We developed an empirical relative wind direction (RWD) model function to represent azimuthal variations of oceanic microwave radiances of vertical and horizontal polarizations. The RWD model function was based on radiance measurements from the Advanced Microwave Scanning Radiometer and Special Sensor Microwave Imager/Sounder (SSMIS). Ocean surface wind vector data from SeaWinds on board the Advanced Earth Observing Satellite-II and European Centre for Medium-range Weather Forecasts (ECMWF) Integrated Forecasting System were utilized for the RWD model function development. The RWD model function was introduced to a microwave ocean emissivity model: a FAST microwave Emissivity Model (FASTEM) in a radiative transfer model for satellite radiance assimilation. Performances of the RWD model function were much more realistic than present azimuthal model functions in FASTEM for low wind speed and high-frequency channels. Assimilation experiments using the RWD model function were performed in the ECMWF system. The experiment demonstrated reductions of first-guess departure biases arising from modelling of the azimuthal variations in areas of high wind-speed and low variability of wind direction. For example, bias reductions in ascending and descending SSMIS 19 GHz vertically polarized radiance in the Somali jet over the Arabian Sea were approximately 0.6 and 0.7 K. The bias reductions were found for all assimilated microwave imager channels in a wide wind-speed range. Moreover, analysis increments of specific humidity in the lower troposphere were reduced (e.g. 0.2 g kg⁻¹ reduction at 1000 hPa in the Somali jet). We found improvements of relative humidity and temperature in short-range forecasts in the lower troposphere. The experiment results clearly showed the importance of modelling the azimuthal variation of emissivity for assimilation of microwave imager observations. The new RWD model function, combined with the other components of FASTEM, will be available as FASTEM-6.