EOF分解与Kalman滤波相结合的副高位势场数值预报优化

用经验正交分解和Kalman滤波相结合的方法,建立了夏季500 hPa位势高度场数值预报误差修正模型,以改进副高数值预报效果,提高副高预报准确率.首先用经验正交分解（EOF）方法将T106数值预报500 hpa位势高度场分解为彼此正交的空间结构模态和相应的时间系数的线性组合,随后选取前15个模态的时间系数（其方差贡献98.7%）序列,分别建立了各自的Kalman（卡尔曼）滤波模型,最后用优化出的时间系数与相应的空间结构场进行EOF重构,进而得到修正后的副高位势预报场.修正后的位势场与原始的数值预报场的对比结果表明,该修正模型可对副高数值预报误差进行有效修正,优化后的预报效果较原始数值预报场有明显改进提高.

Potential height field optimization of numerical forecast products based on EOF and Kalman filter

Based on the method of empirical orthogonal function (EOF) and Kalman filter, a compositive prediction model of the five day 500 hPa height in summer was established to improve and promote the forecasting accuracy of subtropical high of T106. With this method, 500 hPa height was separated into functions based only on time and space, then the first 15 principal time coefficients were selected to establish the independent Kalman filter prediction models of the 15 time coefficients (square contribution 98. 7%), and the independent predicted results were integrated finally. The results show that under the premise of keeping the object's main characters, the errors of 500 hPa height system are reduced and decreased, the precision is evidently improved and promoted compared with those of T106 prediction.