A Logistic Model Predicting Occurrence Probability of Debris Flow

Abstract

This paper presents a logistic model for predicting the occurrence probability of debris flows based on rainfall intensity and duration. The data from a total of 354 rainfall events were used to calibrate the model, among which 249 were triggering a debris flow while 105 were not. The model will be useful to the decision making of debris flow early warning in the future. That is, given the estimated occurrence probability = 70% subject to a combination of rainfall intensity and duration, there is a 30% probability that the early warning will be a false alarm. By contrast, if decision makers decide not to issue an early warning, then there is a 70% chance leading to a missed alarm. Subsequently, integrating the consequences of missed alarm and false alarm into the equation, the respective risks can be computed, based on which decision makers can make a more robust decision whether an early warning is needed or not by choosing the scenario with a lower risk.