多气候模式的全国月降水预测能力评价及偏差校正  

Evaluation of Monthly Precipitation Prediction Based on Climate Model and Bias Correction in China

作　　者：林广洪 朱碧莹 陈杰[2] 邱元霖 刘建华 陈华[2] LIN Guang-hong;ZHU Bi-ying;CHEN Jie;QIU Yuan-lin;LIU Jian-hua;CHEN Hua(Hydroelectric Power Plant of Tianshengqiao-I Hydropower Development Limited Liability Company,Xingyi 562400,Guizhou Province,China;State Key Laboratory of Water Resources Engineering and Management(Wuhan University),Wuhan 430072,Hubei Province,China)

机构地区：[1]天生桥一级水电开发有限责任公司水力发电厂,贵州兴义562400 [2]水资源工程与调度全国重点实验室(武汉大学),湖北武汉430072

出　　处：《中国农村水利水电》2023年第6期47-56,65,共11页China Rural Water and Hydropower

基　　金：国家自然科学基金面上项目(52079093);湖北省自然科学基金资助项目(2020CFA100)。

摘　　要：月降水预测对水资源配置和规划管理等具有重要意义,但其受到多种因素的影响,预测难度和不确定性均较大。为探究基于气候模式的月降水预测在中国区域的表现和偏差校正方法对预测能力的影响,以中国内地为研究区域,选取1981-2014年为研究时段,评价了九种气候模式(CFSv2,SEAS5,CanSips,GEMNEMO,CCSM4,GFDL,CanCM3,CanCM4,GEOSS2S)在不同预见期下对月降水的预测能力,采用聚类分析方法分析了气候模式的预测能力随预见期的变化规律,并采用线性偏差校正方法(Linear Scaling,LS)和分位数映射校正方法(Quantile Mapping,QM)对降水进行后处理,比较了两种偏差校正方法在验证期(2008-2014年)的校正效果。结果表明:(1)不同气候模式之间对降水的预测精度差异较大,对夏季降水的预测能力因预见期和预测区域而异,其中SEAS5模式在不同经纬度和不同预见期下的综合表现均最优,且其预测能力随预见期的延长变化稳定;(2)偏差校正对所有气候模式的降水预测均有明显的改进效果,两种偏差校正方法的效果相近,但经过LS方法校正后降水的平均绝对相对误差小于50%,总体上略优于QM方法,此外经过偏差校正后SEAS5模式的综合表现依然最优。研究结果揭示了SEAS5模式对中国内地月降水预测的优势和偏差校正方法对气候模式预测能力的提升作用,可为基于气候模式的降水预测应用提供参考。Monthly precipitation prediction is crucial for water resources allocation,planning and management.However,the prediction is influenced by various factors,thus it is extreamly difficult and uncertain.To investigate the performance of monthly precipitation predictions based on climate models in China and the influence of bias correction methods,this paper selects China’s mainland as the research area,adopts 1981 to 2014 as the research period,and evaluates the predictive capability of nine climate models(CFSv2,SEAS5,CanSips,GEMNEMO,CCSM4,GFDL,CanCM3,CanCM4,GEOSS2S)for monthly precipitation prediction in different forecasting periods.Clustering analysis is adopted to analyze the patterns of prediction ability of climate models with forecasting period.The linear scaling(LS)and quantile mapping(QM)bias correction methods are used for the post-processing of the predicted precipitation,and the effects of the two methods are compared in validation(2008 to 2014).The results show that the accuracy of precipitation prediction varies among climate models,and the capability to summer precipitation prediction depends on the forecasting periods and forecast regions.The SEAS5 climate model performs optimal in different regions and forecasting periods,and its forecasting capability stabilized with the extension of forecasting period.In terms of bias correction,the bias correction methods can significantly improve the prediction ability of precipitation for all of the climate models and the two methods exhibited similar performs,however,the mean absolute relative error of LS method is less than 50%,which indicates that the LS method is generally slightly outperforms QM method.In addition,the SEAS5 climate model still performs best after bias corrections.This paper reveals the superiority of the SEAS5 climate model for monthly precipitation prediction in China’s mainland and the enhancement of the prediction capability of climate models by the bias correction methods.This study provides a reference for the application of preci

关 键 词：月尺度降水预测 气候模式 偏差校正方法 预测能力 

分 类 号：TV125[水利工程—水文学及水资源]