INVESTIGATING HYDRODYNAMIC AND TRANSPORT PROPERTIES OF KARST AQUIFERS USING CORRELATION, SPECTRAL AND WAVELET ANALYSES OF RAINFALL, WATER LEVEL, TURBIDITY AND ELECTRICAL CONDUCTIVITY TIME SERIES

In many places in the world, a frequent type of contamination of drinking water consists of turbidity. On the other hand, turbidity, as representative of particle transport, can be used as a tracer of certain features of transport properties. In order to compare particle and dissolved element transport to hydrodynamics in karst systems, correlation and spectral analyses were first performed on rainfall (input signal), water level, electrical conductivity and turbidity (output signals) time series at a karst system in the chalk aquifer of the lower Seine valley for spring data collected in 2000. This system is composed by a spring connected to a main sinkhole where a small creek is swallowed on the plateau. The autocorrelation functions for water level and turbidity showed a short memory effect (around 4 days), demonstrating a short duration of influence of flood events across time, whereas electrical conductivity (low-mineralization storm-derived water) had a quite longer memory effect (28 days). These results could be interpreted as a rapid reactivity of the spring to rain events, although a significant water storage in chalk explained the long memory effect for conductivity (transit through the porous/fissured matrix). Energy spectra computed by fast Fourier transform of autocorrelation functions showed a strong structure in the output signals compared to rainfall which seemed more or less random, and proved the system organizes flow and mass transport. The impulse response functions estimated by cross-correlation confirmed a low inertia of the system regarding water level and turbidity, and a quite higher inertia with respect to conductivity. In addition to the main peak, two other peaks in the impulse response functions suggested the existence of additional point-source recharge features, which was supported by field observations. Wavelet analyses were investigated as well and allowed a more accurate study of such strongly time-localized signals. Continuous wavelet spectra highlighted a strong relationship between turbidity and rainfall, which was impossible to deduce from energy spectra. Peculiar structures in water level could be isolated and a time-domain reconstruction using inverse wavelet transform could provide a separation scheme of a quickflow (related to turbidity) and a slowflow component.