Advanced CMD predictor screening approach coupled with cellular automata-artificial neural network algorithm for efficient land use-land cover change prediction

The prediction of future land use and land cover (LULC) change is essential for natural resource management and the development of environmental monitoring strategies especially for a non-resilient system. Thus, a novel LULC prediction model was proposed by coupling a hybrid predictor screening approach with the Cellular Automata coupled Artificial Neural Network (CA-ANN) approach. The hybrid predictor screener was formed by integrating the outcomes of Correlation Coefficient (CC), Mutual Information (MI), and Decision Tree (DT) methods and termed as the CMD technique. The CA-ANN model was trained by LULC maps of 1995 and 2005 and tested by the map of 2020 for these four sets of variables (i.e., CC, MI, DT, and CMD). The results revealed that the CMD predictors were the most influential variables for LULC prediction, followed by the DT-based variables. In contrast, the CC-based predictors were found to be the least effective, followed by the MI variables. Thus, the CMD-CA-ANN model was established as the benchmark model in this study. The study showed that elevation and distance from roads were more critical variables for LULC prediction compared to slope and aspect parameters. The proposed model demonstrated that by 2080, there would be an increase in the area of built-up land (5.79%), cropland (4.35%), shrubland (1.62%), and water bodies (0.3%) and a decrease in the area of forested land (1.34%) and barren land (10%) in the study area. The research has the potential to significantly contribute to the identification and management of significant environmental changes.