Temperature-based model for monthly average hourly global solar radiation for the Caribbean island of Trinidad

A  temperature-based empirical model, whose only input predictor is hourly temperature, predicts monthly average hourly global solar radiation on a horizontal surface. The model was evaluated and compared to five well-established semi-empirical models due to Whillier/Liu and Jordan, Collares-Pereira and Rabl, Gueymard, Newell, and Jain. Advantageously, the temperature-based model is
independent of global solar radiation input unlike the other models.  All models were trained from a five-year dataset and validated against an independent five-year dataset as well as against the five individual years therein. This allowed for long-term and short-term comparative performances of all the models, respectively.  For an average year, the best performing models were the temperature-based and Gueymard models, which yielded root mean square errors of 0.12 and 0.10 MJ/m²/h, respectively.  The same two models outperformed the others over the five individual years, with root mean square errors of 0.17 and 0.14 MJ/m²/h, respectively.  The new temperature-based model's simplicity, robustness and predictive abilities that are on par with or better than its well-established competitors make it a desirable candidate for predicting monthly average hourly global solar radiation.