Validation of downward surface radiation derived from MSG data by in-situ observations over the Atlantic ocean

The present work investigates the quality of the shortwave and and longwave downward radiation (DSR, DLR) at the sea surface over the Atlantic Ocean as retrieved from Meteosat Second Generation (MSG) measurements and EUMETSAT's Climate Monitoring-Satellite Application Facility (CM-SAF) algorithms. The observations taken at two transatlantic research cruises have been an ideal basis to be compared with the MSG products for DSR and DLR derived from Meteosat-8 and Meteosat-9. Onboard the research vessels "Akademik Ioffe" and "Polarstern" high quality in situ measurements of both radiation fluxes have been performed. Continuous full sky imagery and standard meteorological observations enable a comprehensive evaluation of the skills of MSG DSR- and DLR-retrievals in different climate zones and under various cloud and weather conditions. The DSR was retrieved by MSG with a positive bias of 2.77 W m(-2) during the Ioffe cruise, and 22.23 W m(-2) during the Polarstern cruise. The bias for the DLR was -1.73 W m(-2) and 2.76 W m(-2), respectively. The differences between the two cruises mainly arise from the different weather conditions. No significant differences between the satellite products from Meteosat-8 and Meteosat-9 were found. In general DSR and DLR for clear sky conditions are captured with a high accuracy. Largest retrieval errors occur for fast fluctuating broken cloud conditions, though on average the MSG algorithm match the in-situ observations well. Semitransparent cirrus was found to cause a negative bias for the retrieved DSR. In tropics and subtropics the errors for DLR are smaller compared to higher latitudes. Most importantly, no significant dependencies of the satellite retrieval errors for both the DSR and the DLR on the solar elevation, near-surface humidity, cloud cover, SST and the shift of day and night were found, indicating that the CM-SAF radiation products are not subject to significant systematic errors.