Based on the linear relationship between solar radiation and sunshine duration, the Angstrom model is widely used to estimate solar radiation from routinely observed meteorological variables for energy exploitation. However, the relationship may have changed in quickly developing regions in the recent decades under global "dimming" and "brightening" context, with increasing aerosols due to industrial pollutions. Solar radiation stations under different climate conditions in six large cities in China are selected to test this hypothesis. Analysis of the related meteorological items shows that Guiyang has the lowest solar radiation with the average annual value of 10.5 MJm(-2)d(-1), while Lhasa on the Tibetan Plateau has the highest of 20.1 MJm(-2)d(-1). Both radiation and sunshine hours decreased from 1961 to 2010, but at different rates. A moving linear regression method is used to investigate the changes in the relationship between radiation and sunshine duration, the results indicate an abrupt change in the correlation coefficients in 1980-1990s, which can be attributed to the aerosol load resulting from air pollution caused by the industrial development in 1980s under China's Open Door Policy. The sky condition has been changing from clean to dirty, thus the relationship between solar radiation and duration changes in the 1980's and has recovered in the recent decades. This finding implies that it might not necessarily be right to use long data sets for model calibration. Further investigation confirms that the Angstrom model performs the best with higher NSE (nash-sutcliffe efficiency) of 0.914 and lower MAPE (mean absolute percentage error) and RMSE (root mean square error) values of 13.7 w/m(2) and 23.9 w/m(2) respectively, when calibrated with a 10-year data set. In contrast, the model performs worst when it is calibrated with a 40-year data set, with NSE, MAPE and RMSE values of 0.891, 15.1 w/m(2) and 25.3 w/m(2), respectively. Based on the findings of this research, a 10-year data set is recommended as the national standard for model calibration in rapidly developing regions of China. Further analogous investigations are needed in other industrial regions to make an international standard for Angstrom model calibration. (C) 2015 Elsevier Ltd. All rights reserved.