Abstract: Spaceborne microwave radiometers are an attractive tool for observing Antarctic climate because their measurements are related to the snow temperature. However, the conversion from microwave emission to snow temperature is not simple and strongly depends on the emissivity through S snow properties. This difficulty in predicting the snow property profile for Antarctic conditions is the main bottleneck in the retrieval of accurate climate information from microwave radiometers. We attempt to explain the vertically polarized emissivity at 19.3 and 37 GHz derived from brightness temperatures acquired by the Special Sensor Microwave/Imager (SSM/I) and physical temperature from the ERA-40 re-analysis. In Antarctica the snow emissivities at 19.3 and 37 GHz are nearly equal although a decrease with frequency is expected. To explain this, we consider various profiles of snow grain size and density and predict their emissivity using a dense-medium radiative transfer (DMRT) model. The results show that the emissivities cannot be explained by constant profiles of grain size and density. Heterogeneous snowpacks need to be considered. We first test random variations of snow density and grain radius with depth and then monotonic and continuous variations in the snow grain radius. In both cases, we show that an overall increase of the snow grain radius with depth is required to match the observed emissivity in Antarctica. In addition, two parameters characterizing the snow grain profiles are retrieved and compared with (1) in situ measurements of grain size at various locations in East Antarctica, (2) grain size estimated using visible spaceborne radiometers and (3) a semi-empirical relationship for grain growth.