Hydro-meteorological landslide triggering thresholds based on artificial neural networks using observed precipitation and ERA5-Land soil moisture

Landslide prediction is key for the development of early warning systems. In this work, we develop artificial neural networks (ANNs) that can identify landslide triggering conditions using soil moisture data in addition to precipitation. In particular, we use observed precipitation and ERA5-Land reanalysis soil moisture data at four different depth layers at the beginning and end of the precipitation events. Two different case studies, Sicily region (Italy), and a group of catchments in the Bergen area (Norway), are used to test the proposed approach against different climatic and geomorphological conditions. As a first step, traditional power law thresholds based on cumulative precipitation and duration (E-D) are derived by maximizing the true skill statistic (TSS) as a benchmark. For both study areas, ANNs using 87 different input combinations of precipitation characteristics and soil moisture data at multiple depth layers are analyzed. The developed ANN classifiers using soil moisture information in addition to precipitation outperform those using precipitation data only. Specifically, while power law E-D thresholds lead to a TSS maximum of 0.50 for both areas, the use of single-layer soil moisture yields a maximum TSS of 0.76 (0.78) for Sicily (Bergen area), while the use of multilayer soil moisture taken at both the start and the end of precipitation events yields a TSS = 0.79 (0.89). These results demonstrate that the proposed methodology is particularly promising for improving landslide prediction.