基于随机森林算法的芜湖市汛期降水预测模型  

A Precipitation Prediction Model for the Flood Season in Wuhu City Based on Random Forest Algorithm

作　　者：付伟 文想成 司红君 Fu Wei;Wen Xiangcheng;Si Hongjun(Wuhu Meteorological Bureau,Wuhu 241000,China;Anhui Province Key Laboratory of Atmospheric Science and Satellite Remote Sensing,Hefei 230031,China)

机构地区：[1]芜湖市气象局,安徽芜湖241000 [2]大气科学与卫星遥感安徽省重点实验室,安徽合肥230031

出　　处：《气象与减灾研究》2024年第4期279-286,共8页Meteorology and Disaster Reduction Research

基　　金：国家重点研发计划(编号:2023YFC3007703);安徽省气象局复盘总结专项(编号:kY202308).

摘　　要：使用芜湖市国家气象观测站逐月降水量数据,分析了芜湖市汛期降水的气候特征,并以汛期前期的大气环流和海温指数为自变量,借助随机森林算法构建了汛期降水预测模型。结果表明:芜湖市汛期平均降水量652.6 mm,占全年平均降水量53.1%,总体无显著变化趋势。模型成功生成了降水量距平百分率的分类预测规则,筛选出了3月北太平洋副高面积指数等5个关键气候指数。模型对建模变量拟合的定量和定性准确率分别为64.8%和83.1%,并通过了模型自检和对2023年汛期降水预测的检验,其预测规则可在汛期前对降水距平百分率进行分类预测,丰富了芜湖市汛期降水预测的方法,具有较好的可用性。Using monthly precipitation data from the National Meteorological Observatory in Wuhu City,the climate characteristics of precipitation during the flood season were analyzed.Atmospheric circulation and sea surface temperature indices from the National Climate Center during the early stage of the flood season as independent variables,a precipitation prediction model for the flood season was constructed using the random forest algorithm.The results showed that the average precipitation during the flood season in Wuhu City was 652.6 mm,accounting for 53.1%of the annual average precipitation,with no significant overall trend observed.The constructed model successfully generated classification prediction rules for precipitation anomaly percentages and identified five key climate indices,including the Pacific Subtropical High Area Index for March.The quantitative and qualitative accuracies of the model’s fitting results to the modeling variables were 64.8% and 83.1%,respectively,and it passed the model’s self-inspection and the validation tests of precipitation during the 2023 flood season.Its prediction rules can classify and forecast precipitation anomaly percentages before the flood season,enriching precipitation prediction methods for Wuhu City with strong practical applicability.

关 键 词：随机森林算法 预测模型 汛期降水量 芜湖市 

分 类 号：P466[天文地球—大气科学及气象学]