Global solar radiation (Rs) is important in many areas, including agriculture, hydrology and meteorology. In this study, three non-calibrated models, five calibrated models and six multiple linear regression models (MLR) were evaluated in relation to measured Rs​​ based on the following statistical parameters: Mean Bias Error (MBE), Root Mean Square Error ), d (Willmott's coefficient), R² (coefficient of determination) and t-test. The average monthly data used for the study include maximum temperature, minimum temperature, mean relative humidity-RH, sunshine hours-n and Rs, from the city of Maputo dating from 1983 to 2006. The data from 1983 to 1999 were used for the calibration of the models based on the minimization of the residual sum of squares, and the remaining data were used for the test. The results showed that the MLR models which used insolation ratio (n/N) as one of the input variables performed better, especially the MLR3 models (MBE = 0.04 MJ m^-2 day^-1, RMSE = 1.47 MJ m^-2 day^-1, d = 0.97 and R² = 0.87) and MRL4 (MBE = -0.02 MJ m^-2 day^-1, RMSE = 1.44 MJ m^-2 day^-1, d = 0.97 and R² = 0.88). The MRL3 and MRL4 models require the following input variables: extra-terrestrial radiation (Ra), n/N and T; and, Ra, n/N, T and RH, respectively. Both models were not statistically different with measured Rs at 5%. The worst results were observed in the calibrated models.

Keywords: Calibration of models; Empirical models; Multiple linear regression; Solar radiation.