This thesis aims to improve the measurement, observation and statistical analysis of precipitation which is generally defined as falling aqueous particles. A particular interest is dedicated to rainfall, which is defined as liquid precipitation on the ground. Weather phenomena associated with precipitation impacts human life and economies over a wide range of temporal and spatial scales. The study of precipitation in Belgium, which has been based on rain gauge measurements, can now benefit from more than 10 years of radar measurements. The high resolution observations provided by radars are needed to study convective storms and for the verification of weather and climate models. Accurate rainfall estimation is required for flood risk management, especially at the urban scale. Unfortunately radars are not yet widely used due to the many sources of error in their measurements and the uncertainties associated with the derived rainfall estimation. The archived data of a C-band weather radar are used over the longest period available to improve the knowledge of precipitation in Belgium. Different methods combining a basic radar rainfall estimation with rain gauge measurements are evaluated. The characteristics of convective storms are analysed based on their identification and tracking in the volume reflectivity data. The reflectivity measurements are processed to control their quality and to obtain the best rainfall estimation. The radar-based rainfall estimation is verified by comparison with independent rain gauges and by looking at maps of descriptive statistics. Its potential for the study of local extreme rainfall is investigated by comparison with rain gauge analysis. For this purpose a new regional approach is proposed to fully exploit the radar information. The overall methodology is partly based on the refinement of existing approaches with simplicity and robustness in mind. Given the computational challenge, efforts have been made to ensure the high quality of the software implementation. A simple gauge bias correction significantly improves the performance of radar rainfall estimation. A further increase in performance is obtained by using spatial interpolation. Locally convective storms are observed 6 hours per year on average with no significant spatial variations. The distribution of convective activity indices (number of storms, area coverage and total mass) and individual storm properties (volume, mass, top, duration, speed) follows a power law with significant diurnal, monthly and annual variations. The main benefit for the rainfall estimation is obtained by taking into account the vertical profile of reflectivity. The new algorithm allows a 50% increase in areal coverage compared to a basic algorithm. Radar artifacts are significantly reduced by the removal of non-meteorological echoes and the hail threshold. Estimation errors and sampling explain the few differences between radar-based and gauge 1\,hour rainfall extremes. Using regression on a quantile-quantile plot, the fit of an exponential extreme value distribution model to both datasets are consistent thanks to the high uncertainty. The novel regional approach to the radar extremes allows to reproduce the results of long rain gauge records and to significantly reduce the model uncertainty. The availability and quality of the precipitation observations are fundamental to obtain robust results. The statistical results can be easily extended by using other radars and longer periods. The precipitation datasets have been used for research applications mainly for weather model verification. The radar-based rainfall estimation has been implemented operationally for hydrological purposes. Information on estimation uncertainty is important for many applications but has not been addressed in this thesis. The study of areal rainfall extremes is a natural extension to the study of local rainfall extremes. The use of the radar-based rainfall estimation for flood risk management is promising. The statistical results obtained in this thesis could enrich the climatic information regarding precipitation in Belgium.