滑坡危险性评估是山区进行防灾减灾工作的重要步骤,但以往研究多是对现阶段危险性的分析,用于区域尺度上滑坡危险性预测的有效方法较少.针对该问题,首次提出一种考虑物理模型与有效降雨入渗的新型框架用于对未来给定降雨重现期条件下的滑坡危险性进行预测.以青藏高原山南市雅鲁藏布江流域地区为例,首先使用Easy_Balance软件计算1998～2017年降雨的平均有效入渗率和前期有效降雨量Pa,并使用该期间夏季每月连续3日降雨最大值的95％分位数作为临界降雨量Pe,基于物理模型FSLAM完成土壤参数和土地利用参数的校准.使用Gumbel方法得到Pa和Pe在不同重现期时的极值,并分别计算各栅格破坏概率(probability of failure,PoF)对Pa和Pe的响应,根据栅格响应的敏感程度对研究区进行重分类后,使用危险性评估矩阵进行滑坡危险性预测.计算结果表明:当降雨重现期增大到50和100 a时,有92.1％的滑坡点落入到了极高和高危险区内,与传统的未考虑降雨对栅格PoF影响程度的TRIGRS模型相比,AUC精度高出7.6％～9.0％.当前方法能够有效预测未来极端降雨工况下的滑坡发生情况,且同时考虑了前期降雨入渗和极值降雨的影响,进一步提高了区域滑坡危险性预测的合理性.
A landslide hazard prediction framework by considering physically-based model and effective rainfall infiltration——A case study from Yarlung Zangbo River of Shannan City in the Tibetan Plateau
Landslide hazard assessment is an important step in disaster prevention and mitigation in mountainous areas.However,most of the previous studies focused on the risk analysis at the current stage,and there are very few effective methods for landslide risk prediction on a regional scale.Aiming at this problem,this study proposed a new framework by considering physical model and effective rainfall infiltration to predict the landslide hazard under the condition of a given rainfall return period in the future for the first time.Taking the Yarlung Zangbo River catchment in Shannan City of the Tibetan Plateau as an example,the average effective infiltration rate and the previous effective rainfall Pa from 1998 to 2017 were calculated by using the Easy_Balance software.The 95％quantile of the maximum rainfall for three consecutive days in each month in summer was used as the critical rainfall Pe,and the calibration of soil parameters and land use parameters was completed based on the physical model FSLAM.The extreme values of Pa and Pe at different return periods were obtained using the Gumbel method,and the response of probability of failure(PoF)of each grid to Pa and Pe was calculated.Then the study area was reclassified according to the sensitivity of the grid response,and the hazard assessment matrix was used to predict the landslide hazard in the whole area.The calculation results showed that when the rainfall return period reached 50 years and 100 years,92.1％of the landslide points fell into the very high and high hazard areas.Compared with the traditional TRIGRS model which didn't consider the effect extent of rainfall on grid PoF,the AUC accuracy of the proposed method increased by 7.6％-9.0％.The current method can effectively predict the occurrence of landslides under extreme rainfall conditions in the future,and the influence of rainfall infiltration and extreme rainfall are considered at the same time,which can further improve the rationality of regional landslide hazard prediction.
万方数据
维普
