Air-sea exchange play a key role in the dynamics of the Earth's climate system and its accurate representation in numerical models is essential for obtaining reliable climate change predictions. The fundamental problem is that numerous processes comprising air-sea exchanges have small spatial scales not explictly resolved by general circulation models. Most of surface schemes in GCMs use formula based on local measurements, in assuming horizontally homogeneous parameters at gridscale. This assumption is not valid when subgrid motions
inducing horizontal wind variability exist but are not explicitly represented. Two different sources of horizontal wind variability have to be distinguished: deep convection and boundary layer free convection. The scales involved (time and length) and convective patterns are very different for each of these cases.
To address theses issues a combined approach will be proposed, by using observations and numerical simulations of TOGA-COARE. Numerical simulations enable one to scan time-space scales corresponding to variability induced by any convective activity (from Large Eddy Simulations to Precipitating Systems Simulations). LES are evaluated against vertical fluxes in ABL as estimated by aircrafts For boundary layer free convection, the horizontal wind variability can be related to the free convection velocity. To establish this relationship,
both observations and numerical simulations are used. COARE data are revisited to propose changes in the COARE bulk algorythm.
For deep convection, the dominant source of wind variability is the occurence of downdraft generated by convective cells. For low wind conditions downdraft can produce large enhancement of surface fluxes. Results indicate that the gustiness velocity can be related either to the precipitation or either to updraft and downdraft mass fluxes.
A general parameterization of mesocale enhancement of surface fluxes is proposed. That distinguishes effects of free convection in boundary layer and precipitating deep convection.