Rainfall is characterized by an extreme variability over a wide range of space- and timescales. Underlying nonlinear phenomena produce strong and localized extremes that are poorly represented in current meteorological models. There is a need of stochastic representations that could reproduce rainfall statistics whatever the scale. Multifractal cascades, initially introduced in statistical physics of turbulence, are possible candidates. In this work, the "Universal Multifractal" model (Schertzer & Lovejoy, 1987) is considered. This model enables multiscale statistical characterization of a field by the means of three fundamental ("universal") parameters. Multifractal analysis tools have been applied to rainfall datasets representative of su...