Integration of multi-layer perceptron neural network and cellular Automata-Markov chain approach for the prediction of land use land cover in land change modeler

Land use and land cover (LULC) significantly influence the hydrological cycle and various earth processes. Understanding these dynamics is essential for effectively managing environmental issues within river basins. The study focuses on a highly dynamic and flood-prone sub-basin of the Upper Krishna River, where major urban settlements and intensive agricultural activities are concentrated along the riverbanks. The uniqueness of this research comes from the selection of this hydrologically sensitive landscape, shaped by both natural processes and anthropogenic pressures, which presents a critical case for land use and land cover modeling. Utilizing high-resolution satellite data (10 m), combined with the advanced Multi-Layer Perceptron Neural Networks (MLPNN) and Cellular Automata-Markov Chain (CA-Markov) modeling techniques within TerrSet's Land Change Modeler (LCM), which is not only capable of generating spatial transitions and dynamic maps but also identifies the key contributors in gain and loss of various land use classes. We projected LULC scenarios for the mid-century (2049) and end-century (2099) using data from 2015 to 2020. Our model was validated against the actual LULC map from 2024 and showed a strong correlation (Kappa = 0.85). The results indicate significant urban growth along the riverbank and predict an increase in built-up area from 6.53 % in 2024 to 9.59 % in 2049 and further to 15 % by 2099 of the total geographical area. We observed consistent declines in forest cover, cropland, and barren land. These findings are valuable for future hydrological studies and provide important insights for policymakers to support sustainable urban planning and flood risk management.