海南岛空气质量概况及预报效果检验

Overview of Air Quality in Hainan Island and Verification of Forecasting Effectiveness

摘要: GRACEs模式是海南岛空气质量预报业务主要参考模式产品之一，为全面了解该模式对海南岛的预报效果，提高业务人员对模式产品的订正能力，使用海南岛18个市县的空气质量实况观测数据对GRACEs模式的预报产品进行预报效果检验。结果显示：(1) 2019~2021年，海南岛空气质量以优良为主，优良天气占比98.74%，良等级及以上天气中，首要污染物以O₃为主(88.72%)，其次为PM_2.5 (5.88%)和PM₁₀ (5.33%)；(2) GRACEs模式预报的PM_2.5和PM₁₀ (平均误差)和RMSE (均方根误差)最低，AQI和O₃误差水平较高，归一化偏差(NMB)显示O₃和AQI数值上均接近0，PM₁₀和PM_2.5预报值普遍高于实测值；(3) 空气质量为优等级时，空气质量等级TS评分最高，模式欠缺对良二级以上等级的预报能力；无首要污染物时，首要污染物预报TS评分最高，首要污染物为O₃、PM_2.5、PM₁₀时，TS评分较低。

Abstract: The GRACEs model was one of the main reference model products for air quality forecasting in HainanIsland. In order to comprehensively understand the forecasting effect of this model on Hainan Island and improve the correction ability of forecaster for model products. The air quality observation data of 18 stations in Hainan Island were used to verify the air quality forecasted of GRACEs model. The results showed as follows: (1) During 2019~2021, the air quality of Hainan Island was mainly good, with the proportion of 98.74%. In the second-class weather and above, the main pollutant was O₃ (88.72%), followed by PM_2.5 (5.88%) and PM₁₀ (5.33%). (2) The MB (mean error) and RMSE (root mean square error) of PM_2.5 and PM₁₀’s concentration were the lowest, followed by AQI and O₃. The Normalized Mean Bias (NMB) showed that O₃ and AQI are close to 0, and the forecast results of PM_2.5 and PM₁₀ are generally higher than the observation. (3) When the air quality belongs to the level of excellent, the TS was the highest, however, the model lacked the ability to predict levels above the second class. When there was no primary pollutant, the TS of the primary pollutant forecast was the highest, while the TS became lower when the primary pollutant is O₃, PM_2.5 or PM₁₀.