Research On Rainfall-induced Landslide Forecast Model Of GRAPES-LFM

China has suffered from the frequent landslide hazards, which have caused huge casualties and property loss each year. It is essential to carry out the research on the early warning of the rainfall-induced landslides for the geologic hazard weather forecasting and early warning. This study analyses the spatial and temporal distribution characteristics of rainfall-induced landslides in China, improves the models of the landslides hazard assessment, sets the threshold of the rainfall-induced landslides, establishes the regional rainfall-induced landslide predicting model GRAPES-LFM (GRAPES:Global/Regional Assimilation and PrEdiction System; LFM:Landslide Forecast Model) based on the numerical weather-forecasting model GRAPES (Global/Regional Assimilation and PrEdiction System) and the landslide predicting model TRIGRS (Transient Rainfall Infiltration and Grid-based Regional Slope-Stability Model). Main results and conclusions are summarized below.(1) Each year most of the landslides occur from May to September in China, which account for 90% of the total landslides of the whole year. The distribution of landslides is not uniform. For example, the landslides occur frequently in Sichuan and Chongqing because of the mountainous topographic environment and lots of rain. Taking the rainfall-induced landslides in Sichuan province in July,2013 for example, the spatial and temporal distribution of the landslides are consistent with the distribution of the rainfall.(2) The threshold of rainfall-induced landslide is calculated by using the SHALSTAB model (SHAllow Landslide STABility model), and the rainfall is predicted by GRAPES model. The rainfall threshold model is applied to forecast the landslide during a typhoon rainfall process in 2013. Because the soil is supposed to be saturated, the threshold is smaller than the observed rainfall which induced the landslides, and the predicting areas are larger than the areas where the landslides occurred actually.(3) In order to improve the accuracy of the landslide forecast, the rainfall-induced landslide forecast model GRAPES-LFM is established. The rainfall-triggered landslide disaster early warning model GRAPES-LFM is a physical deterministic model which couples the GRAPES model and the TRIGRS model. Results show that the observed landslide areas locate in the high risk areas.(4) To reduce the uncertainty of a single model, we propose the GRAPES-En-LFM model for landslide prediction, taking into consideration the uncertainty of the input parameters and rainfall predicting. Using cumulative distribution for each random variable (such as cohesion and friction angle) and a random number generator, a series of the parameter values are randomly generated. Meanwhile the rainfall is predicted by the ensemble predicteion model. The landslide prediction is probabilistic forecasting instead of deterministic forecsting by using the GRAPES-En-LFM model. Comparing with the operational landslide forecasting, the prediction result of the GRAPES-En-LFM model is more meticulous. The GRAPES-En-LFM model provides a new probability prediction method for landslide hazard.(5) This study proposes two probabilistic analysis methods to assess shallow landslide susceptibility by integrating the infinite slope model SHALSTAB model and TRIGRS model with Monte Carlo simulation, taking into consideration the inherent uncertainty and variability of the input parameters. The proposed approaches are applied to assess the landslide hazard in Dehua to evaluate the feasibility, and the assessment results are meticulous. The results show that the GRAPES-LFM model which is combined with the Monte Carlo-SHALSTAB model can be used to assess the landslide hazard.