A Risk Forecast Method for Southwest Road Damages Based on Precipitation

Landslides, debris-flows and other disasters along roads caused by precipitation occur frequently, becoming one of the most important factors of roads damages. Yunnan, Guizhou, Sichuan and Chongqing are especially prone to road damages. Based on the information of road damages, the corresponding precipitation data from January 2007 to July 2013 and 24 h precipitation forecast data from July 2012 to July 2013, probability forecast models are adopted to describe probabilistic relations between precipitation and road damages. First, precipitation factors of the day and over the past two, three, four, five, six, seven days and effective precipitation over the past 15 days are analyzed by the method of Kendall correlation, and precipitation of the day and the past effective precipitation are identified because of small correlation. Second, after the normality process to two factors, polynomial fitting, Fourier fitting and Gaussian fitting are applied to the frequency distribution of the disaster and two kinds of precipitation factor. According to the analysis of fitting correlation and the fitting error, Gaussian fitting method is selected to apply to the scattering distribution of precipitation and road damages. Finally, universal probability forecast models of road damages based on effective precipitation (Model Ⅰ) and comprehensive of the day and effective precipitation (Model Ⅱ) are established, and the fitting adjustable coefficients are 0.9108 and 0.8333, respectively. According to critical precipitation thresholds of two models, combining the grade of hazards risk and precipitation risk to road damages, two kinds of warning classification scheme based on precipitation are proposed. Two risk forecast models for road damages are developed. Risks of road damages are divided into five levels by probability of damage occurrence: Very small, small, medium, large and very large. Two risk forecast methods are tested, showing they are both applicable to describe the relation between precipitation and road damages, and have a high forecasting accuracy and strong reference value in disaster forecast. In comparison, two models have the same trend and results of Model Ⅱ are generally greater than Model Ⅰ in number. In the flood season and disaster-prone period, Model Ⅱ is more sensitive to subjective forecasts than Model Ⅰ.The risk forecasting systems of road damages are created for Southwest China based on two methods, and used in risk operation since the end of 2012 achieving good effects.