基于ECMWF细网格模式的伊宁市大雾预报方法研究

Research on Fog Forecast Method of Yining City Based on ECMWF Fine Grid Model

摘要: 利用2016~2020年逐日伊宁站雾观测数据、地面常规气象资料，对大雾天气进行分级统计，采用统计法分析伊宁市大雾的时间变化特征，通过相关分析法找出与大雾生消有密切关系的气象因子；用ECMWF细网格模式资料，利用等级分类和逐步回归法建立大雾预报模型。结果表明：1) 伊宁市大雾天气集中出现的月份是11月至次年2月，多发时段是21:00至次日12:00。2) 与大雾相关性高的因子为T₈₅₀、T₉₂₅、T_925-2m和T_2m及T_d；地面相对湿度 < 87%、露点温度差T-T_d > 4℃和地面风速 > 3.2 m/s不宜出现大雾天气。3) 用建立的伊宁市24 h大雾预报的多元回归方程，对2021年1月~3月大雾过程进行预报效果检验，0~2级雾的预报正确率 > 66%，预报正确率远高于空报率，3级的正确率为0、空报率为100%，级别越低正确率越高；各级雾的漏报率均 < 34%。检验效果较理想，具有较好的预报水平和较高的业务应用价值。

Abstract: Based on the daily fog observation data and ground routine meteorological data of Yining station from 2016 to 2020, the classification statistics of heavy fog weather were carried out. The temporal variation characteristics of heavy fog in Yining were analyzed by statistical method, and the meteorological factors closely related to the occurrence and elimination of heavy fog were found by correlation analysis method. Using ECMWF fine grid model data, the fog prediction model is established by hierarchical classification and stepwise regression method. The results showed that: 1) The fog weather in Yining City concentrated from November to February of the following year, and the frequent occurrence period was from 21:00 to 12:00 of the next day. 2) The factors highly correlated with fog were T₈₅₀, T₉₂₅, T_925-2m, T_2m and T_d; Ground relative humidity < 87%, dew point temperature difference T-T_d > 4˚C and surface wind speed > 3.2 m/s were not suitable for foggy weather. 3) The multiple regression equation of 24-hour heavy fog forecast in Yining City was used to test the prediction effect of heavy fog from January to March in 2021. The prediction accuracy of class 0 to class 2 fog forecast is more than 66%, and the accuracy is much higher than the empty forecast rate. The accuracy of class 3 fog forecast is 0 and the absent forecast rate is 100%, and the lower the class is, the higher the accuracy is. The missing rate of fog at all classes was less than 34%. The test effect is ideal, and it has better prediction level and higher operational application value.