Ocean Surface Gravity Waves and Climate Modeling

Yalin Fan


Climate modeling, to a great extent, is based on simulating air-sea interactions. However, the small-scale interactions and related phenomena, such as the ocean waves are sub-grid processes, and cannot be simulated by climate models. Thus, Global circulation models(GCMs) usually parameterize the air-sea fluxes based on wind speed only, while as we all know that the air-sea fluxes are closely coupled with the sea state in the ocean. Furthermore, Stokes drift from ocean waves can create Langmuir circulations. Many Large Eddy Simulation studies have found that the maximum entrainment flux can be increases by a factor of two to five by Langmuir turbulence. There are also studies indicate that the interaction of stokes drift and Coriolis force has an appreciable influence on the Ekman transport at the high latitudes, which could affect the dynamic height and hence large circulation patterns in the ocean model.

Motivated by all these research questions, we have developed a fully coupled atmosphere-wave-ocean model at GFDL. NCEP’s WAVEWATCH III is used for this coupling. The wave climatology generated by this couple model was extensively evaluated against NDBC buoy observations, satellite measurements, and ERA-40 reanalysis and shown very good agreements in terms of magnitude, spatial distribution, and climate variabilities. A new momentum roughness (zo) parameterization as a function of both the wind speed and the sea state is implemented based on recent field observations, which is similar to Beljaar (1994) that was commonly used in GCMs when the wind speed is less than 20 m/s, but becomes significantly lower as the wind speed increases (Fan et al. 2012). The new zo has shown improvements in surface wind simulations for the 1990 control run following Delworth et al (2006). Two different Langmuir turbulence parameterizations are also implemented in the coupled model and evaluated using the 1990 control run. Preliminary results have shown that the ocean stratification should be taken into account for Langmuir turbulence parameterizations in GCMs.