The Best Analyzed Air-Sea Fluxes for Seasonal Forecasting

Initial conditions for both the atmosphere and ocean are needed to initialize a coupled atmosphere-ocean forecast model for seasonal forecasts. Development of a new global coupled atmosphere-ocean forecast system at NCEP has forced us to consider what are the most appropriate sources of these initial conditions. Initial conditions for the ocean are analyzed by an ocean data assimilation system driven by air-sea fluxes from atmospheric data assimilation. To generate ocean initial conditions for the real-time seasonal forecasts, air sea fluxes from the operational global atmospheric data assimilation for daily weather forecasts have been used. An essential part of developing a new seasonal forecast system is the generation of hind-casts to provide bias corrections and estimates of forecast reliability. To generate ocean initial conditions for the hind-casts, air-sea fluxes from the NCEP-2 reanalysis (done with a different model than the current GDAS) have been used. The use of different sources of analyzed air-sea fluxes in generating ocean initial conditions produces differences in seasonal forecasts; the significance of these differences are being investigated. This experience implies that a consistent source of analyzed air-sea fluxes are needed for seasonal forecasts. Ways to provide such a source will be discussed.

This paper considers differences in air-sea fluxes between NCEP-1 (NCEP/NCAR), NCEP-2 and ERA-40 reanalyses and differences in air-sea fluxes between NCEP-2 and GDAS. GDAS provides higher resolution fluxes and has a better parameterization of evaporation at high wind speed than the NCEP reanalyses. It has a more realistic pattern of precipitation but its surface energy balance over the ocean has a greater imbalance. GDAS and NCEP2 surface stresses are similar in the east equatorial Pacific, but have larger differences in middle to high latitudes and over the west Pacific and Indian Oceans. They also have large differences in surface net short wave and net heat flux over the tropical oceans. GDAS has too much surface net short wave and net heat flux overall, but may have a more reasonable pattern than NCEP2.