Vers un modèle d'erreur pour la mesuredes pluies par radar météorologique en région Cévennes-Vivarais
Résumé
Toward an error model for radar quantitative precipitation estimation (QPE) Characterizing the error structure of radar QPE is recognized as a major issue for applications in hydrological modelling. Due to the variety of QPE sources of error, the problem can only be addressed practically by means of radar – raingauge comparisons. Using the radar and rain gauge datasets of the Bollène 2002 experiment, we propose a preliminary investigation of this subject within a geostatistical framework. First, rain gauge measurements are critically analysed and the kriging technique is used to map the rain fields over the considered region. The block kriging estimation variance is used to select a number of reliable reference rainfall values for various integration domains and timesteps. Then, the statistical distribution and the spatial and temporal structure of the residuals between radar and reference values are established and analyzed. It is shown that the requested radar QPE error model depends on the space-time scales considered and needs to be conditioned on the rain amount, on the radar range, and possibly on the rain type.