.A.; Salem, B.B. Land use/land cover change detection and prediction in the north-western
coastal desert of Egypt using Markov-CA. Appl. Geogr. 2015, 63, 101–112. [CrossRef]
6. Zheng, H.W.; Shen, G.Q.; Wang, H.; Hong, J.J. Simulating land use change in urban renewal areas: A case study in Hong Kong.
Habitat Int. 2015, 46, 23–34. [CrossRef]
7. Telo da Gama, J. The role of soils in sustainability, climate change, and ecosystem services: Challenges and opportunities. Ecologies
2023, 4, 552–567. [CrossRef]
8. Chen, C.; Yang, X.; Jiang, S.; Liu, Z. Mapping and spatiotemporal dynamics of land-use and land-cover change based on the
Google Earth Engine cloud platform from Landsat imagery: A case study of Zhoushan Island, China. Heliyon 2023, 9, e19654.
[CrossRef]
9. Wang, X.; Liu, G.; Xiang, A.; Xiao, S.; Lin, D.; Lin, Y.; Lu, Y. Terrain gradient response of landscape ecological environment to land
use and land cover change in the hilly watershed in South China. Ecol. Indic. 2023, 146, 109797. [CrossRef]
10. Jones, K.R.; Venter, O.; Fuller, R.A.; Allan, J.R.; Maxwell, S.L.; Negret, P.J.; Watson, J.E.M. One-third of global protected land is
under intense human pressure. Science 2018, 360, 788–791. [CrossRef] [PubMed]
11. Zhou, J.; Wang, L.; Zhong, X.; Yao, T.; Qi, J.; Wang, Y.; Xue, Y. Quantifying the major drivers for the expanding lakes in the interior
Tibetan Plateau. Sci. Bull. 2022, 67, 474–478. [CrossRef]
12. Zhang, P.; Liu, L.; Yang, L.; Zhao, J.; Li, Y.; Qi, Y.; Ma, X.; Cao, L. Exploring the response of ecosystem service value to land use
changes under multiple scenarios coupling a mixed-cell cellular automata model and system dynamics model in Xi’an, China.
Ecol. Indic. 2023, 147, 110009. [CrossRef]
13. Pande, C.B.; Diwate, P.; Orimoloye, I.R.; Sidek, L.M.; Pratap Mishra, A.; Moharir, K.N.; Pal, S.C.; Alshehri, F.; Tolche, A.D. Impact
of land use/land cover changes on evapotranspiration and model accuracy using Google Earth engine and classification and
regression tree modeling. Geomat. Nat. Hazards Risk 2024, 15, 2290350. [CrossRef]
14. Zabihi, M.; Moradi, H.; Gholamalifard, M.; Khaledi Darvishan, A.; Fürst, C. Landscape management through change processes
monitoring in Iran. Sustainability 2020, 12, 1753. [CrossRef]
15. Meng, Z.; Dong, J.; Ellis, E.C.; Metternicht, G.; Qin, Y.; Song, X.-P.; Löfqvist, S.; Garrett, R.D.; Jia, X.; Xiao, X. Post-2020 biodiversity
framework challenged by cropland expansion in protected areas. Nat. Sustain. 2023, 6, 758–768. [CrossRef]
16. Debnath, J.; Sahariah, D.; Lahon, D.; Nath, N.; Chand, K.; Meraj, G.; Kumar, P.; Singh, S.K.; Kanga, S.; Farooq, M. Assessing the
impacts of current and future changes of the planforms of river Brahmaputra on its land use-land cover. Geosci. Front. 2023,
14, 101557. [CrossRef]
17. Lee, J.-G.; Kang, M.J.B.D.R. Geospatial big data: Challenges and opportunities. Big Data Res. 2015, 2, 74–81. [CrossRef]
18. You, N.; Dong, J.; Huang, J.; Du, G.; Zhang, G.; He, Y.; Yang, T.; Di, Y.; Xiao, X. The 10-m crop type maps in Northeast China
during 2017–2019. Sci. Data 2021, 8, 41. [CrossRef] [PubMed]
19. Li, W.; Wang, C.; Liu, H.; Wang, W.; Sun, R.; Li, M.; Shi, Y.; Zhu, D.; Du, W.; Ma, L.; et al. Fine root biomass and morphology in
a temperate forest are influenced more by canopy water addition than by canopy nitrogen addition. Ecol. Indic. 2023, 11, 1132248.
[CrossRef]
20. Jin, Y.; Liu, X.; Chen, Y.; Liang, X. Land-cover mapping using Random Forest classification and incorporating NDVI time-series
and texture: A case study of central Shandong. Int. J. Remote Sens. 2018, 39, 8703–8723. [CrossRef]
21. Ma, L.; Liu, Y.; Zhang, X.; Ye, Y.; Yin, G.; Johnson, B. Deep learning in remote sensing applications: A meta-analysis and review.
ISPRS J. Photogramm. Remote Sens. 2019, 152, 166–177. [CrossRef]
22. Kaur, H.; Tyagi, S.; Mehta, M.; Singh, D. Time series (2001/2002–2021) analysis of Earth observation data using Google Earth
Engine (GEE) for detecting changes in land use land cover (LULC) with specific reference to forest cover in East Godavari Region,
Andhra Pradesh, India. J. Earth Syst. Sci. 2023, 132, 86. [CrossRef]
Sustainability 2024, 16, 11130 18 of 19

23. Tamiminia, H.; Salehi, B.; Mahdianpari, M.; Quackenbush, L.; Adeli, S.; Brisco, B. Google Earth Engine for geo-big data
applications: A meta-analysis and systematic review. ISPRS J. Photogramm. Remote Sens. 2020, 164, 152–170. [CrossRef]
24. Gorelick, N.; Hancher, M.; Dixon, M.; Ilyushchenko, S.; Thau, D.; Moore, R. Google Earth Engine: Planetary-scale geospatial
analysis for everyone. Remote Sens. Environ. 2017, 202, 18–27. [CrossRef]
25. Xu, Z.; Li, X.; Li, J.; Xue, Y.; Jiang, S.; Liu, L.; Luo, Q.; Wu, K.; Zhang, N.; Feng, Y.; et al. Characteristics of source rocks and
genetic origins of natural gas in deep formations, Gudian Depression, Songliao Basin, NE China. ACS Earth Space Chem. 2022,
6, 1750–1771. [CrossRef]
26. Amani, M.; Ghorbanian, A.; Ahmadi, S.A.; Kakooei, M.; Moghimi, A.; Mirmazloumi, S.M.; Moghaddam, S.H.A.; Mahdavi, S.;
Ghahremanloo, M.; Parsian, S.; et al. Google earth engine cloud computing platform for remote sensing big data applications:
A comprehensive review. IEEE J. Sel. Top. Appl. Earth Obs. Remote Sens. 2020, 13, 5326–5350. [CrossRef]
27. Parashar, D.; Kumar, A.; Palni, S.; Pandey, A.; Singh, A.; Singh, A.P. Use of machine learning-based classification algorithms in the
monitoring of Land Use and Land Cover practices in a hilly terrain. Environ. Monit. Assess. 2024, 196, 8. [CrossRef] [PubMed]
28. Pande, C.B.; Srivastava, A.; Moharir, K.N.; Radwan, N.; Mohd Sidek, L.; Alshehri, F.; Pal, S.C.; Tolche, A.D.; Zhran, M.
Characterizing land use/land cover change dynamics by an enhanced random forest machine learning model: A Google Earth
Engine implementation. Environ. Sci. Eur. 2024, 36, 84. [CrossRef]
29. Patil, A.; Panhalkar, S. A comparative analysis of machine learning algorithms for land use and land cover classification using
google earth engine platform. J. Geomat. 2023, 17, 226–233. [CrossRef]
30. Yuh, Y.G.; Tracz, W.; Matthews, H.D.; Turner, S.E. Application of machine learning approaches for land cover monitoring in
northern Cameroon. Ecol. Inform. 2023, 74, 101955. [CrossRef]
31. Srivastava, A.; Chinnasamy, P. Water management using traditional tank cascade systems: A case study of semi-arid region of
Southern India. SN Appl. Sci. 2021, 3, 281. [CrossRef]
32. Rahman, K.U.; Ejaz, N.; Shang, S.; Balkhair, K.S.; Alghamdi, K.M.; Zaman, K.; Khan, M.A.; Hussain, A. A robust integrated
agricultural drought index under climate and land use variations at the local scale in Pakistan. Agric. Water Manag. 2024,
295, 108748. [CrossRef]
33. Khwarahm, N.R.; Qader, S.; Ararat, K.; Fadhil Al-Quraishi, A.M. Predicting and mapping land cover/land use changes in
Erbil/Iraq using CA-Markov synergy model. Earth Sci. Inform. 2021, 14, 393–406. [CrossRef]
34. Kamel, M. Monitoring of land use and land cover change detection using multi-temporal remote sensing and time series analysis
of qena-luxor governorates (QLGs), Egypt. J. Indian Soc. Remote Sens. 2020, 48, 1767–1785. [CrossRef]
35. Soliman, M.; Gendy, O.; Abdellatif, A.D. Land Resources Evaluation of some Soils in the Western Qena Governorate Using
Remote Sensing and Gis. J. Soil Sci. Agric. Eng. 2017, 8, 837–846. [CrossRef]
36. CLAC. Central Laboratory for Agriculture Climate, Ministry of Agriculture Climatic Elements from Luxor Station (2010–2015); CLAC:
Cambridge, ON, Canada, 2015.
37. Kumar, L.; Mutanga, O. Google Earth Engine applications since inception: Usage, trends, and potential. Remote Sens. 2018,
10, 1509. [CrossRef]
38. Breiman, L. Classification and Regression Trees; Routledge: Abingdon, UK, 2017.
39. Bhargava, N.; Dayma, S.; Kumar, A.; Singh, P. An approach for classification using simple CART algorithm in WEKA. In
Proceedings of the 2017 11th International Conference on Intelligent Systems and Control (ISCO), Coimbatore, India, 5–6 January
2017; pp. 212–216.
40. Lian, W.; Nie, G.; Jia, B.; Shi, D.; Fan, Q.; Liang, Y. An Intrusion Detection Method Based on Decision Tree-Recursive Feature
Elimination in Ensemble Learning. Math. Probl. Eng. 2020, 2020, 2835023. [CrossRef]
41. Tangirala, S. Evaluating the impact of GINI index and information gain on classification using decision tree classifier algorithm.
Int. J. Adv. Comput. Sci. Appl. 2020, 11, 612–619. [CrossRef]
42. Clarke, K.C.; Hoppen, S.; Gaydos, L.J. A self-modifying cellular automaton model of historical urbanization in the San Francisco
Bay area. Environ. Plan. B Plan. Des. 1997, 24, 247–261. [CrossRef]
43. Sang, L.; Zhang, C.; Yang, J.; Zhu, D.; Yun, W. Simulation of land use spatial pattern of towns and villages based on CA–Markov
model. Math. Comput. Model. 2011, 54, 938–943. [CrossRef]
44. Nath, B.; Wang, Z.; Ge, Y.; Islam, K.; Singh, R.P.; Niu, Z. Land use and land cover change modeling and future potential landscape
risk assessment using Markov-CA model and analytical hierarchy process. ISPRS Int. J. Geo-Inf. 2020, 9, 134. [CrossRef]
45. Ma, C.; Zhang, G.; Zhang, X.; Zhao, Y.J. Application of Markov model in wetland change dynamics in Tianjin Coastal Area, China.
Procedia Environ. Sci. 2012, 13, 252–262. [CrossRef]
46. He, D.; Zhou, J.; Gao, W.; Guo, H.; Yu, S.; Liu, Y. An integrated CA-markov model for dynamic simulation of land use change in
Lake Dianchi watershed. Beijing Daxue Xuebao Ziran Kexue Ban/Acta Sci. Nat. Univ. Pekin. 2014, 50, 1095–1105.
47. Singh, S.K.; Laari, P.B.; Mustak, S.; Srivastava, P.K.; Szabó, S. Modelling of land use land cover change using earth observation
data-sets of Tons River Basin, Madhya Pradesh, India. Geocarto Int. 2018, 33, 1202–1222. [CrossRef]
48. Nouri, J.; Gharagozlou, A.; Arjmandi, R.; Faryadi, S.; Adl, M. Predicting urban land use changes using a CA–Markov model.
Arab. J. Sci. Eng. 2014, 39, 5565–5573. [CrossRef]
49. Tali, J.; Divya, S.; Murthy, K. Influence of urbanization on the land use change: A case study of Srinagar City. Geoinformatics 2013,
1, 271–283.
Sustainability 2024, 16, 11130 19 of 19

50. Congalton, R.G.; Green, K. Assessing the Accuracy of Remotely Sensed Data: Principles and Practices; CRC Press: Boca Raton, FL, USA,
2019.
51. Mishra, V.N.; Rai, P.K.; Mohan, K. Prediction of land use changes based on land change modeler (LCM) using remote sensing:
A case study of Muzaffarpur (Bihar), India. J. Geogr. Inst. Jovan Cvijic SASA 2014, 64, 111–127. [CrossRef]
52. Omar, N.Q.; Ahamad, M.S.S.; Wan Hussin, W.M.A.; Samat, N.; Binti Ahmad, S.Z. Markov CA, multi regression, and multiple
decision making for modeling historical changes in Kirkuk City, Iraq. J. Indian Soc. Remote Sens. 2014, 42, 165–178. [CrossRef]
53. Eastman, J.R. IDRISI Kilimanjaro: Guide to GIS and Image Processing; Clark Lab, Clark University: Worcester, MA, USA, 2003.
54. Soomro, A.G.; Babar, M.M.; Arshad, M.; Memon, A.; Naeem, B.; Ashraf, A. Spatiotemporal variability in spate irrigation systems
in Khirthar National Range, Sindh, Pakistan (case study). Acta Geophys. 2020, 68, 219–228. [CrossRef]
55. Raj, S.; Rawat, K.S.; Tripathi, V. Multi-Temporal Image Processing for LULC Classification and Change Detection. Eur. J. Remote
Sens. 2024, 42, 1349–1357. [CrossRef]
56. Puttinaovarat, S.; Khaimook, K.; Horkaew, P. Land use and land cover classification from satellite images based on ensemble
machine learning and crowdsourcing data verification. Int. J. Cartogr. 2023, 1–21. [CrossRef]
57. Basheer, S.; Wang, X.; Farooque, A.A.; Nawaz, R.A.; Liu, K.; Adekanmbi, T.; Liu, S. Comparison of land use land cover classifiers
using different satellite imagery and machine learning techniques. Remote Sens. 2022, 14, 4978. [CrossRef]
58. Mondal, M.S.; Sharma, N.; Kappas, M.; Garg, P.K. CA Markov modeling of land use land cover change predictions and effect of
numerical iterations, image interval (time steps) on prediction results. Int. Arch. Photogramm. Remote Sens. Spat. Inf. Sci. 2020,
43, 713–720. [CrossRef]
59. Hasan Aksoy, S.K. Monitoring of land use/land cover changes using GIS and CA-Markov modeling techniques: A study in
Northern Turkey. Environ. Monit. Assess. 2021, 193, 507. [CrossRef] [PubMed]
60. Khwarahm, N.R.; Najmaddin, P.M.; Ararat, K.; Qader, S. Past and future prediction of land cover land use change based on earth
observation data by the CA–Markov model: A case study from Duhok governorate, Iraq. Arab. J. Geosci. 2021, 14, 1544. [CrossRef]
61. Aniah, P.; Bawakyillenuo, S.; Codjoe, S.N.A.; Dzanku, F.M. Land use and land cover change detection and prediction based on
CA-Markov chain in the savannah ecological zone of Ghana. Environ. Chall. 2023, 10, 100664. [CrossRef]
62. CAPMAS. Central Agency for Public Mobilization and Statistics. Egypt in Figures. 2023. Available online: https://www.capmas.
gov.eg/HomePage.aspx (accessed on 1 November 2024).
63. Biondini, M.; Kandus, P. Transition matrix analysis of land-cover change in the accretion area of the Lower Delta of the Paraná
River (Argentina) reveals two succession pathways. Wetlands 2006, 26, 981–991. [CrossRef]
64. Munthali, M.; Mustak, S.; Adeola, A.; Botai, J.; Singh, S.; Davis, N. Modelling land use and land cover dynamics of Dedza district
of Malawi using hybrid Cellular Automata and Markov model. Remote Sens. Appl. Soc. Environ. 2020, 17, 100276. [CrossRef]
65. Maviza, A.; Ahmed, F. Analysis of past and future multi-temporal land use and land cover changes in the semi-arid Upper-
Mzingwane sub-catchment in the Matabeleland south province of Zimbabwe. Int. J. Remote Sens. 2020, 41, 5206–5227. [CrossRef]
66. Huang, G.; Jiang, Y. Urbanization and socioeconomic development in inner Mongolia in 2000 and 2010: A GIS analysis.
Sustainability 2017, 9, 235. [CrossRef]
67. Gidey, E.; Dikinya, O.; Sebego, R.; Segosebe, E.; Zenebe, A. Cellular automata and Markov Chain (CA_Markov) model-based
predictions of future land use and land cover scenarios (2015–2033) in Raya, northern Ethiopia. Model. Earth Syst. Environ. 2017,
3, 1245–1262. [CrossRef]
68. Gashaw, T.; Tulu, T.; Argaw, M.; Worqlul, A.W. Evaluation and prediction of land use/land cover changes in the Andassa
watershed, Blue Nile Basin, Ethiopia. Environ. Syst. Res. 2017, 6, 17. [CrossRef]
69. Pontius, R.G., Jr.; Schneider, L. Land-cover change model validation by an ROC method for the Ipswich watershed, Massachusetts,
USA. Agric. Ecosyst. Environ. 2001, 85, 239–248. [CrossRef]
70. Keshtkar, H.; Voigt, W. A spatiotemporal analysis of landscape change using an integrated Markov chain and cellular automata
models. Model. Earth Syst. Environ. 2016, 2, 10. [CrossRef]
71. Luo, G.; Amuti, T.; Zhu, L.; Mambetov, B.T.; Maisupova, B.; Zhang, C. Dynamics of landscape patterns in an inland river delta of
Central Asia based on a cellular automata-Markov model. Reg. Environ. Chang. 2015, 15, 277–289. [CrossRef]
72. Huang, Y.; Yang, B.; Wang, M.; Liu, B.; Yang, X. Analysis of the future land cover change in Beijing using CA–Markov chain
model. Environ. Earth Sci. 2020, 79, 60. [CrossRef]

Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual
author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to
people or property resulting from any ideas, methods, instructions or products referred to in the content.