Validating modelled, satellite based, and reanalysis based global horizontal irradiance data sets in South African cities

Abstract

The dissertation presents and compares the alternative methods that can be used to quantify high quality global horizontal irradiance (GHI) in areas with sparse in-situ radiometric measurements like South Africa. This was done by evaluating the performance of six GHI clear sky models, namely: Berger–Duffie, Simple Solis, Haurwitz, Ineichen–Perez, McClear and Bird by comparing 1-minute model outputs to relative clear sky reference 1-minute GHI in-situ data for sites in South Africa. This was also accomplished by calibrating and validating the annual regression coefficients, a and b, of the Ångström–Prescott (AP) model, which can be used to approximate daily GHI from in-situ sunshine hours. The validation of two hourly reanalysis-derived GHI datasets namely: Modern-Era Retrospective Analysis for Research and Applications (MERRA2), and Fifth generation European Center for Medium-Range Weather Forecasts atmospheric reanalysis (ERA5) and three satellite-derived GHI datasets namely Satellite Application Facility on Climate Monitoring (CMSAF SARAH), Copernicus Atmosphere Monitoring Service (CAMS) and SOLCAST for sites in South Africa was also done. Graphs, tables and statistical metrics namely relative Root Mean Square Error (rRMSE), relative Mean Bias Error (rMBE), relative Mean Absolute Error (rMAE) and the coefficient of determination (R2), were used to assess the capability of the datasets in estimating hourly GHI. The results showed that the Bird clear sky model was the overall best for sites in South Africa, with rRMSE of 4 %, rMBE of 2 % and R2 of 0.998. The AP model results for sites in South Africa were as follows: maximum rMBE was 4 %, rMAE was 9 %, rRMSE was 11 % and the worst correlation was 0.91. The hourly satellite and reanalysis-based results were as follows: A very good correlation for the satellite-derived and a poor correlation for the reanalysis-derived datasets. There was a consistent overestimation of GHI by both satellite and reanalysis-derived datasets. MERRA2 reanalysis-derived was the overall worst performing while SOLCAST satellite-derived was overall best performing. When considering only the freely available datasets, CAMS and CMSAF performed better. The AP model’s performance was similar to CMSAF and CAMS daily estimates. The results suggest that: The Bird clear sky model can be used with quantitative assurance for different applications in South African sites. The AP model coefficients determined can be used to precisely approximate daily GHI data at sites in South Africa where there is in-situ sunshine duration data. CMSAF and CAMS satellite-derived datasets are feasible freely available alternative GHI in South African sites.