Rainfall downscaling in a space‐time multifractal framework

Abstract

A space‐time multifractal analysis on radar rainfall sequences selected from the Global Atmospheric Research Program Atlantic Tropical Experiment database is presented. It is shown that space‐time rainfall can be considered with a good approximation to be a self‐similar multifractal process, so that a multifractal analysis can be carried out assuming Taylor's hypothesis to hold for rainfall over a wide range of spatial and temporal scales. The advection velocity needed to rescale the time dimension is estimated using different tracking techniques. On each selected rainfall sequence, a very good scaling is observed for spatial scales ranging from 4 to 256 km and for timescales from 15 min to 16 hours. A recently developed scale‐covariant multifractal model is then reformulated for numerical simulation of space‐time rainfall fields. The two parameters of the log‐Poisson distribution used as cascade generator within the model are systematically estimated from each selected rainfall sequence, and the dependence of one of these parameters on the large‐scale rain rate is highlighted. The model is then applied to disaggregate large‐scale rainfall, and some comparisons between synthetically downscaled and observed rainfall are discussed.