-
利用奇异值分解(SVD)方法对国家气候中心气候业务模式CGCM的预报结果进行订正, 进而得到改进后的预报结果. 通过对1983–2011年中国夏季降水的模式预报和实际降水值各年份之间的耦合场进行SVD分解, 选取模式预报和实况降水相关性最好的前3–7个模态, 对这5种不同模态个数的订正结果加以对比, 选出对中国区域订正效果最好的模态个数, 得到模式预报的订正结果. 对2004–2009年6年进行交叉回报试验, 以距平相关系数和均方根误差作为评判标准来检验回报结果, 得出2004–2009大部分年份取前5个模态作为当年订正的模态个数时,订正效果最好. 利用已有的2004–2009年的模式和实际降水资料进行检验, 证实在大部分年份5个模态订正效果最好, 并将结果与系统误差订正法的结果相比较, 表明SVD法对2004–2009年 6年的订正结果中, 有4年比系统误差法的订正效果好. 以2010年作为预测年, 以1983–2009年27年交叉回报检验的结果确定模态个数, 对预测结果加以分析, 显示该方法具有潜在的业务应用价值.In this paper, singular value decomposition (SVD) is used to correct the National Climate Center business model CGCM, thus improving the forecast results. By performing the SVD decomposition using field coupling between the summer rainfall patterns forecasted in 1983-2011 and actual rainfall values, by selecting the mode forecast and 3-7 modals that are best correlated with the live precipitation, and comparing these revised results of the different numbers of modal number, and by selecting the number of modals with the best correction effect in the China region, the revised results are obtained from the model predictions. The 2004-2009 cross hindcast experiments and the return results tested by using the anomaly correlation coefficient and the root mean square error as evaluation criterion show that the best corrections are obtained by taking the first five modals in 2004-2009 as the number of modal revised in the year. The examinations of 2004-2009 mode and actual precipitation data demonstrate the correction effect obtained by the five modals is indeed best. A comparison between the result and the systematic errors shows that the Dhamma revised results are better than the systematic errors. Taking the year of 2010 as the forecast year, using the results of the cross-examination of 27 years 1983-2009 to determine the number of modals, and analyzing the forecast results by the determined number of modals constitute a method that has a potential business value.
[1] [2] [3] [4] [5] [6] [7] [8] [9] [10] [11] [12] [13] [14] [15] [16] [17] [18] [19] [20] [21] [22] [23] [24] [25] [26] [27] [28] [29] [30] -
[1] [2] [3] [4] [5] [6] [7] [8] [9] [10] [11] [12] [13] [14] [15] [16] [17] [18] [19] [20] [21] [22] [23] [24] [25] [26] [27] [28] [29] [30]
计量
- 文章访问数:5873
- PDF下载量:703
- 被引次数:0
下载: