基于动力统计模型重构的副热带高压中长期预报  

A MID-LONG RANGE FORECAST OF PACIFIC SUBTROPICAL HIGH BASED ON DYNAMIC STATISTIC MODEL RECONSTRUCTION

作　　者：洪梅[1] 张韧[1] 何金海[2] 

机构地区：[1]中国科学院大气物理研究所大气科学和地球流体力学数值模拟国家重点实验室(LASG) [2]南京信息工程大学江苏省气象灾害重点实验室KLME,南京210044

出　　处：《气象学报》2006年第6期780-789,共10页Acta Meteorologica Sinica

基　　金：国家自然科学基金项目(40375019);江苏省气象灾害重点实验室开放课题(KLME0507);LASG开放课题

摘　　要：基于T106数值预报产品500 hPa位势高度场序列,用EOF方法对位势场序列进行时、空分解,在考虑相邻时段位势场空间模态基本稳定的前提下,引入动力系统重构的思想,以EOF分解的空间模态时间系数序列作为动力统计模型变量,用遗传算法全局搜索和并行计算优势,进行了模型参数的优化反演,建立了EOF分解的时间系数的非线性预报模型。通过模型积分和EOF的时、空重构,实现了副热带高压活动场的中长期预报。试验结果表明,在短期预报(1—5 d)该模型与T106模式的预报效果大致相当;但对于中长期预报(≥5 d),其预报效果优于T106预报产品;特别是10 d以上的预报结果对副热带高压的形态和范围仍然能较为准确地描述。该研究方法对副热带高压等复杂天气系统和要素场预报提供了一种新的思路。Based on the 500 hPa potential height field series of T106 numerical forecast products, by EOF time-space separation, and on the hypothesis of the spatial modes of EOFs being relatively stable and changeless in two successive short time periods, the time coefficients series of EOFs was taken as dynamical statistic mode variable, the dynamic system reconstruction idea and genetic algorithm was introduced to optimize the dynamical model parameters, and then a non-linear dynamic statistic prediction model of the temporal coefficients series of EOF modes was established. By the time integral of the model and the time-space reconstruction of EOFs, a mediumlong range forecast of Western Pacific subtropical high was carried out. The results of the forecast experiments of the dynamic statistical model show that both the accuracies of short range forecasts （≤5 days） by the dynamic statistical model and the T106 numerical forecast model is approximately equivalent, but for the medium-long range forecasts（≥5 days）, the dynamic statistical model is superior to the T106 numerical forecast model, especially the former is obviously better in the above 10-day prediction of the shape and spatial range of the subtropical high. The research methodology provides a new method and idea for the diagnosing and forecasting of weather systems as such complicated as the subtropical high, and shows a better application perspective.

关 键 词：动力统计 模型重构 遗传算法 EOF 副热带高压预报 

分 类 号：P46[天文地球—大气科学及气象学]