Uncertainty and error propagation in the channelized debris flood forecasting

Recently the methods for building rainfall threshold were proposed based on the physical process aims to predict the debris flood occurrences. However, due to uncertainties in rainfall and water & soil supply processes in mountainous regions, significant uncertainty remains in the forecasting. This study evaluates the error propagation mechanisms of rainfall patterns, hydrological process, and soil mobilization under specific rainfall constraints using a typical small watershed in the Wenchuan earthquake area as a case study. By setting parameters for rainfall, models, and initial soil conditions based on critical conditions derived from actual monitoring, we observe an amplifying trend in errors throughout the threshold establishment process. From rainfall to water flow, after transforming rainfall patterns and selecting runoff model parameters, the maximum positive error is 0.10, while the maximum negative error is -0.18. Incorporating the uncertainty of D₅₀ into soil mobilization increases the maximum positive error to 1.16 and expands the maximum negative error to -0.43. Considering the uncertainty in the proportion of soil mobilization within the catchment during the threshold building process, the maximum positive error further increases to 1.72, while the maximum negative error rises to -0.48. It is evident that errors introduced by rainfall patterns and runoff model parameters are relatively minor compared to those caused by the uncertainty of D₅₀. Based on these findings, a probabilistic forecasting model is proposed, providing a scientific basis for debris flow forecasts.

Key words: Debris flood, runoff yield, soil mobilization, critical thresholds, probabilistic forecasting.