贵州省降雨型滑坡预报模型及检验

本文利用贵州省2010—2017年630次降雨型滑坡资料以及国家气象台站和区域自动气象站逐小时降水资料，分析了贵州省降雨型滑坡的时空分布特征。基于5个不同时效降水指标与滑坡累积发生概率的拟合曲线公式，得到不同概率对应的5个降水指标的降水阈值，建立贵州省降雨型滑坡预报模型,并利用2018年新增滑坡样本对预报模型进行检验。结果表明：贵州省降雨型滑坡高发期主要集中在5—7月，其中6月滑坡发生概率最高，约48.6 %；7月次之，约34.9 %。当R1h、R3h、R6h、R24h、R3d分别大于等于26.9 mm、48.9 mm、62.9 mm、79.5 mm、92.0 mm时，发布滑坡预报预警。在降雨型滑坡预报模型中，随着5个不同时效指标的降水量增大，降雨型滑坡累积发生概率变化呈现前后增加缓慢，中间快速增加的变化特征。模型检验结果表明贵州省降雨型滑坡预报模型有一定的准确性，且5个降水指标均具有一定的实用性，其中利用R3d开展降雨型滑坡预报准确率最好。

Analysis and Validation of Forecast Model for Rainfall-induced Landslides in Guizhou Province

Based on data from 630 cases of rainfall-induced landslides from 2010 to 2017 and hourly precipitation data from observation stations, this study analyzed the spatial and temporal characteristics of rainfall-induced landslides in Guizhou province. The formulas containing precipitation indexes in different time scales and cumulative probabilities of landslides by curve fitting are built, from which the thresholds for five precipitation indexes in different time scales with different probabilities are defined. The forecast model of rainfall-induced landslides is built and then further validated by using cases of rainfall-induced landslides in 2018. The results are as follows. Rainfall-induced landslides occur most frequently from May to July with the maximum frequency occurring in June about 48.6%, and the next highest in July about 34.9%. Issue rainfall-induced landslides warning when R1h, R3h, R6h, R24h and R3d greater than or equal to 26.9 mm, 48.9 mm, 62.9 mm, 79.5 mm, 92.0 mm respectively. In the forecast model, with the increasing of precipitation amount related to indexes in five time scales, the cumulative probabilities of rainfall-induced landslides show rapid increase in the middle while slow decrease in the beginning and the end. The validation results indicate the accuracy of model and the usage of five precipitation indexes with the most accuracy by utilizing R3d.