Precipitation is an important component of the Earth' water cycle and needs to be carefully monitored. Its large variability over a wide range of spatial and temporal scales must be taken into account. For example, hydrological models require accurate rainfall estimates at high spatial and temporal resolutions (e.g., 1 km and 5 min or higher). Obtaining accurate rainfall estimates at these scales is known to be difficult. So far, the only instruments capable of measuring rainfall over extended domains at such resolutions are weather radars. Their estimates are, however, affected by large errors and uncertainties partly due to the spatial and temporal variability of the drop size distribution (DSD). Major progress in the field is slowed down by the lack of knowledge about the spatial and temporal variability of DSD at scales that are relevant in remote sensing. This lack of reference data can be addressed through two different methods : (1) experimental investigations and (2) stochastic simulation. In this thesis, a comprehensive framework for the stochastic simulation of DSD fields at high spatial and temporal resolutions is proposed. The method is based on Geostatistics and uses variograms to describe the spatial and temporal structures of the DSD. The simulator' ability to generate large numbers of DSD fields sharing the same statistical properties provides a very useful theoretical framework that nicely complements experimental approaches based on large networks of weather sensors. To illustrate its potential, the simulator is applied to different rain events and validated using data from a network of disdrometers at EPFL. The results show that the simulator is able to reproduce realistic spatial and temporal structures that are in adequacy with ground measurements. The second part of this thesis focuses on the simulation and parametrization of intermittency (i.e., the alternating between dry/rainy periods). Simple scaling functions that can be used to downscale/upscale intermittency at different spatial and temporal resolutions are proposed and used to parametrize a new disaggregation method that includes the DSD as an output. Finally, different methods to identify dry and rainy periods and to quantify rainfall intermittency using telecommunication microwave links are proposed. The false dry/wet classification error rates of each method are estimated using data from a new and innovative experimental set-up located in Dübendorf, Switzerland. The results show that the dry/wet classification is significantly improved when data from multiple channels are used.