Landslide hydro-meteorological thresholds in Rwanda

<p>For the development of regional landslide early warning systems, empirical-statistical thresholds are of crucial importance. The thresholds indicate the meteorological and hydrological conditions initiating landslides and are an affordable approach towards reducing people&#8217;s vulnerability to landslide hazards. This research defined different landslide hydro-meteorological thresholds in Rwanda and evaluated their predictive capabilities. We identified the key hydro-geological and meteorological parameters in landslide-prone hillslopes in Rwanda. Precipitation was identified as the most important trigger of landslides and is of course the source of other hydrological processes and stocks such as local and regional groundwater levels and soil moisture content that predispose the slope to near failure. We therefore defined the precipitation-based threshold using rainfall and antecedent precipitation index as a proxy for the soil wetness state. The landslide precipitation thresholds served as an important first step towards the development of early warning system in Rwanda but constrained by the resulting quite high rate of false alarms. To improve the landslide prediction capability, the regional groundwater level was integrated in the landslide hazard assessment thresholds as a proxy for the catchments storage, an important hydrological processes that predispose the slope to near failure. This approach significantly improved the landslide prediction capability with high rate of true alarms and low rate of false alarms. However, given the scarcity of on-site data, the coarse spatial distribution network of the hydrological and meteorological recording equipment, themselves recording data at point scale, we have tested the potential of satellite and hydrological model-derived information in landslide hazard assessment thresholds. The GPM-based IMERG satellite precipitation product was found as a good alternate source of rainfall data for landslide hazard assessment in Rwanda. The hydrological model-derived soil moisture time series broadly reproduce the most important trends of the in-situ soil moisture and proved to be useful for landslide hazard assessment. The hydro-meteorological threshold models that incorporate the antecedent soil moisture content at root zone and the cumulative 3day rainfall reveal promising results. Based on the overall comparison, the landslide hydro-meteorological thresholds that consider the pre-wetting conditions of the terrain using groundwater levels and soil moisture at the root zone improved the landslide prediction as compared to the exclusive use of the classical precipitation thresholds such as intensity-duration I-D and event-duration E-D. The hydrological processes are concluded to be the most important landslide predictors potentially useful for a robust landslide early warning system development in Rwanda.</p>