Time-series InSAR monitoring and evolution predictive analysis of filling-area subsidence in Yan’an New District

Since the completion of large-scale filling in 2015, Yan’an New District (YAND) has experienced notable ground deformation during rapid urban development. To reveal the spatial–temporal evolution of subsidence, this study integrates SBAS-InSAR monitoring with a newly developed Extended Long Short-Term Memory (xLSTM) prediction model. Using 153 Sentinel-1A images from 2020 to 2025, SBAS-InSAR provided continuous five-year ground deformation results, indicating maximum cumulative uplift of 57.7 mm and maximum cumulative subsidence of 141.1 mm, with peak subsidence rates reaching 26.29 mm/a. Three major subsidence belts were identified, mainly distributed in filled construction areas, with seasonally accelerated subsidence occurring in summer and autumn. The core innovation of this study lies in application in ground deformation of the xLSTM model, which effectively overcomes the temporal lag and limited long-term dependency issues inherent in conventional LSTM models, thereby improving subsidence trend prediction accuracy and responsiveness to evolving deformation patterns. The integration of historical deformation signals and xLSTM prediction results suggests that ground deformation in YAND is gradually entering a stabilization stage. This research provides a robust and scalable framework for high-precision urban subsidence monitoring and early-warning, offering valuable references for land development and risk management in rapidly expanding loess-region cities.