Author(s): Annalisa Molini; Paolo La Barbera; Luca G. Lanza
Linked Author(s):
Keywords: Rainfall field reconstruction; Global correlation; Information flows
Abstract: The issue of quantifying global information flows within observed rainfall fields is addressed in this paper. First we discuss the relations between correlation functions and mutual information, which respectively represent a statistical measure of the linear and general (linear and non linear) dependence governing the coupled evolution of the system at different locations in space and time. Since an analytical expression relating mutual information and linear correlation can only be derived in the case of binary time series, in dealing with the rain process the two statistics cannot be directly compared. However, a qualitative comparison is possible in terms of their general structure and sensitivity to the climatological and orographic characteristics of the considered space-time domain. We also discuss the influence of the finite-size effect on inferring mutual information from relatively “short” hydrological time series and the problem of obtaining a normalized expression for global correlation in order to compare results that are obtained at different aggregation scales in time. Finally we show how additional insights can be gained on the structure of the rain process by a deeper understanding of the information flow within its space-time structure.
Year: 2002