Beyond Numerical Modeling: An Integrated Remote Sensing and AI Methodology for Rapid Slope Stability Analysis in Mining Regions

This research presents an integrated methodology for enhanced slope stability analysis in mining areas by merging remote sensing, artificial intelligence, and temporal deep learning. It advances beyond traditional numerical models by utilizing multi-source satellite data (Sentinel-1, Sentinel-2, DEM) to extract critical stability parameters—including slope angle, deformation, and rainfall intensity, among others within a multimodal geographic information system (GIS) environment. Current research focuses on generating a pixel-level risk-susceptibility map of the stability of mining slopes and classifying them into different risk zones — high, medium, and low by integrating fuzzy logic and multi-criteria decision-making (MCDM) techniques. Subsequently, the identified high-risk zones are processed to analyze temporal patterns and mine expansion/deformation in land from time-series imagery using a ConvoLSTM/U-Net deep learning model, thereby improving predictive capability for evolving slope geometries. The methodology has been validated through field surveys using drone imagery and laboratory tests of physico-mechanical properties on rock and dump samples. These support the interpretation of remote sensing–derived slope deformation and stability patterns. Ultimately, this research offers a cost-effective, scalable solution for predicting and monitoring OB dump slope stability by integrating remote sensing and AI, filling gaps left by traditional methods.