Evaluation of the daily sea surface net radiation from nine satellite, reanalysis and reconstructed products and uncertainty estimates

The performance of all-wave net radiation ( R n ) estimates at the global sea surface is largely unknown. In this study, nine daily mean long-term sea surface R n products were compared and evaluated using observations from 55 moored buoys. The nine products included three satellite products: CERES SYN_1deg_Ed4A (CERES-4A), the Japanese Ocean Flux Datasets with Use of Remote Sensing Observations, Version 3 (J-OFURO3), and GEWEX– SRB; five reanalysis products: ERA-Interim, ERA5, JRA55, MERRA2, and NCEP R2; and one reconstructed product: the Objectively Analyzed Air-Sea Fluxes from the International Satellite Cloud Climatology Project radiative flux D-series product (OAFlux+ISCCP). The results indicate the following: (1) large discrepancies appeared in all products, particularly in the tropics and high-latitude seas; (2) the satellite products generally outperformed the reanalysis products, among which J-OFURO3 performed the best with an R 2 of 0.88, a root-mean-square difference (RMSD) of 21.88 Wm −2 , and a bias of 0.28 Wm −2 , and NCEP R2 was the worst, with an R 2 of 0.34, an RMSD of 60.30 Wm −2 , and a bias of -13.11 Wm −2 , during the study period from 2000 to 2013, and the conclusions were almost unaffected by the spatial resolution and time period, but these products performed worse over the Indian Ocean; and (3) the long-term trends, variations, and magnitude in the annual average sea surface R n from all products were remarkably different, especially before 2000, and thus, it is difficult to tell which product is reliable. Overall, these nine products show significant disagreements, and each product has its own advantages and drawbacks; therefore, users are advised to make selections with cautions.