The main goal of this thesis is to explore the influence of sea surface temperature (SST), on surface wind at seasonal and intraseasonal timescales. At seasonal timescales, momentum and convergence budget were first documented by using a simple atmospheric mixed layer model, and two reanalyses. This approach allows us to identify the main processes that control the surface wind dynamics, in order to explore their sensitivity to the SST. Results show that these processes vary strongly in different regions of the tropical Atlantic. In addition, the comparison of the representation of theses processes in observations and reanalyses show that, as in all climate models (coupled or not), the reanalyses have the same flaws. Eventually, this work proposes a method to better assess the capacity of an atmospheric model to answer the SST fluctuations, and investigate potential wrong atmospheric parameterizations, such as boundary layers. The second part of this study focuses on tropical Atlantic regions of strong SST gradients, where SST intraseasonal variability is the largest. Several technics of spectrum and statistical analysis were performed in order to investigate the atmospheric patterns associated to these fluctuations of oceanic fronts. Except in the equatorial region (where we found a clear coupling already described in previous studies), no clear hint of a surface wind response to the SST fluctuations was observed in the two coastal upwelling fronts. In addition, the oceanic patterns associated to the SST indexes were also investigated. In all three upwelling fronts, as expected for such upwelling regimes, the vertical oceanic mixing clearly dominates the mixed-layer heat budget. In the equatorial band, as found in previous studies, the horizontal advection is equally important, while it appears surprisingly weak in the coastal fronts. Eventually, potential signals of equatorial and coastal Kelvin waves were also followed to these coastal fronts.