Trends of the yearly global radiation E-g, annual sunshine hours S, yearly precipitation P and yearly averaged air temperature Tat 51 stations in 1961-1998 in China were detected, and the significance test method was the F-test. A total of 47 stations showed decreasing trend in the E-g series of which 35 passed the F-test at the 5% significant level. At 42 stations, the trend of S was also decreasing, and 36 of them passed the F-test at the 5% significant level. The P series did not change largely in 1961-1998 and at 37 stations showed a positive trend, while the positive trend or reduced trend at 39 stations did not pass the F-test at the 25% significant level. The positive trend of the T series was shown at 49 stations, of which 29 passed the F-test at the 1% significant level. In the yearly scale, the empirical relationship between E-g and S at any station was very low because the length of the E-g series was too short. When using all the data at the used 51 stations together to simulate E-g, the results were good. Using the longitude lambda, latitude phi and altitude H, or/and P and T, of the used stations to adjust the parameters a and b of the Angstrom model, respectively, or just to adjust the parameter a, the results will be better. The parameter b of the Angstrom model was little affected by the geographical position of the used stations. At last, a simple equation is recommended to use for simulation of the yearly global radiation in China. (c) 2006 Elsevier Ltd. All rights reserved.