SASRAD: An hourly-timestep solar radiation database for South Africa

A methodology is presented for the correction filling of solar radiation data at sites within South Africa, with the aim of creating a continuous, hourly-timestep dataset for multiple locations. Data from twenty sites, collected by the Agricultural Research Council, are analysed with regard to the amount of data requiring offset or multiplier adjustment, as well as the amount of bad data. A range correction algorithm is implemented based on the 90th percentile (10% exceedance) hourly irradiance, as a function of site latitude elevation. The resulting, corrected data set is given the title: South African Solar Radiation Database (SASRAD). Comparisons are made with two other solar radiation datasets, the South African Atlas of Agrohydrology Climatology, a limited set of older historical data from the South African Weather Service (SAWS). Results indicate that the SASRAD dataset matches well with other datasets, with major discrepancies apparently due to problems with the other data sets, rather than the SASRAD data. The Coefficient of Multiple Determination (R²) between the Atlas SASRAD for monthly radiation is 0.927, the mean error between three of the SASRAD sites the corresponding SAWS data is 1.1 MJ m^-2 d^-1. The fraction of data requiring correction varied from 11% to 100%, depending on the site. The range correction algorithm was successful at correcting data that had been subject to incorrect calibration, did not remove annual trends in mean radiation levels.