基于Logistic回归和多指标叠加的短时强降水预报模型

利用2001—2011年中国西北地区东部10个特征站地面常规资料和MICAPS系统特征参数资料,分别运用改进的二元Logistic回归法和综合多指标叠加法,通过短时强降水天气学概念模型识别入型、水汽条件消空、敏感物理参数诊断等方法逐级判别,建立了两种短时强降水预报模型,并运用模型试预报2012年该区域的短时强降水过程。结果表明:两种新建预报模型相比平均气候概率模型试预报效果都有明显提高,而且前者高于后者;其中二元Logistic回归模型试预报TS得分高达46.6%,综合多指标叠加模型试预报TS得分19.6%,而平均气候概率模型试预报TS得分仅9. 7%;除西南气流型两者预报效果相当外,不同概念模型下二元Logistic回归模型试预报效果均优于综合多指标叠加模型。

The forecast models of short-time heavy precipitation based on logistic regression and multi-index superposition

Based on the conventional data and the characteristic parameters of MICAPS from 10 stations in east of Northwest China from 2001 to 2011, using the improved binary logistic regression method and comprehensive multi-index superposition method, two short-time heavy precipitation prediction models were established through the step-wise discrimination by short-time heavy precipitation synoptics conceptual model recognition, water vapor conditions emptying, sensitive physical parameter diagnosis and other methods. The models were used to forecast the short-time heavy precipitation events in 2012 in the region. The results show that the forecast effects of the two new forecast models were both obviously improved, compared with the average climate probability model, and the former model is better than the latter. Specifically, the experimental forecast TS of the binary logistic regression model scored as high as 46.6%, and the experimental forecast TS of the comprehensive multi-index superposition model was 19.6%, while that of the average climate probability model was only 9.7%. Except that the forecast results of the two models for the southwest airflow are quite equivalent, the forecast effect of the binary logistic regression model under different conceptual models is better than that of the comprehensive multi-index superposition model.