Etude des pluies intenses dans la region Cévennes-Vivarais à l'aide du radar volumique. Regionalisation des traitements radar et analyse granulométrique des pluies au sol.

摘要

The quantitative interpretation of the weather radar signal in terms of rainfall is complex since it depends (i) on the rainfall variability at all scales (scales of the raindrops, of the radar resolution volume and of the precipitating system itself), (ii) on the radar detection domain, constrained by the surrounding relief and the vertical development of precipitations, and (iii) on the parameters and operating protocol of the radar system(s) employed. A pronounced relief obviously adds complexity to the radar quantitative precipitation estimation (QPE) problem by reducing the visibility and increasing environmental noise. We have addressed part of these problems in the present thesis with the development of an automated rainfall typing procedure into convective and stratiform regions based on the use of 3D weather radar data. First, we have shown the strong influence of the radar sampling properties for two algorithms already proposed in the literature by Steiner et al. (1995) and Sanchez-Diezma et al. (2000) for the detection of convective and stratiform precipitation, respectively. This problem was partially overcome by a decision tree and a coupling of the rain typing and the vertical profile of reflectivity (VPR) identification. The final algorithm is shown to significantly improve the rain-typing at long ranges (e.g., greater than 60 km).On the other hand, we have conducted an experiment in Alès during the autumn 2004 to document the Cévennes drop size distributions (DSD) at ground level by using an optical disdrometer. We have first implemented various methods to establish the reflectivity - rain rate relationship (Z-R relationship). Their respective merits were assessed through a self-consistency procedure based on DSD data alone. Then we have studied the seasonal, inter-storm and inner-storm variability of the Z-R relationship and shown the major influence of the inner-storm variability. Finally, we have performed a preliminary study of the link between the 3D radar data and the ground-based DSD data through a comparison of DSD-based and radar-based rain typing. The significant (though modest) benefit of the proposed rain-typing algorithm and Z-R relationships was established for the Bollène-2002 dataset. This work contributes to the development of the LTHE radar QPE software called TRADHy.