Linkages Between Hydrology and Biogeochemistry, Explored via Spectral Analysis of Catchment Hydrochemical Data

The hydrology and biogeochemistry of small catchments present two linked paradoxes. The first paradox is that streamflow responds promptly to rainfall inputs, but fluctuations in passive tracers (such as water isotopes and chloride) in streamflow are often strongly damped, indicating that storm flow is predominantly 'old' water. The mechanisms by which catchments can store water for months, but then release it in minutes during storm events, are not well understood. The second paradox is that although streamflow concentrations of passive tracers fluctuate very little during storm events, reactive tracers (such as pH, alkalinity, Al, Ca, Mg, Si, DOC, and NO₃) are often highly sensitive to discharge. Thus, although baseflow and highflow are both composed mostly of 'old' water, they can have very different chemical signatures. The physical and chemical mechanisms underlying this phenomenon are poorly understood. Here we explore these puzzles using spectral analysis of hydrochemical time series from small upland catchments. Spectral analysis shows that catchments act as fractal filters for passive tracers like chloride, converting "white noise" rainfall inputs into fractal "1/f noise" runoff time series (Kirchner et al., Nature 403, 524-527, 2000). These fractal time series can be generated by advection and dispersion of spatially distributed rainfall tracer inputs, as long as the flow system is highly dispersive (Kirchner et al., J. Hydrol. 254, 81-100, 2001). By contrast, spectral analyses of rainfall and streamflow water fluxes show that hydrologic signals are transmitted downslope more rapidly, and with much less dispersion, than passive tracer signals are. Thus small upland catchments transmit hydraulic potentials (which drive runoff) much less dispersively than they transport water itself. Alternative conceptual models for subsurface flow and transport in catchments have distinct spectral signatures; thus spectral methods can be used for model testing as well as for exploratory analysis of catchment data.