基于I-D阈值的滑坡气象预警双指标模型

Double-index model of landslide meteorological warning based on the I-D threshold

确定降雨阈值对于管控降雨型滑坡灾害风险具有重要意义。以浙江衢州市98处降雨型滑坡为例进行降雨阈值研究。在统计分析1970—2019年梅汛期间诱发滑坡降雨特征的基础上,首先对比分析I-D、E-D和E-I曲线建立的降雨阈值模型,然后基于I-D阈值建立以当日降雨量R0和5 d有效降雨量R5为预警指标的滑坡气象预警双指标模型,最后采用历史降雨和滑坡数据检验模型的合理性和可靠性。结果表明:(1)I-D模型对于降雨型滑坡的预测能力优于E-D和E-I模型;(2)按双指标模型反演衢州市历史50 a滑坡气象预警情况,红色、橙色和黄色预警的年均预警次数为1.5,3.2,9.3次,具有较高的合理性;20处验证滑坡点被双指标模型准确预警,其中15处为红色和橙色预警,此外模型成功预警了2014年全部8处降雨型滑坡;(3)用于检验模型预测能力的7处滞后性滑坡全部成功预警,其中3处橙色预警、3处黄色预警和1处蓝色预警。本研究提出的预警模型有助于衢州市滑坡气象预警发布,为政府部门开展风险评价和管理提供新思路与方法。

[Objective]Determining the rainfall threshold is of great significance in controlling the risk of rainfall-induced landslides.[Methods]A total of 98 rainfall-induced landslides in Quzhou,Zhejiang Province,were used to study the rainfall threshold.Based on the statistical analysis of the rainfall data from 1970 to 2019,the rainfall threshold models established by the I-D,E-D and E-I curves are compared and analyzed.Subsequently,a landslide meteorological warning double-index model with daily rainfall R 0 and 5-day effective rainfall R 5 as warning indices was established based on the I-D model.Finally,the rationality and reliability of the model were confirmed using historical rainfall and landslide data.[Results]The results show that:(1)I-D is better than E-D and E-I in predicting rainfall-induced landslides;(2)The double-index model reasonably forecasts the average annual warning times of red,orange and yellow landslides in Quzhou,which are 1.5,3.2 and 9.3,respectively;20 landslides can be accurately predicted by the double-index model,of which 15 landslides are red and orange,and the model has successfully predicted all eight landslides in 2014;(3)The 7 hysteretic landslides used to test the prediction ability of the model all fell into the warning area,including 3 orange,3 yellow and 1 blue events.[Conclusion]The results proposed in this paper are helpful for the meteorological warning of landslides in Quzhou and provide new methods for government risk assessment and management.