基于EOF分解的中期平均气温距平客观预报方法

本文通过将500 hPa高度、850 hPa温度、海平面气压的预报与历史要素场进行EOF分解,并假定EOF分解空间函数(即特征向量)是基本稳定的,建立以EOF分解主分量系列的时间系数、预报站点附近基本要素值的时间序列与预报站点平均气温距平的多元线性逐步回归预报方程,结合ECMWF集合预报数值产品,生成全国范围未来10 d平均气温距平客观预报产品投入业务应用,并采用同号率方法检验。结果表明,客观预报产品在检验时段内,同号率平均值为0.77,明显高于ECMWF集合预报模式2 m温度直接计算得出的同类产品同号率0.63;该客观预报产品不仅在分布范围上,同时在距平幅度上预报效果均较好。此外,相关分析也证明同号率方法在检验中期时效气温距平预报中有其合理性。

Objective forecast of medium-term average temperature anomalies based on EOF technique

Based on the EOF decomposing of 500 hPa height, 850 hPa temperature, sea level pressure field forecasts with historical elements, and assuming that the EOF spatial functions (feature vectors) are basically stable, the time function based on EOF principal components series, multiple linear regression prediction equation of time sequence of basic element value and the average temperature anomaly near and at the forecasting and prediction sites were established, moreover, combining with the ECMWF ensemble prediction products, the objective forecasting products with nationwide medium term average temperature anomaly were generated and applied into work, which was verified by using the same symbol ratio. Results show that the objective forecasting products in the test period has 0.77 symbol ratio, significantly higher than the same symbol ratio (0.63) of the similar products directly calculated by ECMWF 2 m temperature products; the objective forecasting products perform well not only in the distribution scope, but in the anomaly amplitude effects. In addition, the correlation analysis also proved that the same symbol ratio method is reasonable in the medium-term temperature anomaly forecast verification.