Rainfall-induced landslide hazard analyses using spatiotemporal retrievals of soil moisture and geomorphologic data

Abstract

Rainfall-induced landslides threaten residential and civil infrastructure. As extreme rainfall events increase with climatological variability, so does the need to effectively monitor these occurrences. However, physical monitoring of landslide occurrence requires costly instrumentation over vast areas. Therefore, a means for large scale spatial monitoring is desired. This study conducts infinite slope stability analyses on known spatially distributed rainfall-induced shallow colluvial landslides. Infinite slope analyses were chosen due to applicability to the investigated shallow landslides. These analyses were investigated as functions of spatial geomorphologic and spatiotemporal soil moisture data. The underlying assumption of these analyses was that soil moisture would act as a hydro-mechanical precursor for rainfall-induced landslides. A majority of geomorphologic data for these analyses was obtained via web databases. Contrarily, it was observed that measurements of friction angle were not spatially available. To remedy this, an Artificial Neural Network (ANN) machine learning workflow was developed to yield these requisite measurements. For spatiotemporal soil moisture, the Land Information System (LIS) was utilized to conduct assimilation of NOAH 3.6 LSM and NASA SMAP L3SMP_E moisture estimates. The LIS workflow yielded soil moisture estimates at various depths and fine resolutions. With spatial geomorphologic and spatiotemporal soil moisture available, this study moved towards the associated stability analyses. These analyses were focused upon a region of Eastern Kentucky, USA, which experienced an extreme rainfall and subsequent landslide event. Through these analyses, a majority of occurred landslides were able to be detected in areas observed to experience increases in soil moisture. Therefore, this study confirmed the underlying assumption that soil moisture can serve as a hydro-mechanical precursor for rainfall-induced landslide occurrence.