Estimating the average diffuse solar radiation based on multiple parameters: a case study of arid climate zone of Pakistan

In this study, 50 empirical models from 2 categories, diffuse fraction and diffusion coefficient, were defined with 4 subcategories to estimate monthly mean diffuse solar radiation over the arid climate zone selecting Karachi as a case study. Commonly used statistical performance indicators are analysed to check the accuracy of models. For diffuse fraction models, MAD, MARE, RMSE, RRMSE and R2 ranged from 0.0552 to 0.1789 kWh/m2-day, 0.0272 to 0.0807, 0.0702 to 0.2238 kWh/m2-day, 0.0026 to 0.0081 and 0.90 to 0.99, respectively. Similarly, for diffusion coefficient models, MAD, MARE, RMSE, RRMSE and R2 ranged from 0.0719 to 0.1618 kWh/m2-day, 0.0336 to 0.0724, 0.0847 to 0.2145 kWh/m2-day, 0.0031 to 0.0078 and 0.92 to 0.98, respectively. Results displayed that (1) diffuse fraction models have higher performance than diffusion coefficient models. (2) Sunshine hour fraction models outperformed clearness index, correlated models. (3) Combined models are more accurate than single variable-based models. (4) The proposed models showed better performance with models available in the literature.