Due to the increase of damage associated with floods enhanced by expanding urbanisation, the quantitative estimation of precipitation on a local scale is a societal need. However, such estimations are difficult because of the high variability of precipitation. Moreover, these estimations are requested by small organisations such as local authorities which cannot afford top level research instruments. Hence, new methods of estimation applicable to a cheap experimental set are needed. Toward this goal, an innovative experimental set is used in this work. It consists of ground instruments such as raingauges and disdrometers, and two radars, a prototype of a scanning radar based on a modified marine radar and a vertically pointing MRR (Micro Rain Radar), which give estimation of rain aloft, over an area and along a profile, respectively. The scanning radar works at X-band, meaning that it uses a longwave very attenuated by precipitation, but which provides a high range resolution. The MRR yields a detailed description of microphysics of precipitation and fills the gap between ground measurements and X-band radar measurements aloft. As both these radars are innovative, a large part of this PhD thesis was spent on the measurements validation : radar calibration, abnormal echoes filtering, attenuation consideration, etc. Using these corrected measurements, this PhD focus then on the study of the variability of precipitation, and aims to propose and develop several classification methods based on precipitation type or local variability, and to check their potential for the improvement of precipitation estimation. Results show that this goal can be reached only if the radar measurements quality is further improved : less interference echoes for the X-band radar, and consideration of vertical wind for the MRR.
