An ensemble neural network approach for space–time landslide predictive modelling

Abstract

There is an urgent need for accurate and effective Landslide Early Warning Systems (LEWS). Most LEWS are currently based on temporally-aggregated measures of rainfall derived from either in-situ measurements or satellite-based rainfall estimates. Relying on a summary metric of precipitation may not capture the complexity of the rainfall signal and its dynamics in space and time in triggering landslides. Here, we present a proof-of-concept for constructing a LEWS based on an integrated spatio-temporal modelling framework. Our proposed methodology builds upon a recent approach that uses a daily rainfall time series instead of the traditional scalar aggregation. Specifically, we partition the study area into slope units and use a Gated Recurrent Unit (GRU) to process satellite-derived rainfall time series and combine the output features with a second neural network (NN) tasked with capturing the effect of terrain characteristics. To assess if our approach enhances accuracy, we applied it in Vietnam and benchmarked it against a modelling counterpart where we replaced the rainfall time series with the corresponding scalar representative of the cumulated precipitation. The corresponding duration was set at 14 days as it proved to produce the best performance. Our results show that our protocol leads to better performance in hindcasting landslides when making use of the rainfall as a continuous signal over time. While not tested here, our approach can be extended to rainfall obtained from weather forecasts, potentially leading to actual landslide forecasts.

Keywords

• Space-time modelling
• Deep learning
• Gated recurrent units
• Landslide early warning systems
• Vietnam