Open access peer-reviewed chapter - ONLINE FIRST

CA-Markov Approach in Dynamic Modelling of LULCC Using ESA CCI Products Over Zambia

Written By

Charles Bwalya Chisanga, Chizumba C. Shepande and Edson Nkonde

Submitted: January 28th, 2022Reviewed: February 3rd, 2022Published: April 7th, 2022

DOI: 10.5772/intechopen.103032

IntechOpen
Geographic Information SystemEdited by Yuanzhi Zhang

From the Edited Volume

Geographic Information System [Working Title]

Prof. Yuanzhi Zhang and Dr. Qiuming Cheng

Chapter metrics overview

16 Chapter Downloads

View Full Metrics

Abstract

The Markov, Cell Atom and CA-Markov modules in TerrSet v19.0 have been applied to predict LULC maps for 2030 over Zambia. The European Space Agency Climate Change Initiative (ESA CCI) classified LULC maps for 2000, 2010 and 2020 were used in this study. The ESA-CCI LULC maps were reclassified using QGIS 3.20 into 10 classes. The 2000 and 2010 LULC maps were used to predict the 2020 LULC maps. The Kappa statistics between the 2020 reference and predicted LULC maps was kappa (0.9918). The probability and transition matrix between the 2010 and 2020 LULC maps were used as inputs into the CA-Markov module to generate the 2030 LULC map. The LULCC from 2020-2030 shows an expansion and contraction of different classes. However, Built-up (42.38% [481.82 km2]) constitutes major changes among the LULC classes. However, Cropland, Dense forest, Grassland, Wetland and Bare land will reduce by 376.00, 1087.65, 70.60, 26.67 and 0.36 km2, respectively. Other LULC changes from 2020-2030 are in seasonally flooded grassland (94.66 km2), Sparse forest (497.05 km2), Shrub land (410.11 km2) and Water body (77.63 km2). The prediction of future LULC from historical LULC using CA-Markov model plays a significant role in policy making and land use planning.

Keywords

  • CA-Markov
  • cellular automata
  • ESA CCI
  • LULC
  • Markov

1. Introduction

Land use/land cover (LULC) has been defined as the physical composition and characteristics of land elements on the earth’s surface [1]. The changes in LULC are caused by both natural and anthropogenic factors such as deforestation, and the intensification of agriculture [1, 2, 3]. Leta et al. [2] noted that the changes in LULC involve multifaceted processes that are dynamic, non-linear human-nature interactions that cause substantial land surface changes. The world-wide changes in LULC trajectory in the recent past have been characterised by gains and losses in agriculture and forests, respectively [2, 3, 4]. Researchers like Leta et al. [2], Pérez-Vega et al. [4], and (Kolb et al. [5] argue that LULC changes are associated with major changes in forest land to agriculture, built-up and deforestation.

The dynamics of LULC change may be monitored using remotely sensed data (satellite imagery). The satellite imagery data is provided at different spatial, temporal and spectral resolutions and this is used to detect changes on the earth’s surface [6]. The prediction of LULC change is vital in strategic developmental plans and in the management of LULC [7, 8]. There are myriads of Land Use Land Cover Change (LUCC) models that have been developed and these includes; Markov model, Cellular Automata (CA) models, statistical models, GeoMod, evolutionary models, CLUE-S model, hybrid models and multi-agent models [9, 10, 11, 12, 13]. Markov models can quantitatively predict the dynamic changes in landscape patterns [13]. Unfortunately, these models cannot resolve the spatial patterns of landscape change [10, 14]. Conversely, the CA models can predict the spatial distribution of landscape patterns but fail to predict temporal changes [13]. Because of the above reasons, researchers use a combination of CA and Markov to model LULCC dynamically [10, 13]. The most convenient tool used in simulating the spatio-temporal LULCC and processes in the landscape is the CA-Markov model [11, 15, 16]. Researchers have used the CA-Markov model to model dynamically the spatio-temporal LULCC and predict future scenarios [15, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26].

Using a Markov process, the future state of a system can be simulated using the immediately preceding state [8]. The Markov model describes LULC change from one time period to another and this is the basis used to project future changes [8, 27]. The Markov outputs are the transition probability, area matrices and probability images of LULC change from the initial time to time two which displays the nature of change as the basis for projecting a future time period. The LULCC has been analysed in Europe, USA, South America, Asia and Africa [2, 28]. The expansion of agricultural fields in Africa has been influenced by population growth [2]. Agriculture has been recognised as the main driver of land use change (LUC). The dynamics of urban growth are linked to factors associated with demography especially in developing countries.

The future LULC condition can be determined using modelling such as CA-Markov models. The CA-Markov is a combination of cellular automata and Markov. It is a LULC prediction procedure that adds an element of spatial contiguity and the likely spatial distribution of transitions to Markov change analysis [10]. The CA-Markov models can be used to model and monitor LULC change at spatial and temporal scale [8]. It has been used by [27, 29, 30, 31] to successfully model future LULCC.

Researchers involved in LULCC agree that the intellectual foundation of validation for land use change models is insufficient [2]. The literature reviewed suggest insufficient research on rigorous validation of LUCC models has been conducted [32, 33]. The objectives of this study were to investigate and analyse the spatio-temporal ESA CCI LULCC for 2000, 2010 and 2020 and to predict LULCC for 2030. The study attempts to analyse the spatio-temporal LULC change based on 2000, 2010, 2020 and 2030 using Zambia as a case study. The study also simulated the spatial variation in LULC change in 2000, 2010 and 2020 and used CA-Markov model to predict 2030 LULCC.

Advertisement

2. Materials and methods

2.1 Study area

The study covers the whole country, Zambia is used in this as shown in Figure 1. It is located in southern Africa and its spatial extent is between −8.27o to −18.075o latitude South and 22.00o to 33.7o longitude East. Zambia is divided into three Agro-ecological Regions (AERI, AERII and AERIII) and ten provinces as shown in the figure below.

Figure 1.

Location of Zambia in southern Africa.

2.2 Sources of classified satellite imagery data

The global land cover maps utilised in this study were the European Space Agency Climate Change Initiative (ESA CCI) datasets [34, 35]. The LULC maps for 2000, 2010 and 2020 were downloaded from the ESA web site (https://cds.climate.copernicus.eu/cdsapp#!/dataset/satellite-land-cover?tab=form). The LULC maps for Zambia were clipped using the shapefiles for the administrative boundary. The annual ESA CCI LULC maps v2.1.1 (2000, 2010) and v2.0.7 (2020) have better quality in the classification and representation of change compared to v1.6.1 epoch-based datasets. The spatial resolutions of the LULC maps are at 300 m. These maps are reliable and cover the whole globe [34, 35]. The individual pixel value matches the labels of the land cover classes as defined by the United Nations (UN) Land Cover Classification System LCCS [35]. The UN LCCS classifiers having 22 classes makes provisions for further conversion into the Plant Functional Types distribution required by the Earth System Models. The global land cover datasets are produced in Coordinate Reference System (CRS).

2.3 Reclassification of land use types

The ESA CCI LULC datasets have 22 classes [34, 35] which were reclassified into 10 classes (Table 1). The reference for integration into 10 classes is the Food and Agriculture (FAO) Openforis EarthMap and Vito Landcover viewer tools located at http://www.openforis.org/newwebsite/tools/earth-map.html and https://lcviewer.vito.be. The ESA 22+ classes were reclassified into 10 classes using r.reclassmodule in QGIS software.

Class name200020102020
Area in km2
1Cropland45733.40146380.68646055.166
2Seasonally, flooded grassland16348.38416284.21216369.564
3Sparse forest66244.03565002.90365425.752
4Dense forest444321.515450001.335449153.425
5Shrub land129878.716123798.324123985.97
6Grassland8333.3168286.1978216.149
7Wetlands27777.51128531.53628540.857
8Built-up401.316642.1541102.362
9Bare land41.9345.20744.571
10Waterbody13947.96614055.53514134.276
Total753028.09753028.09753028.09

Table 1.

Classified LULC map of 2000, 2010 and 2020 in square kilometres.

2.4 Land use and land cover change using CA-Markov model

Researchers have used CA-Markov to monitor LULCC and predictions [12, 15, 17, 19, 20, 21, 22, 25, 26]. The CA-Markov model in TerrSet v19.0 has been adopted in this study to obtain reliable results for Zambia. The 2030 LULC map was predicted based on the state of 2020 LULC. The CA-Markov model in TerrSet v19.0 uses a Markov, CellAtom (Cellular Automata) and CA-Markov modules [10, 11]. The CA-Markov model was used to calculate the amounts of change that may occur to some selected locations in the future [10, 11]. It analyses two qualitative LULC maps from different dates and produces a transition matrix, a transition areas matrix, and a set of conditional probability images [10]. The CA-Markov model is a stochastic process model that describes the probability of change from one date to another [2]. The transition probability would be the probability that a land cover type (pixels) at the initial time (t0) changes to another land cover type at the second time (t1). The changes in LULC between 2000 and 2010 were used to develop a transition probability, area matrices and probability images.

The Markov modules were used to generate the suitability transition images and change matrix. The cellular automata transition rule was implemented in TerrSet v19.0 using a combination of FILTER and RECLASS modules [10]. The CellAtom module was used to generate transition suitability images. The FILTER and RECLASS modules were used to create filters and reclassifying the LULC maps before implementing the cellular automata transition rules [10]. On the other hand, the CA-Markov module output was the 2020 predicted LULC map. The VALIDATE module was used to produce the kappa statistics using the 2020 reference LULC map and the 2020 predicted LULC map. More details on how the process was implemented are provided by Eastman [11].

2.5 Validation of the simulated map

An important stage in the development of any predictive change model is validation [10]. Validation is a procedure used to assess the quality of the predicted LULC map against a reference map [2]. The 2020 predicted LULC map was validated using the 2020 reference LULC map. The VALIDATE module in TerrSet v19.0 [11] was used as it provides a comparative analysis based on the Kappa Index of Agreement (KIA). KIA is essentially a statement of proportional accuracy, adjusted for chance agreement [36]. The kappa for stratum level location (KlocationStrata) is a quantification of the spatial accuracy within pre-identified strata, and it indicates how well the grid cells are situated within the strata [11]. The blend of Kstandard, Kno, Klocation, and Klocation strata scores is considered for a comprehensive evaluation of the overall accuracy both in terms of location and quantity. The statistics; AgreementQuantity, AgreementChance, AgreementGridCell, DisagreementGridCell, and DisagreementQuantity are used to ascertain the strength of the agreement. The Kstandard, Kno, Klocation, KlocationStrata, AgreeGridcell and AgreeQuantity were used in validating the accuracy of the prediction LULC map. The possible ranges of map comparison and level of kappa agreement values are shown in Table 2.

NumberValuesStrength of agreement
1<0Poor
20.01–0.40Slight
30.41–0.60Moderate
40.61–0.80Substantial
50.81–1.00Almost Perfect

Table 2.

Comparison and level of kappa agreement values [2, 37].

Advertisement

3. Results and discussion

3.1 Gains and losses in LULC

The gains and losses by category are shown in Table 3. All LULC classes experienced gains and losses except Bare land and Built-up. Built-up and Bare land categories had gains and losses from 2000 to 2010, 2010–2020 and 2000–2020, respectively. The Built-up class from 2000 to 2020 increased by 697.62 km2. From 2000 to 2020, the net gains in cropland, dense forest, wetland and water body were relatively large. Sparse forest, shrub land and grassland LULC were converted to other LULC classes. The sparse forest, shrub land and grassland LULC classes experienced net losses to other classes.

Class nameActualActualActualPredictions
2000–20102010–20202000–20202020–2030
GainslossesGainslossesGainslossesGainslosses
Cropland1001.85351.11409.61733.821388.321061.81424.08798.78
Seasonally, flooded grassland197.81262.67272.80188.04464.71444.81282.57188.50
Sparse forest1257.642481.921497.221076.092710.593514.151599.301104.65
Dense forest9205.643492.132377.903218.7811426.646554.022453.633533.13
Shrub land1597.377723.251763.111578.303234.909175.972016.561609.81
Grassland84.58131.7537.41107.61120.78238.1638.51109.27
Wetlands995.02247.65140.41131.101117.37360.69143.91170.44
Built-up239.820.00457.800.00697.620.00479.360.00
Bare land7.834.615.626.2610.3210.325.625.99
Waterbody210.98103.00141.6163.48329.83143.72144.4667.44

Table 3.

Gains and losses by category.

The LULCC from 2000 to 2020 in km2 is shown in Table 4. Cropland, dense forest, wetlands, built-up, bare land and water body had an overall increase of 1.40%, 1.26%, 2.64%, 37.5, 7.25% and 0.77% from 2000 to 2010, respectively. However, there was an overall reduction of 0.39%, 1.91%, 4.91% and 0.57% in seasonally flooded grassland, sparse forest, shrub land, and grassland from 2000 to 2010, respectively. During 2010 and 2020, cropland, dense forest, grassland and bare land experienced a reduction of 0.71%, 0.19%, 0.85% and 1.43%, respectively. There was an overall increase in seasonally flooded grassland, sparse forest, shrub land, wetland, built-up and water body as shown in Table 4. Leta et al. [2] observed that the expansion of agricultural land for both domestic and commercial production are drivers of LULC change. Furthermore, cropland is on an increase while shrub land, forest and grassland are decreasing in the world.

Class name2000–2010% ∆2010–2020% ∆2000–2020% ∆
Cropland647.291.40−325.52−0.711550.233.28
Seasonally, flooded grassland−64.17−0.3985.350.52−72.99−0.45
Sparse forest−1241.13−1.91422.850.65−2104.61−3.28
Dense forest5679.821.26−847.91−0.198689.641.92
Shrub land−6080.39−4.91187.650.15−11318.34−9.55
Grassland−47.12−0.57−70.05−0.85−81.00−0.98
Wetlands754.032.649.320.032503.338.27
Built-up240.8437.50460.2141.75531.6056.98
Bare land3.287.25−0.64−1.4316.3328.03
Waterbody107.570.7778.740.56285.812.01

Table 4.

LULCC from 2000 to 2020 in km2.

From 2000 to 2020, cropland, dense forest, wetlands, built-up, bare land and water body had an overall increase of 3.28%, 1.92%, 8.27%, 56.98, 28.03% and 2.01%, respectively. However, there was an overall reduction in seasonally flooded grassland, sparse forest, shrub land and grassland by 0.45%, 3.28%, 9.55% and 0.98%, respectively. The LULCC analysis of 2000 to 2010, 2010 to 2020 and 2000 to 2020, results indicated a positive increase in built-up and water body. Rapid population growth and migration of people from rural to urban areas have resulted in unprecedented changes in LULC. From 2000 to 2010, 2010–2020 and 2000–2020, built-up class had the largest increase at 37.50%, 41.75% and 56.98% followed by bare land at 7.25% (2000–2010) and 28.03% (2000–2020). Shrub land experienced a loss of 4.91% (2000–2010) and 9.55% (2000–2020) to other LULC classes. An increase in built-up and cropland LULC classes is anticipated to meet the demand of the population for residential and food production. This should be taken into consideration by policy makers and planners in future land use plans for sustainable management of natural resources such as soil fertility, water resources and forests (Figure 2).

Figure 2.

Classified LULC for 2000, 2010, 2020 and 2030.

3.2 Transition matrices for Markov 2020 and 2030 prediction

The transition matrix of Markov 2020 and 2030 predictions is shown in Tables 5 and 6. The CA-Markov model is simple to calibrate and can simulate LULCC dynamically with high efficiency [12]. Furthermore, it can simulate complex and multiple land cover patterns. The combination of Markov and Cellular Automata (CA-Markov) allows simulating the evolution of the geographical area represented by pixels. CA-Markov is a combined cellular automata/Markov change LULC prediction procedure that adds an element of spatial contiguity as well as knowledge of the likely spatial distribution of transitions to Markov change analysis [10]. Each pixel can take a value from a finite set of states. All pixels are affected by a transition function that takes as arguments of the measured values and values of the neighbouring pixels as a function of time [8, 10, 38].

C 1C 2C 3C 4C 5C 6C 7C 8C 9C 10
C 10.98240.00000.00000.01120.00600.00000.00000.00040.00000.0000
C 20.00000.97410.00000.02240.00340.00000.00000.00000.00000.0001
C 30.00000.00000.95270.04180.00510.00000.00000.00030.00000.0000
C 40.00260.00080.00360.98220.00560.00040.00460.00000.00000.0002
C 50.00430.00030.00500.05720.93110.00010.00010.00170.00000.0003
C 60.00170.00020.00020.00550.01130.97420.00000.00310.00000.0038
C 70.00000.00000.00000.00860.00100.00000.98110.00000.00000.0093
C 80.00110.00110.00110.00110.00110.00110.00110.99000.00110.0011
C 90.00950.00000.00240.00000.00000.00000.00000.10710.88100.0000
C 100.00010.00000.00020.00100.00180.00010.01410.00000.00000.9826

Table 5.

Transition matrix of Markov 2020 prediction based on LULC maps 2000 and 2010.

C 1 = Cropland; C 2 = Seasonally & flooded grassland; C 3 = Sparse forest; C 4 = Dense forest; C 5 = Shrub land; C 6 = Grassland; C 7 = Wetland; C 8 = Built-up; C 9 = Bare land; C 10 = Waterbody.

C 1C 2C 3C 4C 5C 6C 7C 8C 9C 10
C 10.97430.00000.00000.01370.01020.00000.00000.00170.00000.0000
C 20.00000.97860.00000.01900.00230.00000.00000.00010.00000.0000
C 30.00000.00000.97350.02180.00380.00000.00000.00090.00000.0000
C 40.00160.00120.00700.98290.00640.00010.00050.00010.00000.0001
C 50.00130.00060.00270.01270.97740.00020.00000.00470.00000.0005
C 60.00200.00010.00020.00180.01480.97700.00000.00390.00000.0003
C 70.00000.00000.00000.00400.00130.00000.98540.00000.00000.0092
C 80.00110.00110.00110.00110.00110.00110.00110.99000.00110.0011
C 90.00870.00000.00000.00220.01300.00000.00000.12380.85230.0000
C 100.00060.00000.00000.00040.00210.00040.01090.00000.00000.9855

Table 6.

Transition matrix of Markov 2030 prediction based on LULC maps 2010 and 2020.

C 1 = Cropland; C 2 = Seasonally & flooded grassland; C 3 = Sparse forest; C 4 = Dense forest; C 5 = Shrub land; C 6 = Grassland; C 7 = Wetland; C 8 = Built-up; C 9 = Bare land; C 10 = Waterbody.

3.3 Validation of CA-Markov LULC prediction results

Table 7 shows the statistics of model validation. The Kappa Index of Agreement (KIA), Kno (kappa for no information), Klocation (kappa for location), Kstandard (kappa for standard), and KlocationStrata (kappa for stratum-level location) [39] shown in Table 7 indicate the accuracy of the prediction. The Kstandard, Kno, Klocation, KlocationStrata, AgreeGridcell, AgreeStrata, AgreeQuantity, AgreeChange, DisagreeQuantity, DisagreeStrata and DisagreeGridcell using user-defined filters of 3x3, 5x5 and 7.7 shown in Table 7 indicate the performance of the prediction. The Kno indicates the agreement between the 2020 reference and 2020 predicted LULC map. The Klocation was almost perfect and indicated the spatial accuracy in the overall LULC in each category between the predicted and reference map [40]. Table 8 shows the comparison between the predicted and reference LULC map. The overall Kappa between the 2020 reference and 2020 predicted map was 0.9918. The results indicate that reference and predicted LULC maps have many similarities. The difference between the maps is due to changes that will take place in the future LULC maps.

Filter 3 x 3Filter 5 x 5Filter 7 x 7
Kstandard0.99180.99180.9917
Kno0.99390.99390.9938
Klocation0.99320.99320.9932
KlocationStrata0.99320.99320.9932
kappa0.99180.99180.9917
AgreeGridcell0.67070.67060.6706
AgreeStrata0.0000.0000.0000
AgreeQuantity0.23290.23290.2329
AgreeChance0.09090.09090.0909
DisagreeQuantity0.00100.00100.0010
DisagreeStrata0.00000.00000.0000
DisagreeGridcell0.00460.00460.0046

Table 7.

Kappa statistics for 2020 reference versus 2020 predicted LULC map.

Class name2020 reference2020 predictedDifference between reference and predicted
Filter5x5Area km2%
1Cropland46055.16647283.635−1228.5−0.16
2Seasonally, flooded grassland16369.56416275.39394.1710.01
3Sparse forest65425.75264139.4281286.320.17
4Dense forest449153.425453011.151−3857.7−0.51
5Shrub land123985.97118560.3815425.590.72
6Grassland8216.1498252.317−36.1680.00
7Wetlands28540.85730280.842−1740−0.23
8Built-up1102.362932.914169.4480.02
9Bare land44.57158.258−13.6870.00
10Waterbody14134.27614233.771−99.495−0.01
Total753028.092753028.092

Table 8.

Comparison of 2020 reference and 2020 predicted LULC map.

3.4 Predicted LULCC in 2030

Figure 2 shows the 2000, 2010, 2020 and 2030 LULCC maps over Zambia. The Built-up constitutes major changes among the LULC classes with an increasing trend at 42.38% (481.82 km2) from 2020 to 2030 (Table 8, Figure 3). Researchers such as Jain et al. [41] have observed that built-up areas have increased while forests have decreased. The growth in built-up consequently causes a reduction in cropland or agricultural land. The 2020 to 2030 LULC prediction maps indicate that Cropland, Dense forest, Grassland, Wetland and Bare land will reduce by 0.82% (376.00 km2), 0.24% (1087.65 km2), 0.86% (70.60 km2), 0.09% (26.67 km2) and 0.80% (0.36 km2), respectively. However, LULC classes of seasonally flooded grassland (0.58%, 94.66 km2), Sparse forest (0.76%, 497.05 km2), Shrub land (0.33%, 410.11 km2), Built-up (42.38%, 481.82 km2) and Water body (0.55%, 77.63 km2) will increase from 2020 to 2030. Analysing historical LULCC from future LULCC using CA-Markov plays a significant role in forest management and land use planning [29, 31]. The CA-Markov model can be applied in modelling and quantifying transition rates and different states of diverse land uses [24]. The CA-Markov model can simulate larger areas efficiently [13] such has been undertaken in this study. However, CA-Markov is very sensitive in simulating smaller regions as noted by [13, 42].

Figure 3.

Gains and losses by categories of LULC from 2020 to 2030 over Zambia.

The gains and losses by category indicate that Built-up will have gains of 479.36 km2 between 2020 and 2030. The Cropland, Dense forest, Grassland, Wetland and Bare land LULC categories will experience greater losses compared to gains as shown in Table 9. On the other hand, seasonally flooded grassland, Sparse forest, Shrub land and Water body LULC categories will experience greater gains than losses. Therefore, LULC prediction is important as it contributes to national, and regional planning and management of natural resources [3].

ActualPredicted
Class200020102020202020302000–2010% ∆2010–2020% ∆2000–2020% ∆2020–2030% ∆
C 145733.4046380.6946055.1746421.8546045.85647.281.40−325.52−0.71321.760.70−376.00−0.82
C 216348.3816284.2116369.5616280.1216374.78−64.17−0.3985.350.5221.180.1394.660.58
C 366244.0465002.9065425.7565019.7265516.76−1241.13−1.91422.850.65−818.28−1.25497.050.76
C 4444321.51450001.34449153.42450072.03448984.385679.821.26−847.91−0.194831.911.08−1087.65−0.24
C 5129878.72123798.32123985.97123621.05124031.17−6080.39−4.91187.650.15−5892.75−4.75410.110.33
C 68333.328286.208216.158285.548214.93−47.12−0.57−70.05−0.85−117.17−1.43−70.60−0.86
C 727777.5128531.5428540.8628562.6728536.00754.032.649.320.03763.352.67−26.67−0.09
C 8401.32642.151102.36655.011136.83240.8437.50460.2141.75701.0563.59481.8242.38
C 941.9345.2144.5745.3044.943.287.25−0.64−1.432.645.93−0.36−0.80
C 1013947.9714055.5414134.2814064.8014142.43107.570.7778.740.56186.311.3277.630.55
Total753028.09753028.09753028.09753028.09753028.09

Table 9.

Percent changes in LULC classes from 2000 to 2030.

C 1 = Cropland; C 2 = Seasonally & flooded grassland; C 3 = Sparse forest; C 4 = Dense forest; C 5 = Shrub land; C 6 = Grassland; C 7 = Wetland; C 8 = Built-up; C 9 = Bare land; C 10 = Waterbody.

Advertisement

4. Conclusions

The prediction of future LULC from historical LULC using CA-Markov model plays a significant role in policy making, land use planning and in natural resource management. There are major changes in built-up areas during the historical (2000 and 2020) and future (2030). An increase in built-up and cropland LULC classes is anticipated to meet the demand of the population for residential and food production. This should be taken into consideration by policy makers and planners in future land use plans for sustainable management of natural resources. This study has shown that the ESA CCI products can be used in LULC analysis and in predicting future LULCC.

Advertisement

Acknowledgments

The authors would like to thank ESA for the provision of the ESA CCI LULC maps and codes for converting NetCDF to Gtiff used in this study.

Advertisement

Author contributions

Downloading, preparation and analysis of the ESA CCI data by CBC. Methodology development by CBC. Model setup, simulation and validation by CBC. Drafting of the original manuscript by CBC. Abstract and conclusion by CCS. Reviews and editing by EN and CCS.

Advertisement

Contribution to the field statement

The information generated from change detection can be used in land use planning.

The European Space Agency Climate Change Initiative (ESA CCI) products can be used in LULCC analysis and in predicting future LULCC at 300 m spatial resolution.

The prediction of future LULC from historical LULC using CA-Markov model plays a significant role in policy making and land use planning.

Advertisement

Funding information

There was no funding for this research.

Advertisement

Competing interests

The authors declare no competing interests.

Advertisement

Data availability statement

The datasets analysed during the current study are available in the ESA CCI repository (https://cds.climate.copernicus.eu/cdsapp#!/dataset/satellite-land-cover?tab=form). The reclassified datasets are available from the corresponding author on reasonable request.

Institutional Review Board Statement

Not applicable.

Informed Consent Statement

Not applicable.

References

  1. 1.Yulianto F, Prasasti I, Pasaribu JM, Fitriana HL, Zylshal Haryani NS, Sofan P. The dynamics of land use/land cover change modeling and their implication for the flood damage assessment in the Tondano watershed, North Sulawesi Indonesia, Modeling Earth Systems and Environment. 2006;2:20. DOI: 10.1007/ s40808-016-0100-3
  2. 2.Leta MK, Demissie TA, Tränckner J. Modeling and Prediction of Land Use Land Cover Change Dynamics Based on Land Change Modeler (LCM) in Nashe Watershed, Upper Blue Nile Basin, Ethiopia. Sustainability. 2021;13:17. DOI: 10.3390/su13073740
  3. 3.Saputra MH, Lee HS. Prediction of land use and land cover changes for north sumatra, indonesia, using an artificial-neural-network-based cellular automaton. Sustainability. 2019;11:1-16. DOI: 10.3390/su11113024
  4. 4.Pérez-Vega A, Mas JF, Ligmann-Zielinska A. Comparing two approaches to land use/cover change modeling and their implications for the assessment of biodiversity loss in a deciduous tropical forest. Environmental Modeling Software. 2012;29:11-23
  5. 5.Kolb M, Mas JF, Galicia L. Evaluating drivers of land-use change and transition potential models in a complex landscape in Southern Mexico. International Journal of Geographical Information Science. 2013;27:1804-1827
  6. 6.Halmy MWA, Gessler PE, Hicke JA, Salem BB. Land use/land cover change detection and prediction in the north-western coastal desert of Egypt using Markov-CA. Applied Geography. 2015;63:101-112. DOI: 10.1016/j.apgeog.2015.06.015
  7. 7.Ibrahim WYW, Ludin ANM. Spatiotemporal Land Use Change Analysis Using Open-source GIS and Web Based Application. International Journal of Built Environment Sustainability. 2014;2:10. DOI: 10.11113/ijbes.v2.n2.64
  8. 8.Nguyen TTH, Ngo TTP. Land use/land cover change prediction in Dak Nong Province based on remote sensing and Markov Chain Model and Cellular Automata. Journal of Vietnamese Environment. 2018;9:132-140. DOI: 10.13141/jve.vol9.no3.pp132-140
  9. 9.Dawn CP, Steven MM, Marco AJ, Matthew JH, Peter D. Multi-agent systems for the simulation of land-use and land-cover change: A review. Annals of the Association of American Geographers. 2016;93:314-337. DOI: 10.1111/1467-8306.9302004
  10. 10.Eastman JR. IDRISI Taiga: Guide to GIS and Image Processing Volume-Manual version. 16.02. Worcester, MA, USA: Clark Labs, Clark University. 2009. p. 325
  11. 11.Eastman JR. The IDRISI Applications Programming Interface for the TerrSet Geospatial Monitoring and Modeling System: A User’s Guide. Version 4.0, Worcester, USA. 2019. p. 1–41
  12. 12.Liping C, Yujun S, Saeed S. Monitoring and predicting land use and land cover changes using remote sensing and GIS techniques - A case study of a hilly area, Jiangle, China. PLoS One. 2018;13:1-23. DOI: 10.1371/journal.pone.0200493
  13. 13.Yang C, Wu G, Chen J, Li Q, Ding K, Wang G, Zhang C. Simulating and forecasting spatio-temporal characteristic of land-use/ cover change with numerical model and remote sensing: A case study in Fuxian Lake Basin China, European Journal of Remote Sensing. 2019;52:374–384. DOI: 10.1080/22797254.2019.1611387
  14. 14.Balzter H, Braun PW, Kohler W. Cellular automata models for vegetation dynamics. Ecological Modelling. 1998;107:113-125. DOI: 10.1016/S0304-3800(97)00202-0
  15. 15.Hamad R, Balzter H, Kolo K. Predicting Land Use/Land Cover Changes Using a CA-Markov Model under Two Different Scenarios. Sustainability. 2018;10:1-26. DOI: 10.3390/su10103421
  16. 16.Eastman RJ. TerrSet Geospatial Monitoring and Modeling System - Manual., Worcester, USA: Clark Labs, Clark University. 2016. p. 395
  17. 17.Abdulrahman AI, Ameen SA. Predicting Land use and land cover spatiotemporal changes utilizing CA-Markov model in Duhok district between 1999 and 2033. Academic Journal of Nawroz University. 2020;9:18. DOI: 10.25007/ajnu.v9n4a892
  18. 18.Adhikari S, Southworth J. Simulating Forest Cover Changes of Bannerghatta National Park Based on a CA-Markov Model: A Remote Sensing Approach. Remote Sensing. 2012;4:3215-3243. DOI: 10.3390/rs4103215
  19. 19.Chaula JA. Ca-Markov Model for Simulating Land use Land Cover Dynamics in Rufiji Delta of Tanzania. International Journal of Innovative Science Research Technology. 2019;4:802-814
  20. 20.Hua AK. Application of CA-Markov model and land use/land cover changes in Malacca river watershed Malaysia, Applied Ecology and Environmental Research. 2017;15:605–622. DOI: 10.15666/aeer/1504_605622
  21. 21.Khawaldah HA, Farhan I, Alzboun NM. Simulation and prediction of land use and land cover change using GIS, remote sensing and CA-Markov model. Global Journal of Environmental Science Management. 2020;6:215-232. DOI: 10.22034/gjesm.2020.02.07
  22. 22.Matlhodi B, Kenabatho PK, Parida BP, Maphanyane JG. Analysis of the future land use land cover changes in the gaborone dam catchment using ca-markov model: Implications on water resources. Remote Sensing. 2021;13:20. DOI: 10.3390/rs13132427
  23. 23.Samat, N., 2009: Integrating GIS and CA-MARKOV model in evaluating urban spatial growth Malaysian. Journal of Environmental Management, 10, 83–100
  24. 24.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. Mathematical and Computer Modelling. 2011;54:938-943. DOI: 10.1016/j.mcm.2010.11.019
  25. 25.Sunday A, Abdulkadir A, Anene NC. Simulation and Prediction of Urbanization in Makurdi City Nigeria using CA-Markov Technique. FUTY Journal of the Environment. 2020;14:12-21
  26. 26.Tadese S, Soromessa T, Bekele T. Analysis of the Current and Future Prediction of Land Use/Land Cover Change UsingRemote Sensingand the CA-MarkovModel in Majang Forest Biosphere Reserves of Gambella. Southwestern Ethiopia Science World Journal. 2021;2021:18. DOI: 10.1155/2021/6685045
  27. 27.Nguyen HTT, Pham TA, Doan MT, Tran PTX. Land use/land cover change prediction using multi-temporal satellite imagery and multi-layer perceptron markov model. International Archives of Photogrammetry Remote Sensing and Spatial Information Sciences ISPRS Archives. 2020;54:99-105. DOI: 10.5194/isprs-archives-XLIV-3-W1-2020-99-2020
  28. 28.Faichia C, Tong Z, Zhang J, Liu X, Kazuva E, Ullah K, et al. Using RS Data-Based CA–Markov Model for Dynamic Simulation of Historical and Future LUCC in Vientiane, Laos. Sustainability. 2020;12:1-20. DOI: 10.3390/su12208410
  29. 29.Behera MD, Borate SN, Panda SN, Behera PR, Roy PS. Modelling and analyzing the watershed dynamics using Cellular Automata (CA)-Markov model - A geo-information based approach. Journal of Earth System Science. 2012;121:1011-1024. DOI: 10.1007/s12040-012-0207-5
  30. 30.Chen C-F et al. Multi-Decadal Mangrove Forest Change Detection and Prediction in Honduras, Central America, with Landsat Imagery and a Markov Chain Model. Remote Sensing. 2013;5:6408-6426. DOI: 10.3390/rs5126408
  31. 31.Yirsaw E, Wu W, Shi X, Temesgen H, Bekele B. Land Use/Land Cover change modeling and the prediction of subsequent changes in ecosystem service values in a coastal area of China, the Su-Xi-Chang region. Sustainability. 2017;9:1-17. DOI: 10.3390/su9071204
  32. 32.Pontius RG, Schneider LC. Land-cover change model validation by an ROC method for the Ipswich watershed, Massachusetts, USA Agriculture Ecosystem Environment. 2001;85:239–248, DOI: 10.1016/S0167-8809(01)00187-6
  33. 33.Pontius RG, Huffaker D, Denman K. Useful techniques of validation for spatially explicit land-change models. Ecological Modelling. 2004;179:445-461. DOI: 10.1016/j.ecolmodel.2004.05.010
  34. 34.ESA. ESA Quick user guide of the Land Cover State products in GTiff and NetCDF formats. UCL-Geomatics, Belgium: ESA CCI partnership. 2015. p. 2
  35. 35.European Space Agency. Land Cover CCI Product User Guide Version 2. Tech. Rep, UCL-Geomatics, Belgium: ESA CCI partnership. 2017. p. 105
  36. 36.Pontius RG Jr. Quantification error versus location error in comparison of categorical maps. Photogramm. Eng. Remote Sensing. 2000;66:1011-1016
  37. 37.Viera AJ, Garrett JM. Understanding interobserver agreement: The kappa statistic. Family Medicine. 2005;37:360-363
  38. 38.Quintero GV, Moreno RS, García MP, Guerrero FV, Alvarez CP, Alvarez AP. Detection and projection of forest changes by using the markov chain model and cellular automata. Sustainability. 2016;8:1-13
  39. 39.Mosammam HM, Nia JT, Khani H, Teymouri A, Kazemi M. Monitoring land use change and measuring urban srawl based on its spatial forms: The case of Qom city Egyptian Journal of Remote Sensing and Space Science. 2016;20:103-116. DOI: 10.1016/j.ejrs.2016.08.002
  40. 40.Nadoushan MA, Soffianian A, Alebrahim A. Predicting urban expansion in arak metropolitan area using two land change models. World Applied Sciences Journal. 2012;18:1124-1132
  41. 41.Jain S, Siddiqui A, Tiwari PS, Shashi M. Urban Growth Assessment using CA Markov Model: A case study of Dehradun City. In: 9th International Geographic Union. Delhi, India. 2016. p. 9
  42. 42.Berling S, Wu J. Modeling urban landscape dynamics: A case study in Phoenix USA. Urban Ecosystem. 2004;7:215-240. DOI: 10.1023/B:UECO.0000044037.23965.45

Written By

Charles Bwalya Chisanga, Chizumba C. Shepande and Edson Nkonde

Submitted: January 28th, 2022Reviewed: February 3rd, 2022Published: April 7th, 2022