This study examines the ability of the National Center for Environmental Prediction (NCEP)'s operational global analysis/forecast system to produce realistic estimates of transfers of water, energy and momentum across the air-sea boundary in its analysis cycle, short- and medium-range forecasts and longer-range forecasts. Air-sea fluxes serve as a good diagnostic of the model's parameterization of the boundary layer, radiation and cloudiness and are essential for coupling atmosphere and ocean models, a key element of dynamical seasonal forecasting. At least some air-sea fluxes from the NCEP global system may be as realistic as air-sea fluxes from any source.

Air-sea fluxes from the NCEP-I and NCEP-II reanalyses have been widely used and critically examined; surface stress and sensible and latent heat fluxes from reanalyses may be as realistic as other independent global estimates. Independent estimates of precipitation and surface radiation, however, appear superior to the reanalyses' estimates. NCEP-I is known to have too weak an equatorial wind stress in the Pacific; NCEP-II lacks low-level stratus clouds in the eastern subtropical oceans. Cloudiness appears to be one of the major problems in improving air-sea fluxes in numerical weather prediction.

This paper emphasizes differences in air-sea fluxes between the NCEP reanalyses and the operational NCEP global system. Since the NCEP/NCAR reanalysis, surface short wave fluxes, equatorial surface stress and the precipitation pattern have improved in the NCEP global system; however, sensible heat flux appears to be too low and the NCEP system still has problems with low-level stratus clouds. Air-sea fluxes in extended integrations of the NCEP global model will also be examined, together with NCEP's experience with air-sea fluxes in coupled atmosphere-ocean integrations and the effect of recent system changes on air-sea fluxes.