The Angstrom-Prescott (A-P) model is one of the most accurate and widely used empirical methods to estimate global solar irradiance (Rs). In order to use this highly rated model, the first step is to choose its coefficients from literature or to calibrate them in a new locality. This seemingly simple matter can be confusing and difficult due to their large variability and many different calibration approaches. This paper addressed the role of time-dependent nature of these coefficients by comparing the Rs predictions of the simultaneously calibrated monthly, seasonal. bi-annual and annual (fixed) coefficients using data from 20 sites of temperate climate in China. When evaluated by performance indicators of RMSE, RRMSE, MBE, R-2 and values of t-statistic, it is demonstrated that time-dependent A-P coefficients are not better than fixed ones in Rs prediction. This indicates that consideration of the time-dependent characteristics of these coefficients has little predictive value and thus unnecessary. The phenomenon can be interpreted by the concept of equifinality and the linearity inherent in the relationship of n/N vs. Rs/Ra. The implication of this study is that it gives more confidence in predictive accuracy of fixed coefficients and contributes to reducing unnecessary complexity of this model and thus easing their choice, calibration and application. (C) 2010 Elsevier Ltd. All rights reserved.