FIRST TO FOURTH ORDER ANGSTROM-PRESCOTT-TYPE MODELS AND THE PREDICTION OF GLOBAL SOLAR IRRADIANCE OVER NIGERIA

Abstract

Daily re-analyzed data of global solar radiation and sunshine hour for ten years 2004-2013 were obtained from Era-Interim dataset at European Centre for Medium-Range Weather Forecast (ECMWF) based at Shinfield Park, Reading, United Kingdom for twenty stations across the four basic climatic regions in Nigeria. The daily clearness index and relative sunshine hour were obtained for each station, region and the entire Nigeria. The distributions of global solar radiation with sunshine hour in all the twenty stations across the four regions showed that the combination of these two variables in a linear equation can be used to evaluate the surface data of global solar radiation in all the stations. The regional distributions of clearness index with the relative sunshine hour for each station, region, and entire country showed that the highest value was recorded in January and the least in September. This also establishes the proof of the two distinct seasons across Nigeria. As revealed in the distributions between the global solar radiation and sunshine hour, a very strong correlation exist between the two meteorological variables. The generally approved and tested Angstrom-Prescott-type bivariate linear regression was used in this work. The Angstrom-Prescott-type regression is given below; = a + where H is the monthly average daily global solar radiation on a horizontal surface, is the monthly average daily extraterrestrial solar radiation, S is the monthly average daily numbers of hours of bright sunshine, is the monthly average daily maximum number of hours of possible sunshine (or day length in hours), is the clearness index, is the relative sunshine hour, while a and b are regression constants to be determined. The second, third and fourth Order Angstrom-Prescott-type linear regression models were later developed, using the ordinary least square (OLS) method, and used in this work. The first seven years (2004-2010) daily data were used to develop the models and the last three years (2011-2013) daily data were used for validation. Results showed that the regression models are all statistically significant at all the stations and regions. The P-value was zero (0.000) for all. Accuracy tests carried out using the coefficient of determination (), the mean bias error (MBE) and the root mean square error (RMSE) also confirmed that all the models, except the fourth Order are significant in an increasing order from the first to the third. It was also observed that value decreases from the southern (Coastal region) part towards the northern (Sahel Savannah region) part of Nigeria. Conclusively, the Angstrom-Prescott-type models are good models for predicting global solar radiation with the third Order having the best performance.