Uncertainties in local and regional mass movement prediction using rainfall

The prediction of debris flows is relevant because this type of natural hazard can pose a threat to humans and infrastructure. Debris-flow (and landslide) early warning systems often rely on rainfall intensity–duration (ID) thresholds. Multiple competing methods exist for the determination of such ID thresholds but have not been objectively and thoroughly compared at multiple scales, and a validation and uncertainty assessment is often missing in their formulation. As a consequence, updating, interpreting, generalizing and comparing rainfall thresholds is challenging. Here, we present the findings of Hirschberg et al. (2021), which focused on (i) uncertainties related to ID thresholds, (ii) differences in local compared to regional ID thresholds, and (iii) how prediction can potentially be improved using statistical learning algorithms. The findings are of interest for debris-flow (and landslide) early-warning developers.

We use a 17-year record of rainfall and 67 debris flows in a Swiss Alpine catchment (Illgraben) to determine ID thresholds and associated uncertainties as a function of record du- ration. This included comparing two methods for rainfall threshold definition based on linear regression and/or true-skill-statistic maximization. The main difference between these approaches and the well-known frequentist method is that non-triggering rainfall events were additionally considered for obtaining ID-threshold parameters. Depending on the method applied, the ID-threshold parameters and their uncertainties differed significantly. We found that 25 debris flows are sufficient to constrain uncertainties in ID-threshold parameters to ±30% for our study site. We further demonstrated the change in predictive performance of the two methods if a regional landslide data set with a regional rainfall product was used instead of a local one with local rainfall measurements. Hence, an important finding is that the ideal method for ID- threshold determination depends on the available landslide and rainfall data sets. Furthermore, for the local data set we tested if the ID-threshold performance can be increased by considering other rainfall properties (e.g. antecedent rainfall, maximum intensity) in a multivariate statistical learning algorithm based on decision trees (random forest). The highest predictive power was reached when the peak 30 min rainfall intensity was added to the ID variables, while no improvement was achieved by considering antecedent rainfall for debris-flow predictions in Illgraben. Although the increase in predictive performance with the random forest model over the classical ID threshold was small, such a framework could be valuable for future studies if more predictors are available from measured or modelled data.