Test Study on Flood Forecast by Merging Multi Precipitation Data  

作　　者：Yin Zhiyuan Shen Tieyuan Yang Fang 

机构地区：[1]Hubei Key Laboratory for Heavy Rain Monitoring and Warning Research, Institute of Heavy Rain, China Meteorological Administration, Wuhan 430205, China [2]Meteorological Information and Technology Support Center of Hubei Province, Wuhan 430074, China

出　　处：《Meteorological and Environmental Research》2018年第2期50-57,共8页气象与环境研究（英文版）

基　　金：Supported by the Science and Technology Development Key Fund of Hubei Provincial Meteorological Bureau(2015Z02)

摘　　要：Shuibuya control basin in upper reaches of Qingjiang River,Hubei Province was taken as the case. By combining grouping Z-I relation with ground meteorological rainfall station,rainfall estimation by radar was calibrated,and actual average surface rainfall in the basin was calculated.By combining genetic algorithm with neural network,the corrected AREM rainfall forecast model was established,to improve rainfall forecast accuracy by AREM. Finally,AREM rainfall forecast models before and after correction were input in Xin＇an River hydrologic model for flood forecast test. The results showed that the corrected AREM rainfall forecast model could significantly improve forecast accuracy of accumulative rainfall,and decrease range of average relative error was more than 60%. Hourly rainfall forecast accuracy was improved somewhat,but there was certain difference from actual situation. Average deterministic coefficient of AREM flood forest test before and after correction was improved from -32. 60% to 64. 38%,and relative error of flood peak decreased from 39. 00% to 25. 04%. The improved effect of deterministic coefficient was better than relative error of flood peak,and whole flood forecast accuracy was improved somewhat.Shuibuya control basin in upper reaches of Qingjiang River,Hubei Province was taken as the case. By combining grouping Z-I relation with ground meteorological rainfall station,rainfall estimation by radar was calibrated,and actual average surface rainfall in the basin was calculated.By combining genetic algorithm with neural network,the corrected AREM rainfall forecast model was established,to improve rainfall forecast accuracy by AREM. Finally,AREM rainfall forecast models before and after correction were input in Xin＇an River hydrologic model for flood forecast test. The results showed that the corrected AREM rainfall forecast model could significantly improve forecast accuracy of accumulative rainfall,and decrease range of average relative error was more than 60%. Hourly rainfall forecast accuracy was improved somewhat,but there was certain difference from actual situation. Average deterministic coefficient of AREM flood forest test before and after correction was improved from -32. 60% to 64. 38%,and relative error of flood peak decreased from 39. 00% to 25. 04%. The improved effect of deterministic coefficient was better than relative error of flood peak,and whole flood forecast accuracy was improved somewhat.

关 键 词：AREM quantitative rainfall forecast Radar quantitative rainfall estimation Genetic algorithm-neural network Flood forecast 

分 类 号：P338[天文地球—水文科学] P468.024[水利工程—水文学及水资源]