Leveraging time series analysis of radar coherence and normalized difference vegetation index ratios to characterize pre-failure activity of the Mud Creek landslide, California

Abstract

Assessing landslide activity at large scales has historically been a challenging problem. Here, we present a different approach on radar coherence and normalized difference vegetation index (NDVI) analyses – metrics that are typically used to map landslides post-failure – and leverage a time series analysis to characterize the pre-failure activity of the Mud Creek landslide in California. Our method computes the ratio of mean interferometric coherence or NDVI on the unstable slope relative to that of the surrounding hillslope. This approach has the advantage that it eliminates the negative impacts of long temporal baselines that can interfere with the analysis of interferometric synthetic aperture (In- SAR) data, as well as interferences from atmospheric and environmental factors. We show that the coherence ratio of the Mud Creek landslide dropped by 50% when the slide began to accelerate 5 months prior to its catastrophic failure in 2017. Coincidentally, the NDVI ratio began a near-linear decline. A similar behavior is visible during an earlier acceleration of the landslide in 2016. This suggests that radar coherence and NDVI ratios may be useful for assessing landslide activity. Our study demonstrates that data from the ascending track provide the more reliable coherence ratios, despite being poorly suited to measure the slope’s precursory deformation. Combined, these insights suggest that this type of analysis may complement traditional InSAR analysis in useful ways and provide an opportunity to assess landslide activity at regional scales.