Spatial resolved and localized uncertainty analysis of the irradiance product from the CAMS Radiation Service v4

Kurzfassung

The Copernicus Atmosphere Monitoring Service (CAMS) through its CAMS Radiation Service (CRS) offers historical all-sky solar irradiance estimates for Europa/Africa since 2004 and for Asia/Oceania since 2014. The provision of this data is an important service to the solar energy community. Moreover, the integration of the CRS data requests in the pvlib library since 2018 has proven an ever-increasing interest from a variety of the solar community stakeholders. The performance of the CRS products is assessed every three months by the Evaluation and Quality Control (EQC) reports using standard statistical error metrics. Looking into a continuous improvement of the CRS service, one of the aims in the CAMs sErvice evOlution (CAMEO) project is to assess and quantify the error patterns of the CRS irradiance on a localized pixel scale. In the framework of this project, we are developing an error model for the CRS irradiance that identifies the local expected error for every individual estimate. The approach chosen for the uncertainty model is based on the identification of variables that have a high impact on the error, e.g. cloud properties, solar and satellite geometries and atmospheric conditions. To select the parameters that are the most important for predicting CRS uncertainty, we propose to identify and analyze spatial patterns of CRS error, whose determination is made possible by high density networks. We have collected ground observations from two densely populated irradiation networks: Météo-France (266 stations) and ISE PVLive (40 stations, Lorenz et al. 2020) , which measure Global Horizontal Irradiance (GHI) with a 1 min time resolution. As a comparative reference, the data from five BSRN stations in the same region is also used. As an example of the analysis, the spatial distribution of the standard deviation of CRS GHI for cloudy conditions is shown in Figure 1. This figure shows three panels representing the monthly standard deviation in the upper panel, seasonal evolution of the standard deviation in the medium panel and maps of the average GHI in the lower panel for each one of the locations retained for the study (Météo-France and PVLive pyranometric network in circles and BSRN stations in rectangles). The most relevant features identified are marked by rectangles. In the sections marked with a), the standard deviation of the error shows a meridional trend with a higher standard deviation in the South than in the North which has been found to be linked to mean solar irradiance. The spatial pattern marked with b) is observed in the west of the Alps and is identified as being related to orography. The dependencies found for the standard deviation, and also other error metrics, are currently being investigated from a modeling perspective in CRS to verify whether this effect results from issues in the coupling with clouds. At the time of writing the quantification of this localized uncertainty model is being developed. This should result on an additional indicator appended to the irradiance product that will inform the CRS user about a specific uncertainty for each individual estimation.