Bias correction of a novel European reanalysis data set for solar energy applications

Abstract

One of the major challenges during the transition phase of the energy system is to maintain the balance between energy supply and demand. Rising questions are often related to site mapping, variability, extremes and compensation effects for example. A fundamental source of information to answer these questions are high quality data sets of renewable energy related variables. As reanalyses provide all relevant data to assess wind and solar power generation over a long period of time (decades) in a gridded consistent way, they exhibit great potential in the field of renewable energy. A new regional reanalysis is COSMO-REA6, which covers the European domain over the years 1995–2014 with a horizontal resolution of about 6 km and a temporal resolution of 15 min. In this paper, we first assess the quality of the Global Horizontal Irradiance (GHI) provided by COSMO-REA6. High quality GHI measurements obtained through the Baseline Surface Radiation Network (BSRN) are used as reference and reveal systematic short comings in the reanalysis: (1) an underestimation of GHI in clear sky situations and (2) an overestimation of GHI in cloudy sky situations. In order to reduce these systematic regime dependent biases, a post-processing is developed. The applied post-processing method is a scaling based on orthogonal distance regressions for two different regimes, i.e., “clear sky” and “cloudy sky”. The two regimes are distinguished by the use of a transmissivity threshold. The post-processed GHI shows a significant reduction of the systematic biases and an improvement in representing the marginal distributions. A spatial cross-validation shows the applicability to the whole model domain of COSMO-REA6. Moreover, COSMO-REA6 as well as the post-processed GHI data reveal an added-value when compared to global reanalysis ERA-Interim and MERRA-2. The higher resolution reanalysis exhibits a significantly better performance of representing GHI variability, as well as biases, RMSE and other conventional scores. The post-processed GHI data are freely available for download.