摘要
利用传统的EOF迭代方法,以梅汛期（6~7月）国家气候中心T63动力延伸模式200 hPa和500 hPa形势场产品为模式气候因子,全国961个站点的降水距平百分率为实况气候,建立了华东区月尺度的降水统计释用模型。在建模中运用交叉检验方法,解决了EOF迭代合理截断解释方差的选择问题,同时对释用建模中普遍存在的小样本拓展问题进行了探讨,并提出了有效的解决方案。由于在上述EOF迭代建模关键问题的研究中获得显著进展,使该模型在2007年和2008年梅汛期的释用中得到较好的效果。另外,还用Bootstrap方法给出了这些模型交叉检验平均释用技巧的置信区间估计,从而进一步揭示了这些模型的释用性能。
A traditional EOF method was engaged to establish the monthly rainfall statistical down- scaling models for the East China Meiyu season （from June to July） by taking the 200 hPa and 500 hPa heights derived from China National Climate Center T63 global model as model climate predictors, and taking the precipitation anomaly percentage at 916 observation stations of China as predictands, in which, cross-validation method was used to determine the reasonable truncated explaining variance of EOF iteration. In order to improve the drownscaling skill, a resolution to expand small sample which is a difficulty common in statistical downscale modeling, was then put forward, was then put forward. Such efforts directly resulted in the exciting predictions during Meiyu season from 2007 to 2008. Finally, the performance of the model was assessed and further its confidence interval of average skill scores was estimated by means of Bootstrap method.
出处
《高原气象》
CSCD
北大核心
2008年第B12期52-63,共12页
Plateau Meteorology
基金
上海市气象局2007年研究型业务专项"上海区域短期气候预测关键技术研发及业务平台完善"资助
关键词
EOF迭代技巧
降尺度方案
梅汛期
降水预测
EOF iteration skill
Downscale scheme
Meiyu-flood-season
Precipitation precdiction
