摘要
应用华东区域大气环境数值预报业务系统对长江三角洲55个城市2019年逐日臭氧(O3)进行了数值预报,结果表明不同时效的预报效果非常稳定,预报与观测在相关系数、倍比等指标上体现出良好的一致性,绝大多数城市的相关系数高于0.7,但数值上存在系统性偏高.时间变化上的高度一致性决定了学习期为5~7d、截距为0的单因子动态适应学习型线性修正算法是适用于长江三角洲的O3数值预报释用方法.修正后O3预报效果得到显著改进,预报偏差从超过25μg/m3降至0.4μg/m^3;预报误差类指标减小30%以上,其中RMSE从42μg/m^3下降至27μg/m^3;除相关系数外,90%以上城市的其他指标得到明显改进;日最大8h滑动平均O3和日最大小时平均O3污染的预报临界成功指数(CSI)指数分别提高28%和17%;空报率和漏报率趋于平衡.修正方法的应用解决了数值预报系统性偏高的不足.
Daily ozone in 2019 over the 55 cities in the Yangtze River Delta(YRD)region was forecasted by using the operational Regional Atmospheric Environmental Modeling System for eastern China(RAEMS).The results showed that the performance was consistent in different forecast lengths of 1~4d.Good agreement of modelled ozone with the observed was found through indices of correlation coefficient(r),FACx,etc.Correlation coefficients over most cities were over 0.7.At the same time,forecast was generally overestimated.The high consistence in temporal trend suggested that the single factor dynamical learning linear method of with interception of zero and learning period of 5~7d was the most appropriate approach for the improvement of ozone forecast over YRD.The forecast performance was significantly improved.Mean bias decreased from over 25 to 0.4μg/m^3.Mean error,normalized mean error and root mean square error(RMSE)was 30%decreased,among which RMSE decreased from 42μg/m^3 to 27μg/m3.Performance indices except for r at over 90% cities were improved.For ozone pollution,CSI of O3-8h and O3-1h was 28% and 17%increased,respectively and the missed detection rate and false alarm rate was balanced.In general,the application of the improvement method resolved the systematical overestimation of ozone forecast over the YRD region.
作者
周广强
瞿元昊
余钟奇
ZHOU Guang-qiang;QU Yuan-hao;YU Zhong-qi(Yangtze River Delta Center for Environmental Meteorology Prediction and Warning,Shanghai 200030,China)
出处
《中国环境科学》
EI
CAS
CSCD
北大核心
2021年第1期28-36,共9页
China Environmental Science
基金
国家重点研发计划项目(2016YFC0201900)。
