In order to estimate the impact of climate change it is necessary to monitor the surface mass balance of the ice sheets. Much effort has gone into developing accurate methods for surface mass balance estimation over the past years. Mass balance is defined as the sum of mass gain and mass loss. Still, it remains difficult to quantify surface accumulation, which is the net gain term of the balance equation. This is due to the fact that the polar ice sheets are difficult to access, so in-situ measurements are sparse. Therefore, it is important to employ remote sensing techniques to obtain accumulation data with a better spatial and temporal resolution. However, the resulting datasets still exhibit large amounts of uncertainty. This thesis seeks to improve the currently available methods to derive snow accumulation rates from microwave remote sensing data. For this purpose, different types of microwave data are systematically evaluated with respect to their suitability for accumulation rate retrieval. The approach taken here makes use of the fact that microwaves interact with a volume of dry polar firn and are sensitive to accumulation rate-dependent firn characteristics. For this reason, firn microstructure is examined in more detail. On the basis of this analysis, an improved parameterization of firn properties (grain size and density) is developed, which is valid for the layers of the firn column that interact with the microwave radiation. The improved microstructure model is used in conjunction with a simple radiative transfer model to simulate firn-microwave interaction, resulting in a synthetic microwave signal. Accumulation rates can subsequently be inverted by matching the signal from the model to microwave data measured by satellite sensors. A number of assumptions made in the radiative transfer model have proven to be invalid. In this work, the radiative transfer model is improved by including Mie scattering instead of the widely-used Rayleigh approximation for the combination of scatterer sizes and microwave frequencies investigated. The results from the accumulation retrieval algorithm developed in this work are validated with in-situ data, and limits in its applicability are evaluated. Accumulation rates inverted from the method introduced in this work are found to agree with field data as well as with accumulation maps from external sources within the range of the model's validity.