Russian Scientists Create Algorithm to Predict Soil Subsidence
Researchers at RUDN University have built a precision model that forecasts dangerous ground subsidence in arid regions with record accuracy.
Scientists from Russia’s Peoples’ Friendship University (RUDN) unveiled an innovative model for predicting soil subsidence in areas with arid climates. The project, which combines artificial intelligence with bio-inspired metaheuristic optimization, was announced by the Russian Ministry of Science and Higher Education.
The main problem the research targets is the risk subsidence poses to building foundations, roads, farmland, and water supply systems. The process is especially dangerous in regions where groundwater is heavily used. Traditional assessment methods often fail to account for the complex interaction of natural factors and human activity.
The new two-stage method employs the k-nearest neighbors (KNN) machine learning algorithm alongside metaheuristic approaches, including a custom algorithm. For training and validation, the team used Sentinel-1 satellite remote sensing data collected between 2014 and 2020. The analysis identified 215 subsidence zones and defined 17 key influencing factors, including geology, topography, well density, and vegetation cover.
The model achieved record accuracy of 95.7%. Its transparency and reliability make it a powerful tool for preventing both economic losses and environmental disasters.
