| Understanding the processes and pathways by which rainfall is transformed into streamflow in headwater catchments is a key prerequisite to understanding pollutant transport, biogeochemical cycling, and aquatic ecosystem function in these regions. We examined a seven-year record of episodic and weekly stream chemistry observations for three forested catchments, each roughly 10 km 2, in Shenandoah National Park, Virginia. Our goal was to explore the uses and limitations of catchment-scale observations as the basis for inferring a regionally applicable conceptual model of catchment hydrochemistry.; The first part of our work focused on episodic data. We observed trends in the relative frequency of clockwise and anticlockwise concentration-discharge ("c-Q") loops for acid neutralizing capacity (ANC) versus stormflow, as well as associations between loop rotation direction and environmental predictors, for all three catchments. These observations exhibited a degree of systematic variation consistent with several equally plausible two- and three-component mixing models. However, even with a large set of observations, we were unable to unambiguously identify a single model based on c-Q loops alone. We conclude that, without additional supporting observations, c-Q loops are not a sufficient basis for constraining models of catchment hydrochemistry, no matter how long the record.; The second part of our work focused on building continuous simulation, two-component mixing models, with the hydraulic response of the two components constrained independently of the solute data using the empirical hydrologic model IHACRES. When calibrated on daily time steps for Paine Run, IHACRES formed the basis of an informative model for chloride flux on weekly to inter-annual time scales, permitting the identification of two parameters associated with solute residence time in the shallow subsurface and along deeper "baseflow" pathways. The model will serve as a useful conceptual framework for comparing Paine Run with the other two study catchments, constraining more detailed models of reactive constituents such as ANC, and speculating as to the effect of hypothetical future climate or atmospheric deposition scenarios. However, in all these applications, credible inferences will require careful consideration of the approximations associated with representing a complex, distributed dynamic system with a low-order, linear stationary model. |
