Abstract: Precipitation is an important triggering factor of railway geological disasters. Every year significant economic losses are caused by railway geological disasters because of rainfall. To solve the problem of geological disaster forecasting in operational weather forecast service, a probability forecasting model is needed. Due to its special terrain and weather conditions, Yingxia Railway suffers from geological disasters more frequently and more severely. The disaster data from 2007 to 2012, as well as the temporal and spatial distribution features along Yingxia Railway are analyzed. Geological disasters happen most frequently at Qingzhou-Zhuozhai segment, especially from April to August.
    4 types of precipitation are the major trigger for the railway geological disasters: Local heavy precipitation, precipitation caused by typhoon, persistent precipitation and convectional weather. Geological disasters caused by typhoon are all relatively concentrated in Meishuikeng-Longhai-Xiamen segment. Persistent rainfall makes railway roadbed soil water saturation imbalance and thus slough or collapse may happen. Strong convective weather caused by rain could lead the soil flow to the air and thus causes the collapse of the shoulder. According to characteristics of different railway geological disasters caused by different types of precipitation, further study of the relationship between railway geological disasters and precipitation are carried out.
    10-min maximum precipitation, maximum hourly rainfall of a day, continuous rainfall and the cumulated rainfall of past 20 days are introduced as forecasting factors. Based on factor correlation analysis and logistic regression methods, the probabilistic forecasting models are established for each railway segment along Yingxia Railway. Although there are differences in precipitation hazard factor of each segment of geological disasters, the intraday precipitation is influencing for all segments. The precipitation one or two days before geological disasters plays an important role in probabilistic forecasting model. In order to verify the accuracy of this model, a test is applied on a heavy rainstorm happened from 20 May to 22 May in 2013 to forecast geological disasters of Yingxia Railway. The outcome indicates that the forecasting accuracy rates have reached above 80%. Effects of the probabilistic forecasting models are tested well. In the future, it can be used to conduct geological disaster forecasts to provide some technical support for railway safety meteorological services.