ABSTRACT
A methodology developed for statistical analysis of rainfall fields is applied to weather-radar statistics on rainfall observed over the tropical Atlantic during the GATE experiment. First, the problem of robust estimation of the intensity of rainfall is addressed using a novel, resistant, edge-preserving technique for multivariate smoothing of structured data from spatial and spatial-temporal processes. Secondly, the problem of the assessment of point-process models for rainfall fields is addressed. The investigation of rainfall fields entails the analysis of data from spatial processes. These data refer to the statistics of random phenomena that distribute in space and evolve in time. Statistical models based on point processes are fundamental in the literature to phenomenological models for rainfall fields. Many of these draw upon the mesoscale features of precipitative systems, taking the convective raincell as the primitive for points.
