基于水土耦合机制的流域滑坡预报研究

降水诱发型滑坡灾害涉及水文与土壤失稳过程,是个非常复杂的预报难题。本研究应用CRESLIDE(Coupled Routing and Excess Storage and Slope-Infiltration-Distributed Equilibrium)模型,研究基于水土耦合机制的流域滑坡预报。模型输入格点降水由基于流域逐小时加密雨量站的反距离权重法插值获取;基于GIS、DEM和遥感技术,提取滑坡流域下垫面信息;采用分布式水文模型CREST模拟预报出滑坡研究区域的流域水文过程,作为中间变量驱动滑坡模型SLIDE,实现降水诱发型滑坡的预报。选择陕南月河流域2012年7月3—5日降水诱发型滑坡过程进行模拟预报,验证模型的适用性。结果表明,基于CRESLIDE模型的流域水土耦合滑坡预报模型在滑坡时空预报上表现稳定;通过ROC曲线分析CRESLIDE模型取得的特异性(87.8%)和敏感性(52.9%)均较好;本研究基于流域水土耦合机制研究滑坡机理预报,耦合了流域水文过程与土力学过程,使得流域水文模拟更为合理,在滑坡的预报与早期预警中表现良好,对同类滑坡预报有一定的借鉴意义。

Landslides Forecasting Using a Physically Based, Coupled Hydrological Geotechnical Framework

Rainfall induced landslide disasters, which involve hydrological processes and geotechnical processes, are a complex prediction problem. In this study, a coupled hydrological geotechnical model called CRESLIDE (Coupled Routing and Excess Storage and SLope Infiltration Distributed Equilibrium) was applied for landslides forecasting in the Yuehe River Basin. The hourly gridded precipitation data as the input of the CRESLIDE model were calculated using the inverse distance weighted interpolation method from reported values by encryption rainfall stations of China Meteorological Administration (CMA). By utilizing GIS, DEM and RS technology, the characteristic information of the test basin was extracted. The CREST distributed hydrological model was applied for simulating hydrological processes and computing the key intermediate variables as forcings of the SLIDE model to forecast rainfall induced landslide events. We chose the Yuehe River Basin, located in the south of Shaanxi Province, as the test region for landslide forecasting. The results show that the CRESLIDE model has a generally good reliability to accurately predict occurrence of landslides (location and timing). Receiver Operating Characteristic (ROC) analysis indicated that the CRESLIDE model perform well with a high specificity (87.8%) and a reasonably good sensitivity (52.9%). Coupled hydrological geotechnical framework like the CRESLIDE model is based on physical processes and has a more realistic representation of hydrological processes, so this type of model is very useful for landslide prediction and early warning. This study provides valuable information and insight for similar studies in this field.