Monsoon-triggered landslide timings derived from Sentinel-1 reveal cloudburst triggering and suggest earthquake-induce hillslope weakening

Heavy rainfall events in mountainous areas can trigger thousands of destructive landslides, which pose a risk to people and infrastructure and significantly affect the landscape. In Nepal, monsoon rainfall triggers hundreds of landslides across the country every year and represent a large share of total sediment production. However, a persistent problem is that monsoon-induced landslides are mapped using multi-spectral satellite images, where cloud cover makes it impossible to constrain precisely landslide timing. This has hampered our understanding of how rainfall is driving landsliding during the monsoon and how the hillslope susceptibility to rainfall could be modulated after earthquakes.

Sentinel-1 SAR images offer a solution to this problem since SAR can be acquired through cloud cover, is sensitive to landslides, and the regular acquisition strategy of the satellite means that images are acquired every 12 days on two tracks globally. A newly developed method based on Sentinel-1 amplitude time series can be used to assign 12-day time windows for 30% of the landslides in an inventory (Burrows et al., NHESS, 2022).

We apply this method to optically-derived inventories of monsoon triggered landslides across Nepal from 2015-2019, obtaining timing information for hundreds of landslides during this period. We use this new landslide timing information alongside satellite rainfall data (GPM IMERG calibrated using rain gauges to better account for orographically-induced precipitation) to further our understanding of landslide triggering during the monsoon. We are able to identify spatio-temporal clusters of landslides that are concurrent with intense peaks in rainfall during the 2017 and 2019 monsoon seasons. This suggests that cloudburst events during the monsoon can drive a large share of the mass-wasting volume associated with a given year.

We also observe that, during the 2015 monsoon season, a large number of landslides failed earlier and after much less rainfall than in 2017-2019. This may reflect weakening of the hillslope following the M_w 7.8 Gorkha earthquake, which occurred around six weeks before the onset of the monsoon season in 2015, triggering co-seismic landslides across central and eastern Nepal and resulting in elevated numbers of landslides during the 2015 monsoon. Using the satellite rainfall data, we model the evolution of soil water content through time for every landslide. By using the modelled soil moisture at the time of failure in the Factor of Safety equation, we can obtain an estimate of cohesion for every landslide. By comparing cohesions for the 2015 dataset against those from 2017-2019, we suggest that the early landsliding in 2015 could be explained by a cohesion loss of 1-5 kPa.