This paper is the result of geological hazard survey engineering.

Objective Debris flow is a major geological disaster widely distributed in Bailong River Basin and often causes mass casualties. Accurate evaluation of the scale and risk of debris flow is the premise of the early warning and prediction of debris flow. Reasonable construction of risk prediction model is the key to debris flow disaster prevention and reduction.

Method Based on the reported debris flow cases and corresponding rainfall data in the study area, the quantitative evaluation model of potential debris flow risk in Bailong River Basin is constructed by analyzing the key geological environmental conditions and their relationship, and two kinds of critical discrimination models of debris flow risk level are proposed.

Results The results show that: (1) The dynamic quantitative calculation model of debris flow risk degree with debris flow activity scale, average gradient of gully bed, watershed cutting density and unstable gully bed ratio can be used as judgment factors to quickly and accurately predict the risk degree of debris flow under different engineering scenarios and rainfall frequencies in the future; (2) The topographic conditions affecting the occurrence of rainfall type debris flow are composed of watershed area, slope area ratio of 10 °-40 °, average longitudinal ratio of gully bed, etc; the rainfall conditions are mainly composed of 24 h cumulative rainfall before the outbreak of debris flow, 1 h rainfall or 10min rainfall triggering debris flow, etc; (3) Based on the risk calculation results of 30 typical debris flow gullies, the critical discrimination value of debris flow risk is obtained, and the 1h prediction model and 10min prediction model of potential debris flow risk of rainfall type are proposed. Among them, the prediction accuracy of debris flow above high risk of class Ⅰ model is greater than 87.5%, the prediction accuracy of debris flow above medium risk of class Ⅱ model is greater than 80%, and the verification accuracy of the two prediction models is 83.3%.

Conclusions The research results provide technical support for the accurate early warning and prediction of debris flow, and have a certain reference significance for the establishment of small and medium-sized debris flow real-time early warning system.