Self‐Entrainment Motion of a Slow‐Moving Landslide Inferred From Landsat‐8 Time Series

Abstract

In mountainous environments, slow‐moving landslides (velocities <100 m/year) are a major concern for local populations. Rainfall is often the dominant forcing, and often result in major changes in kinematics which can mask smaller signals related to internal forcings. We focus here on a major (>40 Mm3) slow‐moving landslide in the desert of southern Peru and take advantage of this arid environment to study the internal processes affecting landslide kinematics. We first estimate the ground displacement from time series analysis of Landsat‐8 images, spanning a 5.7‐year period. Systematic artifacts in the optical time series are shown to correlate with topography, as well as vary seasonally. We apply a novel procedure for correcting these artifacts, which significantly reduces noise in the resulting time series, thereby allowing us to precisely resolve landslide displacements. We find landslide velocities of up to 35 m/year, with complex nonlinear interannual pattern, including a period of rapid acceleration. We validate our optically derived time series using Global Navigation Satellite System field measurements and find uncertainties (root‐mean‐square error) on the moving mass of 1.12 to 1.55 m. Sudden acceleration of the landslide body after March 2016 may originate from a mass collapse due to retrogression of the headscarp. By coupling sparse 3‐D Global Navigation Satellite System measurements with dense 2‐D optical time series data, we show that the headscarp retrogression acts like a wedge, resulting in domino‐like tilting of the downward blocks, and accelerates basal sliding over 2 years. These observations reveal that the dynamics of this retrogressive landslide are predominantly controlled by sediment supply and that succession of retrogressive and advancing motions is a self‐entrainment process.