#LIA MODELS TORRENT SERIES#

Calculation of the importance of the features indicated that 7 rainfall features are important in determining the prediction accuracy of the model. The results showed that the accuracy of the model using I-D features was 0.94, and the AUC was 0.938 while the accuracy of the model using 143 features was 0.98, and the AUC was 0.975. Using a machine learning method, we tested the accuracy of the model in predicting debris flows in the two cases of using rainfall intensity-duration (I-D) (2 features), and using all features (143 features).

#LIA MODELS TORRENT SERIES#

One hundred forty-three rainfall features were extracted from the rainfall data using tsfresh (a time series data processing Python package). Based on rainfall monitoring records for the Goulinping debris flow catchment in central China, 46 rainfall events triggering debris flows and 321 rainfall events not triggering debris flows were analyzed. Therefore, extracting more features from rainfall data is important for improving the prediction accuracy of debris flows. Statistical analysis of the relationship between rainfall characteristics and debris flow events is one of the principal methods of building a threshold model. The early warning of debris flows based on rainfall monitoring is an important method for reducing risk. It also highlights the significant influence of antecedent conditions and the seasonal fluctuations of sediment supply. This emphasizes the critical role played by short intense rainfall sequences that are only detectable using high time-resolution rainfall records. A logistic regression used to separate flow responses above the ID threshold, revealed that the best predictors are the 5-min maximum rainfall intensity, the 48-h antecedent rainfall, the rainfall amount and the number of days elapsed since the end of winter (used as a proxy of sediment supply). A systematic increase in the ID threshold with drainage area was also observed from the comparison of the three stations, as well as from the compilation of data from experimental debris-flow catchments. In the Réal Torrent, a 3-h MDRI appears to be representative of the local rainfall regime.

This is especially the case regarding the minimum duration of rain interruption (MDRI) between two distinct rainfall events.

The results show that parameters used to identify rainfall events significantly affect the ID threshold and are likely to explain part of the threshold variability reported in the literature. The classic method of rainfall intensity–duration (ID) threshold was used, and a specific emphasis was placed on the objective identification of rainfall events, as well as on the discrimination of flow responses observed above the ID threshold.

The study is supported by a 4-year record of flow responses and rainfall events, from three high-frequency monitoring stations equipped with geophones, flow stage sensors, digital cameras, and rain gauges measuring rainfall at 5-min intervals. This paper investigates the occurrence of debris flow due to rainfall forcing in the Réal Torrent, a very active debris flow-prone catchment in the Southern French Prealps.