Ozone Forecasting Performance Evaluation and Error Source Analysis of GRACEs Model Under Different Synoptic Patterns in Guangdong

CHEN Jingyang; LI Tingyuan; DENG Tao; WENG Jiafeng; OUYANG Shanshan; LIN Zifeng

doi:10.16032/j.issn.1004-4965.2025.035

Abstract：Based on the observed data of air quality and meteorological elements,reanalysis data and CMA model forecast products of Guangdong during 2018-2020,along with an objective classification method,this study conducts a comprehensive evaluation of the ozone forecasting performance and error source analysis of the Guangzhou Regional Atmospheric Composition and Environment Forecasting System(GRACEs).The key findings are as follows:(1)The GRACEs system forecast the trend of O3_8h concentration relatively well,but systematically underestimated both the concentrations of O3_8h and its precursor NO2,with the NO2 bias being particularly pronounced.(2)Under the control of typhoon periphery combined with cold high ridge(TPR)and weak cold high ridge(HR),the average ozone concentration and the rate of cities over standard were highest,while the model's skill in forecasting O₃_8h was worst.The forecast deviation of NO2 concentration was an important reason for the O3_8h concentration bias,and the forecast deviation of boundary layer meteorological elements from CMA model might further lead to the underestimation of O3_8h concentration.(3)The GRACEs model had a high omission rate for ozone pollution.Compared with the overall dataset,a larger negative forecasting deviation of NO2 concentration appeared under the control of TPR,and the negative bias in 2 m temperature became more pronounced under the HR conditions.(4)As for the spatial distribution,GRACEs delivered superior O₃_8h forecasts of O3_8h concentration over the eastern and western wings of Guangdong.The distinct patterns of the negative NO2 concentration forecasting deviation by GRACEs and 2 m temperature by the CMA model were the important reasons for the distribution difference of O3_8h concentration forecast skill.