Time-series InSAR ground deformation prediction based on LSTM model
In order to analyze the deformation status and development trend along the Changjiang River,and to maintain flood control safety and river regime stability,SBAS-InSAR technology was utilized to monitor ground deformation with 61 Sentinel-1A images covering the Changjiang River riparian area of Nanjing reach from March 2017 to March 2022.Additionally,the long short-term memory neural network model(LSTM)was used to predict the future trend of feature points.The results revealed that:① Compared with the leveling monitoring results,the accuracy of SBAS-InSAR was verified.The average annual ground de-formation rate in the study area ranged from-31～19 mm/a,and four subsidence funnels were formed along the Changjiang Riv-er in Nanjing reach.② The deformation prediction values of LSTM model exhibited a high degree of consistency with the SBAS-InSAR results,with a maximum absolute error of 3.28 mm.Using LSTM to predict the subsidence trend of feature points in the study area,it is found that the overall trend in the next 2 years will involve slow subsidence and a tendency to stabilize.The results can provide technical reference for relevant departments formulating protection and planning plans for the Changjiang River ripari-an area.
ground settlementground deformation predictionSBAS-InSARLSTMNanjing CityChangjiang River Basin
万方数据
维普
