Fractal water quality fluctuations spanning the periodic table in an intensively farmed watershed.

Recently developed measurement technologies can monitor surface water quality almost continuously, creating high-frequency multiparameter time series and raising the question of how best to extract insights from such rich data sets. Here we use spectral analysis to characterize the variability of water quality at the AgrHys observatory (Western France) over time scales ranging from 20 min to 12 years. Three years of daily sampling at the intensively farmed Kervidy-Naizin watershed reveal universal 1/f scaling for all 36 solutes, yielding spectral slopes of 1.05 ± 0.11 (mean ± standard deviation). These 36 solute concentrations show varying degrees of annual cycling, suggesting different controls on watershed export processes. Twelve years of daily samples of SO4, NO3, and dissolved organic carbon (DOC) show that 1/f scaling does not continue at frequencies below 1/year in those constituents, whereas a 12-year daily record of Cl shows a general 1/f trend down to the lowest measurable frequencies. Conversely, approximately 12 months of 20 min NO3 and DOC measurements show that at frequencies higher than 1/day, the spectra of these solutes steepen to slopes of roughly 3, and at time scales shorter than 2-3 h, the spectra flatten to slopes near zero, reflecting analytical noise. These results confirm and extend the recent discovery of universal fractal 1/f scaling in water quality at the relatively pristine Plynlimon watershed in Wales, further demonstrating the importance of advective-dispersive transport mixing in catchments. However, the steeper scaling at subdaily time scales suggests additional short-term damping of solute concentrations, potentially due to in-stream or riparian processes.