摘要
因其独特的地理位置与气候,重庆市暴雨灾害频发,对其开展暴雨灾害风险评估与区划十分必要。本文利用暴雨过程强度影响因素、孕灾环境影响因素与承灾体暴露度等数据,结合专家打分得到的指标权重获得致灾危险性与承灾体受灾风险性指数,以此构建样本集。基于随机森林(Random Forest, RF)、自适应提升(Adaptive Boosting, AdaBoost)、极端梯度提升(eXtreme Gradient Boosting, XGBoost)、梯度提升回归(Gradient Boosting Regression, GBR)、支持向量回归(Support Vector Regression, SVR)、线性回归(Linear Regression, LR)算法分别进行预测。结果显示,XGBoost算法以最低的平均相对误差值(Mean Relative Error, MRE)1.950%、均方根误差值(Root Mean Square Error, RMSE)0.028,以及最高的相关性值(R-squared,R2)0.896,成为最优算法(以暴雨致灾危险性预测结果为例)。在单场暴雨灾害风险评估中,在缺少暴雨过程持续天数数据的情况下,XGBoost算法仍为最优算法,其预测的MRE值与RMSE值分别为2.066%、0.030,R^(2)值为0.885,利用XGBoost算法在评估区划中划分的各等级受灾风险区域与实际受灾区域基本保持一致,表明XGBoost算法在缺少部分数据的情况下,仍能高效准确地进行评估。
Rainstorm disaster is one of the frequent natural disasters in China.Chongqing has suffered heavy losses due to its unique geographical location and climate.Risk assessment and zoning of rainstorm disaster can effectively prevent and control it.This paper uses the data of rainstorm process intensity influencing factors,disaster-inducing environmental influencing factors and exposure degree of disaster-affected bodies,combines the index weights obtained by expert scoring to obtain the disaster-causing risk and disaster-affected risk index of disaster-affected bodies,so as to construct the sample set.Random Forest(RF),AdaptiveBoosting(AdaBoost),eXtremeGradientBoosting(XGBoost),Gradient Boosting Regression(CBR),Support Vector Regression(SVR)and Linear Regression(LR)are used to predict respectively.Finally,the XGBoost algorithm with the lowest mean relative error(MRE)1.950%,root mean square error(RMSE)0.028,and the highest correlation(R-squared,R°)0.896 becomes the optimal algorithm(taking the prediction results of rainstorm disaster risk as an example).In the risk assessment of single rainstorm disaster,the XGBoost algorithm is still the optimal algorithm in the absence of continuous days data of rainstorm process.The MRE and RMSE of the prediction results are 2.066%and 0.030,and the R^(2) is 0.885.The disaster risk areas of each grade divided by XGBoost algorithm in the evaluation zoning are basically consistent with the actual disaster areas,indicating that the XGBoost algorithm can still be evaluated efficiently and accurately in the absence of some data.
作者
谢涛
余亮
周浩
秦文思
XIE Tao;YU Liang;ZHOU Hao;QIN Wensi(School of Remote Sensing and Geomatics Engineering,Nanjing University of Information Science&Technology,Nanjing 210044,China;Laboratory for Regional Oceanography and Numerical Modeling,Qingdao National Laboratory for Marine Science and Technology,Shandong Qingdao 266237,China;Technology Innovation Center for Integration Applications in Remote Sensing and Navigation,Ministry of Natural Resources,Nanjing 210044,China;Jiangsu Province Engineering Research Center of Collaborative Navigation/Positioning and Smart Application,Nanjing 210044,China;Chongqing Shete Institute of Meteorological Application,Chongqing 401147,China)
出处
《气象科学》
2024年第6期1140-1153,共14页
Journal of the Meteorological Sciences
基金
基于机器学习方法的重庆暴雨灾害风险评估技术研究(YWJSGG-202413)
高分专项航空观测系统科研项目外协子课题。
关键词
暴雨灾害风险评估
致灾危险性指数
受灾风险性指数
极端梯度提升算法
s rainstorm disaster risk assessment
disaster risk index
disaster risk index
eXtreme Gradient Boosting algorithm
