Measuring and Modeling Oceanic Air-Sea Fluxes

There have been numerous studies evaluating model representation of the latent heat flux (LHF) over terrestrial surfaces due to LHF’s role in weather prediction, heat balance, and the hydrological cycle. However, LHF model representation over the ocean, where 86% of global evaporation occurs, has been largely untested due to the scarcity of in-situ measurements and difficulties associated with open ocean observations. This study evaluates the Weather Research and Forecasting Model (WRF) latent heat, sensible heat and momentum surface fluxes over the Sub-tropical North Atlantic Ocean from September 16 to October 30, 2012 under various surface layer and planetary boundary layer (PBL) parameterization schemes. WRF output is validated against bulk and direct covariance flux observations collected during the NASA Salinity Processes in the Upper-Ocean Regional Study (SPURS) from a highly instrumented surface mooring and surveying research vessel. WRF is also compared to the OAFlux hindcast product at an interpolated 1-day, 1-degree resolution. WRF produced a persistent positive bias in LHF when evaluated against buoy and ship measurements at respective locations in all native parameterization schemes. Modifications to surface layer schemes were employed to mimic the functionality of the COARE3.5 bulk flux algorithm. These modifications reduced root mean square error (RMSE) and model bias in LHF for the default surface layer scheme known as MM5. The inclusion of COARE3.5 functionality had a slightly negative impact on a preferred surface physics scheme known as MYNN, which had already been modified to use the COARE3.0 algorithm. The MYNN scheme with COARE3.0 provided the minimum RMSE between model and observations. Results for momentum flux agreed with these findings. OAFlux closely agreed with SPURS validation data and produced lower surface LHF than WRF across the domain throughout the test case. This study has found that the WRF overestimation of surface fluxes is the result of inaccurate model physics; e.g., inappropriate convective velocity scale approximation and the lack of an oceanic cool-skin correction. Other factors contributing to the bias and RMSE are a positive bias in the SST used to initialize and force the model and large variability in wind speed, respectively