In winter, the Northwest Tropical Atlantic Ocean can be characterized by various regimes of interactions among ocean current, surface wind, and wind waves, which are critical for accurately describing surface wind stress. In this work, coupled waveocean-atmosphere model simulations are conducted using two different wave roughness parameterizations within COARE3.5, including one that relies solely on wind speed and another that uses wave age and wave slope as inputs. Comparisons with the directly measured momentum fluxes during the ATOMIC/EUREC4A experiments in winter 2020 show that, for sea states dominated by short wind waves under moderate to strong winds, the wave-based formulation increases the surface roughness length by 40\% compared to the wind-speed-based approach. For sea states dominated by remotely generated swells under moderate to strong wind intensity, the wave-based formulation predicts significantly lower roughness length and surface stress (˜20%), resulting in increased near-surface wind speed above the constant flux layer (˜5%). Further investigation of the mixed sea states in the model and data indicates that the impact of swell on wind stress is over-emphasized in the COARE3.5 wavebased formulation, especially under moderate wind regimes. Various approaches are explored to alleviate this deficiency by either introducing directional alignment between wind and waves or using the mean wave period instead of the wave period corresponding to the spectral peak to compute the wave age.