Improving Atmospheric Fluxes with Ocean Reanalysis

This project explores a method to reduce the level of uncertainty in surface heat flux estimates. Time mean surface heat flux estimates provided by atmospheric reanalyses differ by 10-30W/m². They are generally unbalanced globally, and have been shown, when used to force ocean simulations, to be incompatible with ocean temperature and velocity measurements. Here a method is presented 1) to identify the spatial and temporal structure of the underlying flux errors and 2) to reduce them by exploiting information contained in hydrographic observations through the analysis increments produced by an ocean reanalysis using sequential data assimilation. The method is applied to heat fluxes computed from daily state variables from three widely used reanalyses: MERRA2, ERA-Interim, and JRA-55, during an eight year period 2007-2014. For each of these data sets seasonal heat flux errors/corrections are estimated.

In a second set of experiments the three heat flux data sets are corrected based on the error estimates and the ocean reanalysis experiments are repeated with the improved flux estimates. This second round of experiments shows that the time mean error in the corrected fluxes is reduced to within ±5W/m² over the interior subtropical and midlatitude oceans, with the most significant changes occuring over the Southern Ocean. The global heat flux imbalance of each reanalysis is reduced to within a few W/m² with this single correction. Encouragingly, the corrected forms of the three sets of fluxes are also shown to converge. In the final discussion we present experiments beginning with a modified form of the ERA-Int reanalysis, produced by the DAKKAR program, in which state variables have been individually corrected based on independent measurements. Finally, we discuss the separation of flux error from ocean model error.