基于多特征BiLSTM模型的西宁市地表形变预测OA

SBAS-InSAR technology was employed to obtain the land surface deformation distribution in Xining city from October 2020 to November 2024.By integrating static features such as the spatial lo-cation of deformation points with dynamic climatic characteristics,a multi-feature-based BiLSTM pre-diction model was constructed to forecast deformation in areas with typical deformation characteris-tics.The main conclusions are as follows:1)The average annual deformation rate in the study area ranges from-48 mm/a to 30.2 mm/a,with a maximum cumulative deformation value reaching-187.2 mm.2)Compared with the LSTM model,the prediction accuracy of the multi-feature BiL-STM model in three typical deformation areas is improved by up to approximately 32.3％,and it can more accurately identify local details and abrupt change points.At last,the optimal model was used to predict the deformation trends in each typical deformation area for the next six months,providing a reasonable reference for surface deformation monitoring.