基于EC细网格产品在乌鲁木齐机场低能见度预测中的释用

利用2015—2018年乌鲁木齐机场航空例行天气报告(METAR报)、ECMWF(European Centre for Medium-Range Weather Forecasting)细网格数值预报产品对影响能见度的主要因子进行分析,提取与低能见度相关性高的物理量作为预报因子,采用SVM方法,分别基于Poly、RBF核函数建立乌鲁木齐机场未来21 h能见度预报模型。结果表明:(1)基于预报因子区间分类的SVM模型物理意义明确,试验结果较好;以RBF为函数建立的SVM模型(SVM-RBF)预报能力更好,其训练样本预测的TS评分0.84,准确率89.20%。(2)SVM-RBF模型的检验样本中,预报准确样本的预报误差整体偏小;在漏报样本中则有能见度越低、预报误差越大的特点,模型的振荡性明显。(3)结合NCEP/NCAR再分析资料研究SVM-RBF模型对天气过程的预报表现,发现模型对于特定天气形势下引发的低能见度天气,预报误差较小且预报提前量较大。

Low-visibility weather forecast combined with EC fine mesh product release based on support vector machine (SVM) of Urumqi Airport

Abstract: In this study, key factors, which influencing the visibility, were classified based on the distribution intervals by utilizing pertinent hourly measured data at Urumqi Airport and fine grid numerical forecasting data from ECMWF (European Centre for Medium-Range Weather Forecasting) from 2015 to 2018. According to the correlations to the visibility, some influencing factors that being high related to the visibility were chosen as the forecasting factors. Furthermore, 21-hour-ahead forecasting models on visibility at Urumqi Airport were established using the SVM method with Poly and RBF functions, respectively. Results showed that: (1) SVM models based on our method showed clear physical meaning and results of pertinent experiment performed good. SVM-RBF performed better than SVM-Poly with TS score of 0.84 and accuracy rate of 89.2%. (2) For the samples which predicted well, the deviation was small. By contrary, for the samples not well predicted, the deviation is large. (3) Using NCEP/NCAR reanalysis data to study the performance of SVM-RBF model, it was found that SVM-RBF showed satisfied model performance on low visibility caused by specific synoptic situation which with a small forecasting deviation and an acceptable advance of prediction.