Abstract

Hillslope stability problems experienced in Three Gorges Reservoir Area after the water impoundment in the reservoir have been widely studied in the literature. However, the contributions of morphometric, meteorological and seismic variables altering hillslope deformations and the variation in their roles over time is yet to be explored. This study aims at addressing this gap in the literature. To do so, we generate hillslope deformation time series using Interferometric synthetic aperture radar (InSAR) techniques and calculate mean line-of-sight velocities for eight time windows between March 2017 and April 2021. For each of these time windows, we build a random forest model and regress six variables (slope steepness, distance to river, total precipitation, snow cover, earthquake intensity and territorial water storage) against mean line-of-sight velocities. For each of these models, we quantify the variable importance. Our findings show that earthquakes and precipitation have the highest contribution to the surface deformations, across different time windows. Additionally, we run a suite of bivariate analyses to assess the contribution of reservoir water level, an informative layer whose variability is only measured across time. We show that hillslopes mainly close to the reservoir exhibit strong negative correlation between surface deformation and reservoir water level. Overall, our finding could be considered as a step towards developing a predictive tool to identify expected hillslope deformation in the future.