Моніторинг та прогнозування відновлення агроресурсів методами Geo-Graph Intelligence

Abstract

This research focuses on the development of a hybrid intelligent system for agro-resource monitoring and recovery prediction based on a Geo-Graph Intelligence framework. The approach evolves from isolated models to a unified multi-modal architecture integrating Sentinel-1 SAR and Sentinel-2 optical data. Physics-Informed Machine Learning (PIML) paradigm, embedding mass transfer differential equations directly into the neural network’s loss function. By utilizing Graph Neural Networks (GNNs), the system accounts for terrain-driven pollutant migration and soil-specific diffusion. This methodology enables high-precision vegetation dynamics analysis and remote chemical soil reconnaissance, providing a robust foundation for strategic post-conflict land reclamation.