A new algorithm for landslide dynamic monitoring with high temporal resolution by Kalman filter integration of multiplatform time-series InSAR processing
Jialun Cai,Guoxiang Liu,Hongguo Jia,Bo Zhang,Renzhe Wu,Yin Fu,Wei Xiang,Wenfei Mao,Xiaowen Wang,Rui Zhang
DOI: https://doi.org/10.1016/j.jag.2022.102812
IF: 7.5
2022-06-01
International Journal of Applied Earth Observation and Geoinformation
Abstract:Free and open access to the earth observation synthetic aperture radar (SAR) satellite images has enabled the implementation of landslide time-series monitoring. However, with the restriction of the low revisit period of satellite, it is difficult to meet the requirements of landslide dynamic monitoring with high temporal resolution. Moreover, with the multiplication of SAR satellite platforms, a corresponding data integration algorithm is needed to obtain the complementary displacement information from every single observation orbit. In this work, an integrated algorithm for landslide multi-source displacement optimization estimation based on Kalman filter (KF) is proposed to improve the temporal resolution and realize the dynamic monitoring and prediction of landslide movement. Specifically, the landslide migration coordinate system is established firstly, and the interferometric SAR (InSAR) displacement results in line of sight (LOS) are projected to the downslope direction. Then, with the introduction of the acceleration variable into the process covariance matrix of prediction model and the observation noise variance weight determination, the downslope displacements of multiplatform InSAR observations are dynamically integrated into a unified time series by KF dynamic prediction and correction. So as to achieve the high temporal resolution monitoring of landslide. For validation purpose, the Baige landslide in Tibet, China is selected as the test area, and 55 Sentinel-1A ascending images, 42 Sentinel-1A descending images, and 10 ALOS-2 PALSAR-2 images collected over this area from May 2017 to April 2019 are used to estimate the high temporal time series. The temporal resolution of landslide monitoring is successfully improved from 12 days with a single orbit to the shortest 1 day and repeatedly 2–5 days with multiple platforms, and the prediction of subsequent displacement is also realized. Prospectively, with the continuous multiplication of the InSAR satellite platforms, this proposed algorithm can provide high temporal monitoring data for landslide and better assist relevant emergency response, which is necessary for the dynamic monitoring and early warning of landslide.
remote sensing