Comparison between MRM simulations, CAMS and PVGIS databases with measured solar radiation components at the Methoni station, Greece

Abstract This study examines the performance of the estimated solar radiation components obtained via the Meteorological Radiation Model, satellite-based data sets (CAMS, PVGIS-CMSAF-SARAH) and reanalysis (PVGIS-ERA5) against ground measurements taken with the Sunshine-Pyranometer at Methoni station, Greece. MRM shows satisfactory simulations for the global solar irradiation (R 2  = 0.97, RMSE = 11.5%, MBE = −2.5%) at 15-min time-intervals, while for the diffuse larger biases are found (R 2  = 0.57, RMSE = 45%). Solar irradiation estimates via CAMS at 15-min intervals reveal RMSE values of 19.5%, 38% and 28% for the global, diffuse and direct radiations, respectively. Biases are progressively reduced for hourly, daily and monthly data sets. PVGIS databases simulate the global irradiance reasonably well (R 2  = 0.82–0.92), exhibiting high uncertainties for the diffuse (R 2  = 0.39–0.49) and direct (R 2  = 0.75–0.87), regarding instantaneous measurements. Simulations under clear-sky conditions of all components are found to be significantly improved, from both MRM, satellite-based retrievals and reanalysis. Overcast and partially cloudy skies result in large uncertainties, especially for the diffuse and direct irradiations, since the satellite sensors may detect clouds at time intervals of unobstructed Sun disk by clouds. In addition, broken bright clouds near to the Sun's disk may increase significantly the measured diffuse irradiance, leading to large biases in the simulations from both MRM and satellite databases.