Deep learning forecast of rainfall-induced shallow landslides.

Rainfall triggered landslides occur in all mountain ranges posing threats to people and the environment. Given the projected climate changes, the risk posed by landslides is expected to increase, and the ability to anticipate their occurrence is key for effective risk reduction. Empirical thresholds and physically-based models are used to anticipate the short-term occurrence of rainfall-induced shallow landslides. But, evidence suggests that they may not be effective for operational forecasting over large areas. We propose a deep-learning based strategy to link rainfall to landslide occurrence. We inform and test the system with rainfall and landslide data available for the last 20 years in Italy. Our results indicate that it is possible to anticipate effectively the occurrence of rainfall-induced landslides over large areas, and that their location and timing are controlled primarily by the precipitation, opening to the possibility of operational landslide forecasting based on rainfall measurements and quantitative meteorological forecasts.

Implications of climate change on landslide hazard in Central Italy.

The relation between climate change and its potential effects on the stability of slopes remains an open issue. For rainfall induced landslides, the point consists in determining the effects of the projected changes in the duration and amounts of rainfall that can initiate slope failures. We investigated the relationship between fine-scale climate projections obtained by downscaling and the expected modifications in landslide occurrence in Central Italy. We used rainfall measurements taken by 56 rain gauges in the 9-year period 2003-2011, and the RainFARM technique to generate downscaled synthetic rainfall fields from regional climate model projections for the 14-year calibration period 2002-2015, and for the 40-year projection period 2010-2049. Using a specific algorithm, we extracted a number of rainfall events, i.e. rainfall periods separated by dry periods of no or negligible amount of rain, from the measured and the synthetic rainfall series. Then, we used the selected rainfall events to forcethe Transient Rainfall Infiltration and Grid-Based Regional Slope-Stability Model TRIGRS v. 2.1. We analyzed the results in terms of variations (or lack of variations) in the rainfall thresholds for the possible initiation of landslides, in the probability distribution of landslide size (area), and in landslide hazard. Results showed that the downscaled rainfall fields obtained by RainFARM can be used to single out rainfall events, and to force the slope stability model. Results further showed that while the rainfall thresholds for landslide occurrence are expected to change in future scenarios, the probability distribution of landslide areas are not. We infer that landslide hazard in the study area is expected to change in response to the projected variations in the rainfall conditions. We expect our results to contribute to regional investigations of the expected impact of projected climate variations on slope stability conditions and on landslide hazards.