Интеграция наблюдения за Землей и моделирования CA-ANN для прогнозирования динамики LULC в водно-болотных экосистемах под воздействием городской среды

Integrating earth observation and CA-ANN modelling to predict LULC dynamics in wetland ecosystem under urban pressure

Urban wetlands and water bodies in medium-sized Indian cities are increasingly at risk from anthropogenic influences and rapid development. This study examines the Land Use Land Cover Change (LULCC) of the Chatra wetland in English Bazar Municipality (EBM) using maximum likelihood classification. To assess future LULC, the study employs multi-layer perceptron (MLP) Cellular Automata-Artificial Neural Network (CA-ANN) methods. Landsat data from 1990, 2000, 2010, and 2020 were analyzed, revealing declines in cropland, plantation, and waterbodies by 982.83, 505.24, and 388.30 hectares respectively, while built-up areas expanded by 1721.08 hectares. Accuracy was validated using Google Earth images, with Kappa coefficients of 0.83, 0.86, 0.91, and 0.95. The CA-ANN model predicted LULC for 2030, 2040, and 2050. Kappa index values were used to validate the future LULC scenarios by comparing anticipated outcomes to 2020 LULC, validated with an 84.74% correctness. By 2050, built-up areas will increase by 2344.78 hectares, while cropland and waterbodies will decrease by 1268.06 and 521.18 hectares. The study, which includes a field survey of residents near the wetland, highlights that wetland resources are crucial for livelihoods and the local economy. Our study also finds the fact and figure of the wetland resources as well as present scenario of the wetland vicinity. It reveals a trend of decreasing natural landscapes and increasing urbanization. If these patterns persist, the wetland will fail to serve dependent people and its ecology will suffer. The study aims to enhance land use management including the wetland and waterbodies in medium-sized cities like EBM, which are particularly vulnerable to growing urbanisation. The CA-ANN model may also predict future occurrences and understand spatiotemporal LULC dynamics.