贵阳机场辐射雾天气特征及客观预报研究

Study on Weather Features and Objective Forecast of Radiation Fog at Guiyang Airport

摘要: 利用2010~2021年贵阳龙洞堡机场地面逐时观测数据、贵阳站探空以及再分析资料等，统计辐射雾生消规律和发生辐射雾的天气形势和特征，通过相关分析归纳对辐射雾造成的低能见度天气有高影响的高、低空物理量因子。将与能见度变化相关的相关要素输入神经网络进行训练，针对不同区间的能见度样本，利用BP神经网络方法分类训练了3个统计模型；并与WRF天气模式产品对接，采用分步筛选法，研发了龙洞堡机场24 h时效的逐时能见度预报产品。检验结果表明：基于BP神经网络方法开展逐时预报，预报结果与能见度实况日变化趋势和最低值接近，比本地WRF模式产品预报能力有较大提高，分级命中率平均提高了9%~35%，尤其对0.35~0.8 km，平均提高了30%~45%，对<0.8 km的低能见度过程技巧评分显著提高了39%，达到了73%的命中率。

Abstract: Based on the ground hourly observation data, sounding data and reanalysis data of Guiyang Long-dongbao Airport from 2010 to 2021, the generation and dissipation law of radiation fog and the weather situation and characteristics of radiation fog were statistically analyzed. The high and low altitude physical quantity factors which have high influence on low visibility weather caused by radiation fog are summarized through correlation analysis. The relevant elements related to visibility change are input into the neural network for training, and three statistical models are trained by BP neural network method for visibility samples in different intervals. By docking with WRF weather model products, the 24-hour hourly visibility prediction product of Longdongbao Airport was developed by using the step-by-step screening method. The test results show that the hourly forecast based on BP neural network method is close to the daily variation trend and the minimum value of visibility, which is greatly improved compared with the forecast ability of local WRF model products. The classification hit rate is increased by 9 - 35 % on average, especially for 0.35 - 0.8 km, which is increased by 30 - 45 % on average. The skill score of low visibility process < 0.8 km is sig-nificantly increased by 39 %, reaching 73 % hit rate.