Global solar radiation predictions in Fiji Islands based on empirical models

Many difficulties associated with Global Solar Radiation (GSR)' variability in space and time and with the angular position in sky hemisphere hamper its accurate knowledge. Fiji Islands, with a tropical maritime climate without great extremes of heat or cold, have GSR measurement limited by a small number of stations due to the technical and financial restricts. Yet there are various options of available empirical models to select the most suitable one(s) for estimating GSR accurately using weather-related variables available for each location. This paper, therefore, focuses on evaluation of the performances of twenty (20) empirical models in predicting GSR between 1984 and 2018 in six (6) meteorological stations within Fiji Islands. Simulations were performed for 35-year period (1984-2018) and results were compared with the National Aeronautics and Space Administration (NASA)'s MERRA (Modern-Era Retrospective Analysis for Research and Applications) observational datasets after bias corrections. Model performance indices such as Mean Error (ME), Nash-Sutcliffe Efficiency coefficient (NSE), Percentage Mean Error (PME), Root Mean Square Error (RMSE) and correlation coefficient (r) were adopted. Obtained coefficients of determination from the regression analysis (0.415 <= R-2 <= 0.988; at 95% confidential level) performed for the models' calibration between 1984 and 2018 suggest that the models' potentials to simulate GSR varies across the stations and among the models. About 33% of the models in group I, which relate GSR to the sunshine duration, monthly average day length and/or extra-terrestrial radiation performed favourably well in predicting GSR while 75% of models in group II (with air temperature as additional predictor) and 100% in group III (humidity) performed very well. It is concluded that inclusion of air [GRAPHICS] temperature and humidity as two principal predictors alongside with sunshine duration, day length and extra-terrestrial radiation produce the best GSR prediction (RMSE <= 5 MJ m(-2)) over Fiji Islands. (C) 2022 The Authors. Published by Elsevier B.V. on behalf of Faculty of Engineering, Alexandria University

Automated summary

This study evaluates the performance of twenty empirical models in predicting Global Solar Radiation (GSR) in six meteorological stations within Fiji Islands. The results suggest that including air temperature and humidity as additional predictors, alongside with sunshine duration, day length, and extra-terrestrial radiation, produces the best GSR prediction over Fiji Islands.