Timing landslide and flash flood events from radar satellite

Landslides and flash floods are geomorphic hazards (hereafter called GH) that often co-occur and interact. Such events generally occur very quickly, leading to catastrophic impacts. In this study we focus on the accurate estimation of the timing of GH events using satellite Synthetic Aperture Radar (SAR) remote sensing. More specifically, we focus on a tropical region, i.e. environments that are frequently cloud-covered and where space-based accurate characterization of the timing of GH events at a regional scale can only be achieved through the use of SAR given its cloud penetrating capabilities. In our multi-temporal change analysis method we investigated amplitude, spatial amplitude correlation and coherence time series of four recent large GH events of several hundreds of occurrences each covering various terrain conditions and containing combinations of landslides and flash floods within the western branch of the East African Rift located in tropical Africa. We identified changes that could be attributed to the occurrence of the GH events within the SAR time series and estimated GH even timing from it. We compared the SAR time series with vegetation and rainfall time series to better understand the environmental influence imposed by the variying terrain conditions. The Copernicus Sentinel 1 satellite is the key product used, which next to being open access, offers a dense, high resolution time series within our study area. The results show that SAR can provide valuable information for GH event timing detection. The most accurate GH event timing estimations were achieved using the coherence time series ranging from one day to a 1,5 month difference from the GH event occurrence, followed by the spatial amplitude correlation time series with one day to a 2,5 month difference. Amplitude time series were highly influenced by seasonality and proved to be insufficient for accurate GH event timing estimation. The results provide additional insight into the influence of seasonal vegetation and rainfall patterns for varying landscape conditions on the SAR time series. This research is one of the first to show the capabilities of SAR to constrain the timing of GH events with an accuracy much higher than what can be obtained from optical imagery in cloud-covered environments. These methodological results have the potential to be implemented in cloud-based computing platforms to help improve GH event detection tools at regional scales, and help to establish unprecedented GH event inventories in changing environments such as the East African Rift.