Thursday, 7 August 2003
A comparative study of different rainfall downscaling processes
Simulating rainfall fields with higher resolution than observed would be useful in quantifying the errors of rainfall estimation by weather radar. Several approaches have been proposed to achieve the rainfall downscaling, all of them based on extrapolating some experimental law over the available scales to the smaller ones. Among them, modeling the rainfall as a Multiple Poisson Process, decomposing the rainfall field using a wavelet transform, and considering the multifractal behavior of the rainfall field seem to be the most applied procedures. The aim of this paper is to compare the different alternatives, in particular to compare the wavelet decomposition with the multifractal model, to determine which one reproduces the best the rainfall variability over the different scales. To simplify our comparative study, this has been done over one-dimensional rainfall fields, using time series of rain intensity measured by a Doppler disdrometer with one minute resolution. The analysis has been conducted by testing the ability of each downscaling procedure (wavelet decomposition and multifractal model) to reproduce either the multifractal characteristics or the wavelet analysis results of the observed rainfall sequences. Preliminary results show that none of the approaches is able to reproduce the characteristics of the other, although more cases must be simulated and analyzed.