Landslide deformation and temporal prediction of slope failure in Himalayan terrain using PSInSAR and Sentinel-1 data

Abstract

Landslides in hilly terrains are frequent phenomenon due to various triggering factors such as heavy rainfall, flash flood, slope instability, tectonic disturbances etc., and result in loss of human lives and damage of infrastructure and property. The continuous monitoring of slope instability and temporal prediction of slope failure is of paramount importance for assessing landslide-triggering factors and mitigating landslide hazards. This research presented a comprehensive methodology for time-series displacement measurement of a landslide-affected area prior to its actual failure and temporal prediction of slope failure in a hilly terrain using persistent scatterer SAR interferometry (PSInSAR) technique and Sentinel-1 data in an open source computational environment. On 30 July 2021, a massive landslide witnessed in Sirmaur district of Himachal Pradesh, India near Barwas and brought down around 180 m stretch of national highway (NH) 707 connecting Paonta Sahib with Shillai-Hatkoti. We used 52 single look complex (SLC) images in ascending and 41 SLC images in descending orbit with VV polarization, acquired for a period of almost two years (2019–2021) for time-series displacement measurement along radar line of sight (LOS), horizontal and vertical directions. The pre-failure displacement-time series derived from ascending and descending data stacks was critically analyzed to identify the reliable and accelerating persistent scatterer (PS) locations for predicting temporal failure of slope material and to understand kinematics of the study area. We used inverse velocity (INV) and modified inverse velocity (T-INV) methods at three different accelerating and reliable PS locations for temporal prediction of slope failure. The results show that the mean LOS velocity ranges from −25.49 to + 31.35 mm/y in ascending data and from −20.97 to + 24.16 mm/y in descending data, and horizontal and vertical components of motion vector vary from −30.23 to + 18.99 mm/y and –23.16 to 25.79 mm/y respectively in the study area. Displacement time-series at selected measurement points (PSs) depict the initiation of slope instability from 31 July 2020 and 3 July 2020 using ascending and descending stacks respectively, which are almost one year before the actual failure. INV method predicted temporal windows of slope failure between 27 July 2021 to 1 August 2021 for ascending stack and 24 July 2021 to 29 July 2021 using descending stack, while those of T-INV method lye between 28 July 2021 to 1 August 2021 for ascending data and between 22 July 2021 to 30 July 2021 for descending data. The deformation-time series and predicted days of slope failure are in good agreement at different reliable PS locations. Results in this study show that the approach used for time failure may evolve as an effective tool for landslide early warning systems.