Unravelling chemical and physical processes in Fourier transform spectra of low frequency dissolved oxygen time series in Swiss lakes

Abstract

Frequency domain analysis of hydrological time series can be used to extract information on physical and chemical processes, but such analyses are lacking for low frequency (fortnightly or monthly) time series of concentrations of lake constituents. We have explored the potentiality of these usual frequency time series in understanding processes in lakes. To this end, we have calculated Fourier transform spectra for dissolved oxygen time series at 14 sampling stations of 11 Swiss lakes for the approximate period 1980-2010. Despite noisy spectra, and spectral slopes (α) with a wide range of uncertainty, we have found consistent patterns of α change with the position in the lake and depth of the sampling stations. Spectral power (SP) shows a dependence of power on frequency (f) of the type SP ∝ f^-α with α varying from -0.103 to 1.40. α increases with depth and with distance from the main lake inlet up to 1-1.40. Spectral slopes have a significant parabolic relationship with mean oxygen at each depth (r² = 0.285, N = 182, p < 10^-4), indicating the influence of mixing processes in the lakes. Regression results of α with water travel time from the main lake inlet -’processing time’- and the square of dissolved oxygen concentration (r² = 0.363, N = 182, p < 10^-4) suggest that surface α are sensitive to the rate of recalcitrant dissolved organic matter decomposition, meaning that only processes with rates within the frequency range of the spectra (7.45x10^-5 to 0.028 d^-1) can exert a visible effect on the spectra and α. Potentially, effects on α variation can be used to detect or characterise slow processes in lakes for other compounds using long-term time series.