System for predicting areas of elevated geodynamic risk based on integrated analysis of geophysical fields

Abstract

Relevance. The territories of Southern Russia and the North Caucasus are characterized by a complex fault-block structure, pronounced seismicity, dissected terrain, and hazardous natural and natural-technogenic processes. Therefore, prediction of geodynamic risk zones requires integrated processing of heterogeneous geophysical, geological-structural, and geoinformation data. Aim of the study. The study aims to develop an approach for predicting areas of elevated geodynamic risk based on integrated analysis of geophysical fields and intelligent processing of spatial data. Research methods. Seismic event catalogues, active fault maps, gravity and magnetic anomalies, digital elevation models, remote sensing data, morphometric parameters, and GIS layers were used. Random Forest, gradient boosting, and convolutional neural network analysis of multichannel raster matrices were applied. The results were evaluated by predictor importance, spatial correspondence between calculated zones and known manifestations of geodynamic activity, and geological-geophysical interpretability of the maps. Results. The most significant factors were distance to active faults, seismic event density, horizontal gradient of the gravity field, slope angle, and lineament density; their contribution reached 63–71%. Zones of low, moderate, elevated, and high risk were identified. About 28% of the territory was assigned to elevated and high-risk classes, while correspondence with known manifestations of geodynamic activity reached about 76%. The approach is applicable to geodynamic zoning and selection of areas for detailed monitoring.