Integrated analytical hierarchy process and neural network approaches for assessment of soil erosion risk in Manjira...

Soil erosion poses critical threats to agricultural sustainability and food security in semi-arid regions, necessitating innovative assessment frameworks for effective sustainable land management. This study presents an integrated Analytical Hierarchy Process (AHP)-Artificial Neural Network (ANN) framework for enhanced soil erosion susceptibility mapping in the Manjira River Sub-basin, Maharashtra, India, addressing sustainable development challenges in agriculturally intensive landscapes. The methodology utilized ten environmental sustainability indicators derived from Shuttle Radar Topography Mission Digital Elevation Model (SRTM DEM), Sentinel-2 & Landset-8 imagery, and meteorological data to assess erosion risks across 10,160 km² of predominantly agricultural terrain. AHP analysis established factor importance through expert consultation, identifying slope (0.20) and rainfall (0.15) as dominant sustainability indicators, achieving satisfactory consistency (CR = 0.092). The novel integration employed AHP-derived susceptibility classifications as training targets for a multilayer perceptron neural network, representing a paradigm shift toward sustainable, data-driven environmental assessment. The integrated framework demonstrated significant improvements over traditional approaches, achieving 86.3% overall accuracy with F1-score of 0.88, providing enhanced reliability for evidence-based conservation planning. Spatial analysis revealed 41% of the basin exhibits high to very high erosion susceptibility, concentrated in agriculturally intensive western regions requiring immediate sustainable management interventions. The ANN enhancement refined classification precision by reducing moderate susceptibility areas from 44.33% to 39.91% while providing definitive risk designations crucial for targeted sustainable development strategies. This integrated approach successfully combines expert knowledge interpretability with advanced computational capabilities, offering a robust methodology for sustainable soil conservation planning in semi-arid agricultural environments. The framework provides practical applications for achieving sustainable development goals through improved land management decisions.