Rainfall intensity-duration threshold of debris flows in the Réal Torrent from rain gauges and radar data: comparison and transposability

The Réal Torrent is a very active debris-flow catchment of the Southern French Alps, monitored since 2011. A debris-flow intensity-duration threshold was established in 2017 (Bel et al., 2017), based on rainfall data collected with rain gauges during the 2011-2014 period, where 33 debris-flow events were observed. Our study aimed to update this threshold with rain gauge data collected during the 2014-2023 period, where 51 additional events were observed, to evaluate the stability of the threshold. Secondly, we tested the influence of rainfall data source through the comparison of the rain-gauge-based threshold with a radar-based threshold, established by using the Météo-France ANTILOPE product, which combines radar estimates and rain-gauge observations of precipitation. Finally, we tested the transposability of the latter threshold by comparing it to triggering rainfall events of various regional debris flows recorded on 82 catchments, spread across the Southern French Alps during the 2011-2024 period. The detailed dataset of debris-flow events in the study region was obtained from the ONF-RTM natural hazard database (French National Forest Office service dedicated to natural hazards in mountain regions).

The update of the threshold with 9 additional years of debris-flow monitoring allowed us to conclude that an amount of roughly 30 debris-flow events is sufficient to establish a stable intensity-duration threshold on a single torrent. We also observed that the radar-based threshold is much lower than the rain-gauge-based threshold. Therefore, we showed that the source of rainfall data has a strong influence on the threshold equation. Finally, the analysis of the intensity and duration of the regional debris-flow triggering rainfall events relative to the Réal radar-based intensity-duration threshold led us to conclude that using a threshold on only one very active catchment is not transposable at the regional scale due to the high proportion of false positives induced.