On the importance of the atmospheric coupling to the small-scale ocean in the modulation of latent heat flux

In this study, ocean and atmosphere satellite observations, an atmospheric reanalysis and a set of regional numerical simulations of the lower atmosphere are used to assess the coupling between the sea-surface temperature (SST) and the marine atmospheric boundary layer (MABL) as well as the latent heat flux (LHF) sensitivity to SST in the north-west tropical Atlantic Ocean. The results suggest that the SST-MABL coupling depends on the spatial scale of interest. At scales larger than the ocean mesoscale (larger than 150 km), negative correlations are observed between near-surface wind speed (U-1 (0m)) and SST and positive correlations between near-surface specific humidity (q(2m)) and SST. However, when smaller scales (1 - 150 km, i.e., encompassing the ocean mesoscale and a portion of the submesoscale) are considered, U-10 (m)-SST correlate inversely and the q(2m)-SST relation significantly differs from what is expected using the Clausius-Clapeyron equation. This is interpreted in terms of an active ocean modifying the near-surface atmospheric state, driving convection, mixing and entrainment of air from the free troposphere into the MABL. The estimated values of the ocean-atmosphere coupling at the ocean small-scale are then used to develop a linear and SST-based downscaling method aiming to include and further investigate the impact of these fine-scale SST features into an available low-resolution latent heat flux (LHF) data set. The results show that they induce a significant increase of LHF (30% to 40% per degrees C of SST). We identify two mechanisms causing such a large increase of LHF: (1) the thermodynamic contribution that only includes the increase in LHF with larger SSTs associated with the Clausius-Clapeyron dependence of saturating water vapor pressure on SST and (2) the dynamical contribution related to the change in vertical stratification of the MABL as a consequence of SST anomalies. Using different downscaling setups, we conclude that largest contribution comes from the dynamic mode (28% against 5% for the thermodynamic mode). To validate our approach and results, we have implemented a set of high-resolution WRF numerical simulations forced by high-resolution satellite SST that we have analyzed in terms of LHF using the same algorithm. The LHF estimate biases are reduced by a factor of 2 when the downscaling is applied, providing confidence in our results.