Modeling Spatial and Temporal Changes in Land-Uses and Land Cover of the Urmia Lake Basin Applying Cellular Automata and Markov Chain

Land use and land cover change are critical motivations for environmental changes. It mainly arises from human activities, e.g., the expansion of urban areas, the changes in agricultural land areas, and the destruction of water area which rooted in population growth. The present research used a combination of cellular automata (CA) and the Markov chain to simulate the present land-uses in the Lake Urmia Basin using remote sensing data. First, the land-use map was produced by the maximum likelihood classification method using the Landsat satellite imagery for the years 1998, 2008, and 2018. After the integrated CA-Markov approach assessed the model, the land-use maps were predicted for the years 2028 and 2038. The trend of land-use change between 1998 and 2018 revealed that agricultural areas and urban/human-made areas have increased by 3.31 and 2.74 percent, respectively, but water areas and other uses have decreased by 6.87 and 0.71 percent, respectively. The kappa coefficient was estimated at 80% for the model, implying its high accuracy in predicting land-uses. Based on the simulation results for 2028 and 2038, agricultural land area and urban/man-made areas will expand by 40.12 and 476.36% versus those in 1998 whereas water areas and other uses will shrink by 26.67 and 5.80%, respectively. The results can greatly help policymakers and managers of natural resources to make management decisions on land uses in different regions.
Extended Abstract
1-Introduction
Land use and land cover change (LULCC) has turned into a global issue that has aroused a grave concern of governments and natural resources officials. These changes are mainly caused by rapid population growth and are either causing the conversion of fertile lands into urban lands or destroying naturally occurring wetlands and lakes. Presently, remote sensing (RS) and geographic information system (GIS) have provided researchers with robust tools to explore natural ecosystems and manage them socio-economically (Gashaw et al., 2017). Many research works utilize past models, e.g., cellular automata (CA) modeling, to assess land-use dynamism and growth. The present study used a combination of CA and Markov chain (CA-Markov) to explore land-use changes between 1998, 2010, and 2018 to quantitatively and spatially predict the trend of these changes up to 2028 and 2038.
2-Materials and Methods
The Lake Urmia Basin (35°40'-38°30' N., 44°12'-47°54' E.) is one of the six main river basins of Iran. The present research used the Landsat imagery derived from the United States Geological Survey (USGS) website to estimate land-use changes. After the images were downloaded, they were subjected to atmospheric and radiometric corrections by the FLASH module in the ENVI 5.3 software package. All images were extracted for the case study in terms of geometric corrections in the UTM WGS84^[1] coordinate system and zone 38 north after mosaic forming. After the images were classified by the maximum likelihood method, the land-use map was obtained for the years 1998, 2009, and 2018 within four land-uses including agricultural lands, water areas, urban-human, and man-made regions, and other uses. Finally, the integrated CA-Markov model was employed to predict the map of four land uses for the years 2028 and 2038. The analysis of land-use change trends is crucial for future decision-making. Their changes were explored between 1998, 2008, 2018, 2028, and 2038 to examine the changes in the four land-use classes in the target years.
3-Results and Discussion
The analysis of the variation patterns in the study site indicated that the vegetation covers and build-up areas were increased versus 1998. In contrast, the water areas and other uses have declined in these three decades. Overall, agricultural lands, urban, and human-made regions have been expanded by 1558.39 and 2526.39 ha over the period 1998-2018 while the water areas and other land uses have decreased by 1960.87 and 2175.78 ha, respectively. It is found that the water areas have sharply decreased by 43.99% over the period 1998-2009. Still, the urban and human-made areas and agricultural lands were increased by 160.05% and 1.98% in this period, respectively. Data from the local meteorological station and satellite data showed that the Lake Urmia Basin had been affected by several droughts in this period (Rezaei Banafsheh et al., 2015; Kazempour et al., 2019). The comparison of the predicted map with the real map in 2018 (Figure 3) shows that they are almost similar. The kappa coefficient was >80% for the predictions made by the integrated CA-Markov model. Based on Aburas et al. (2016), Mosammam (2016), and Mansour et al. (2020), the kappa coefficient of >80% reflects the high simulation accuracy of a model. On the other hand, this is supported by the comparison of land-use areas so that the areas of agricultural lands, urban and man-made areas, water areas, and other uses were 17.47, 3.6, 13.41, and 65.51% of the study site in the real map while their areas were 22.38, 4.24, 92.43, and 59.13% of the study site in the predicted map, respectively. Since the model performed well in predicting land-uses in 2018, a transition probability matrix was applied to the period 2008-2018 to predict the land-use changes in the next 20 years up to 2038. The map of 2018 was used as the original map to predict changes in 2028 and similarly, the map of 2028 was used as the original map to make predictions for 2038. The predictions of the changes for 2028 and 2038 show that agricultural lands, urban and man-made areas, and water areas will expand, but the other uses will shrink. From 2028 to 2038, the most significant change will be a 9.48% increase in urban and human-made areas, followed by a 4.18% increase in agricultural lands and a 1.77% increase in water areas, whereas the other uses will change by -2.42%.
4-Conclusion
The present research sought to reveal land-use changes and spatially analyze them in the Lake Urmia Basin over the period 1998-2018 by combining remote sensing and geographic information system. Then, an integrated CA-Markov method was used to make predictions for the period ending in 2038. Research has shown that the integration of CA and Markov chain performs well in monitoring and simulating future land-uses in different regions so that its simulation accuracy has amounted to over 85% in some areas



[1] Universal Transverse Mercator (UTM) projection system datum World Geodetic System (WGS) 1984