(1. Key Laboratory of Active Tectonics and Crustal Stability Assessment，Institute of Geomechanics，Chinese Academy of Geological Sciences，Beijing 100081，China；2. China Three Gorges Resettlement Office，Chengdu，Sichuan 610041，China)
Abstract：Abstract：Based on TS-InSAR(time series-interferometric synthetic aperture radar)，GPS and image offset measurement data，this paper employs inverse velocity model(INV) and slope model(SLO) to calculate landslide failure trend，discusses the existing questions about the combination of the three types of data and prediction models and explores a landslide monitoring and warning technique based on time-series deformation data measured by remote sensing. TS-InSAR monitoring of landslide in Xinmo Village，Maoxian County，Sichuan Province occurred on 24 June 2017，indicates that the main displacement acceleration started two months ahead before failure，and the prediction models can provide early warning two days in advance. Although the deformation accumulated over 1 meter and accelerated in part of the time for 18 months continuously monitored by 6 GPS monitoring stations and observed serious ground fissures at the Zuofangxi slope in Meigu River of Sichuan Province，the predicted and residual life plots of two models showed that this slope is gradually stable. Field investigation verified the judgment and the current state of this slope. The failure time of Baige landslide in the upper stream of the Jinsha River is predicted by INV and SLO methods using satellite imagery offset measure data and predicted accurately in the most rapidly deformed area. The results show that (1) INV and SLO combined with remote sensing time-series deformation monitoring can predict the development trend and even the time of landslide failure，which is an efficient and economic monitoring and forecasting method. (2) Both INV and SLO have their advantages in landslides prediction，the results of the SLO model are more conservative than that of the INV model. But in the process of slope deceleration，the INV model can indicate the trend to restore stability，and the SLO model is more likely to get an invalid forecast. The prediction results of the two models are complementary，and the two models are more reliable when used together than when used alone.
2020, Vol. 39 
