Solar radiation data are essential for the work of energy planners, engineers, and agricultural scientists. However, most solar radiation recording stations measure only global radiation in China. It is, therefore, necessary to elaborate correlations between the rarely available diffuse radiation and other climatic data. In the present study, 10-year data (1995-2004) of daily global diffuse solar radiation and sunshine duration obtained at Beijing meteorological station of China was analyzed to guide future projects. Nine models correlating the diffuse fraction (K(d)) with both the clearness index (K(t)) and percentage possible sunshine (S/S(0)) and each variable separately were tested. The accuracy of the correlations is performed in terms of the two widely used statistical indicators, mean bias error, and root mean square error. The results indicate that the correlations relating K(d) with both K(t) and S/S(0) are more reliable than using each variable separately and Model 9 is preferred for its accuracy. The recorded data of Zhengzhou meteorological station are compared with the corresponding values predicted by Model 9 of Beijing. Model 9 provides predictions very close to the measured values of Zhengzhou. Therefore, it may be concluded that Model 9 can be used for estimating diffuse solar radiation for locations of semiwet region of the north of China where only global solar radiation and sunshine duration are available.