Spectral Approaches to Time Series Analysis of the Short- to Long-Term Hydrological Variability of Karst Hydrosystems

Unlike other hydrogeologic systems, hydrological processes in karst aquifers present the specificity of varying over a large range of time scales. Among the various approaches investigating karst ground water flow and transport, time series analysis provides a most appropriate means for describing the modes of variability of related hydrological processes, but also to quantify their respective contribution to the overall dynamics of karst hydrosystems. In particular, spectral and time-domain methods, including correlation, Fourier spectral, continuous wavelet, wavelet coherence, singular spectrum analyses and filtering can be used for investigating time variability, transient characteristics, external forcings, and causal or non-causal physical relationships between hydrometeorological variables sensu lato over a wide range of time scales. It is a common view to think about spectral methods as a tool for searching periodicities or cyclicities. As a matter of fact, even when dealing with long time series in which cycles are observed (e.g., the water year), the transient nature of hydrometeorological variables rapidly sweeps this paradigm. In karst systems, all time scales may display a strong variability. Taking the example of a few piezometric, spring discharge, electrical conductivity and precipitation time series of different lengths, from 50 days to 30 years, in France (chalk aquifer, Upper Normandy) and in Texas, USA (Edwards aquifer, region of Austin) it can be shown what use can be made of spectral methods to describe karst aquifer dynamics by comparing flow and geochemical time series and by studying the long-term variability of hydrological processes in the context of global (i.e. environmental and climatic) changes.