Spatiotemporal prediction of land subsidence by integrating multi-source data

Land subsidence is a geologic phenomenon characterized by the gradual lowering of ground elevation, primarily caused by the over-extraction of groundwater. This study utilizes hydrogeological, engineering geological data, and small-aperture radar interferometry (InSAR) to establish a three-dimensional flow-solid coupling numerical model integrated the BP neural network for land subsidence in the Weibei Plain. Based on consolidation theory and the BP neural network, the model reconstructs the regional land subsidence development process, analyzes the evolution patterns and trends of subsidence, and provides valuable recommendations for regional groundwater development plans. The study results indicate that land subsidence mainly occurs in Guangling Township, Yangzi Street, and Xiaying Town, with the subsidence amount increasing annually. It is recommended to designate the northern and northeastern parts of Guangling Township as subsidence reduction zones. Considering the numerical simulation of land subsidence caused by the over-extraction of groundwater for agricultural and domestic use in the Weibei Plain, the groundwater extraction in the northern part of Guangling Township can be reduced by 30%, and by 50% in the northeastern region.