Reliability Analyses of Anomalies of NCEP/NCAR Reanalyzed Wind Speed and Surface Air Temperature in Climate Change Research in China

By means of varied statistical methods, such as normalized root mean square error (RMSE), correlation analysis, empirical orthogonal function (EOF) decomposition, etc., the reliability of the varied seasonal anomalies of NCEP/NCAR reanalyzed wind speed and surface air temperature (SAT) data frequently used in the climate change research in China is studied. Results show that RMSEs of meteorological variables are smaller in eastern China than in western China, i.e., the reliability of NCEP/NCAR reanalysis in eastern China is better than that in western China. This could be due to effects of the topography in the reanalysis model and the disposition of "dense-in-eastern-and-sparse-in-western" of meteorological stations in China.The RMSE of anomalies of reanalyzed wind speeds decreases with increasing height, further con rming the possible impact of topography on reliability of reanalysis. Results of correlation analysis inversely correspond to those of RMSE analysis, i.e., if the RMSE is larger, the correlation between reanalyzed and observed data is worse, and vice versa. It is found from comparing the EOF eigenvectors of anomaly of reanalyzed and observed data that if a meteorological variable has smaller RMSE, the spatial patterns of corresponding EOF eigenvectors of anomaly of reanalyzed and observed data are similar and their time coeffcients are significantly correlated, and vice versa. Therefore, the similarity of EOF modes and the consistency of their time coe cients can be used to objectively assess the reliability of the reanalysis. On the whole, the reliability of the reanalyzed wind speed is better in spring, summer, and autumn, but worse in winter; and for the reanalyzed SAT, it is the best in winter and the worst in summer.