Use of wavelet transform for temporal characterisation of remote watersheds

Abstract

French Guyana is a remote, wild and little instrumented French department located in South America. Without rainfall dataset, rivers that can be temporally studied from their outlets offer easy accessible hydrological and climatologic information on the remote areas that they cover. As part of a large project intended to build a biotic index of water quality, we studied the hydrological properties of the main French Guyana basins. By computing a set of temporal indicators for flow variability characterisation, we extracted the most relevant by means of a principal component analysis. A cluster analysis based on the remaining indicators finally classified the nine watersheds. At this stage, French Guyana hydrology shows a surprising uniformity. For this reason, standard methods such as the Fourier analysis appear rapidly ineffective to analyse the runoff variability. Hence, we exploited the potential of continuous wavelet transform on flow curves to detect new periodicities or time annual features. The wavelet coefficient maps made it possible to detect and explain the so-called 'Short March Summer' meteorological phenomenon. The latter, a temporary reduction of rain during the rainy season of the Guyana plateau, has been explained as the result of the Atlantic Ocean influence on the continent. It permitted to characterise and discriminate more accurately the studied watersheds. © 2002 Published by Elsevier Science B.V.