Rainfall downscaling in time: theoretical and empirical comparison between multifractal and Hurst-Kolmogorov discrete random cascades

Abstract

During recent decades, intensive research has focused on techniques capable of generating rainfall time series at a fine time scale that are (fully or partially) consistent with a given series at a coarser time scale. Here we theoretically investigate the consequences on the ensemble statistical behaviour caused by the structure of a simple and widely-used approach of stochastic downscaling for rainfall time series, the discrete Multiplicative Random Cascade. We show that synthetic rainfall time series generated by these cascade models correspond to a stochastic process which is non-stationary, because its temporal autocorrelation structure depends on the position in time in an undesirable manner. Then, we propose and theoretically analyse an alternative downscaling approach based on the Hurst-Kolmogorov process, which is equally simple but is stationary. Finally, we provide Monte Carlo experiments which validate our theoretical results. © 2012 Copyright 2012 IAHS Press.