Error correction on summer model precipitation of China based on the singular value decomposition

Abstract

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.