基于Random Forest的区域性泥石流的预警预测研究——以凤城市为例  

Regional Prediction of Debris Flow Based on Random Forest Regression Tree Model——Case Study in Fengcheng City

作　　者：付建飞[1] 门业凯[1] 侯根群[1] 赵纯福[1] 

机构地区：[1]东北大学资源与土木工程学院,辽宁沈阳110819

出　　处：《东北大学学报（自然科学版）》2012年第11期1641-1644,共4页Journal of Northeastern University(Natural Science)

基　　金：中央高校基本科研业务费专项资金资助项目(N100401006;N090601002)

摘　　要：以GIS为平台,选取16个预测因子(包括8个遥感因子、3个DEM因子、4个土壤因子和1个地层岩性),提取泥石流发生当日和前一日累计降水数据为响应因子,建立了random forest回归树模型,在区域尺度上对凤城市泥石流灾害进行预警预测.预测的结果表明:岩性、海拔和坡向是影响泥石流发生的主要因素,而持续的降水提高了土壤因子在地质灾害发生中的重要性,各遥感指数中,黏土因子对泥石流的影响最大.在东南部,只有当累计降水量较高时,泥石流才会发生;相反,在西北部,少量的累计降水就会造成泥石流的发生.另外,从1日降水预警图和2日降水预警图比较来看,持续降水容易增加泥石流发生机率.On the basis of 16 independent variables （including 8 remote sensing factors, 3 DEM factors, 4 soil factors and 1 formation lithology factor）, the intraday precipitation rainfall and that in preceding day were extracted as the dependent variable respectively, and the random forest regression tree model was built to predict the debris flows in the regional scale on the platform of GIS. The predicting results shows that, the lithology, elevation and aspect are the main factors influencing the occurrence of debris flow, and the increasing precipitation improved the importance of the soil factors in the occurrence of debris flow. Among the 8 remote sensing factors, the clay factor has the greatest impact on the debris flow. In the southeast, the debris flow would occurr when the accumulated precipitation goes higher. On the contrary, in the northwest, a small amount of accumulated precipitation would result in the occurrence of debris flow, and the continuous precipitation would increase the incidence of debris flow from the daily and two day precipitation warning charts.

关 键 词：泥石流 凤城市 RANDOM FOREST 预警预测 降水 

分 类 号：X141[环境科学与工程—环境科学]