This paper aims to interpret the temporal variations of microwave brightness temperature (at 19 and 37GHz and at vertical and horizontal polarizations) in Antarctica using a physically based snow dynamic and emission model (SDEM). SDEM predicts time series of top-of-atmosphere brightness temperature from widely available surface meteorological data (ERA-40 re-analysis). To do so, it successively computes the heat flux incoming the snowpack, the snow temperature profile, the microwaves emitted by the snow and, finally, the propagation of the microwaves through the atmosphere up to the satellite. Since the model contains several parameters whose value is variable and uncertain across the continent, the parameter values are optimized for every 50 km × 50 km pixel. Simulation results show that the model is inadequate in the melt zone (where surface melting occurs on at least a few days a year) because the snowpack structure and its temporal variations are too complex. In contrast, the accuracy is reasonably good in the dry zone and varies between 2 and 4 K depending on the frequency and polarization of observations and on the location. At the Antarctic scale, the error is larger where wind is usually stronger, suggesting either that meteorological data are less accurate in windy regions or that some neglected processes (e.g. windpumping, surface scouring) are important. At Dome C, in calm conditions, a detailed analysis shows that most of the error is due to inaccuracy of the ERA-40 air temperature (∼2 K). Finally, the paper discusses the values of the optimized parameters and their spatial variations across the Antarctic.