2018年11—12月北京市重污染天气过程数值预报能力评估

2018年11—12月北京市发生了4次以PM_2.5为首要污染物的重污染天气过程，为了分析数值模型对4次重污染过程的预报能力，将CMAQ模式提前1~7 d对北京市PM_2.5的小时预报结果与观测结果对比，分别从离散统计和分类统计2个方面评估CMAQ模式对4次重污染天气过程的预报效果，并简要分析了偏差产生的气象方面原因。结果表明：CMAQ模式提前1~6 d对重污染天气过程的预报显示出良好的性能，为日常业务预报提供了可借鉴的参考信息，可较好地预报出PM_2.5小时浓度变化趋势和浓度水平，离散统计结果显示提前1~4 d的预报结果好于提前5~7 d，相关系数r基本大于0.8，但有一定程度的低估趋势；分类统计结果显示不同预报时效预报准确率大于70%，探测准确率高于55%，部分时段可以达到80%~90%，对人工预报起到了良好的参考作用；输入的气象场的变化及其偏差对于重污染的起始时间、持续时间及清除时间有一定的影响，对相对湿度预报偏小和风速预报偏大是造成CMAQ模式低估的一个重要原因。

Evaluation of the Numerical Prediction Performance of Heavy Air Pollution Processes in Beijing During November-December 2018

From November to December,2018,there were four PM_2.5 heavy air pollution processes happened in Beijing.In order to analyze the prediction performance of the numerical model about heavy air pollution,in this paper,the hourly forecast results of PM_2.5 in Beijing output by CMAQ 1-7 days in advance were compared with the observation results.The prediction effects of CMAQ on four of these processes were evaluated by discrete statistics and categorical statistics respectively,and the meteorological reasons for the deviation were briefly analyzed.The results showed that:CMAQ model showed good performance in predicting heavy air pollution processes 1-6 days in advance,which could provide reference information for daily operation forecast,and it could better forecast the PM_2.5 hourly concentration change trend and the concentration level,discrete statistics showed that the forecast of 1-4 days in advance was better than that of 5-7 days in advance,the correlation coefficient was basicly greater than 0.7,but there was a certain degree of underestimation.Categorical statistics showed that the different aging prediction accuracy was higher than 70%.Probability of detection was above 55%,sometimes even reached 80%-90%,which played a good reference role in artificial prediction.The variation and deviation of the input meteorological field had a certain impact on the starting time,duration and clearance time of heavy pollution,and the small relative humidity forecast and the large wind speed forecast were important reasons for the underestimation of CMAQ model.