Assessing the Birkenes Model of stream acidification using a multisignal calibration methodology

Abstract

A revision of the Birkenes model of streamwater acidification has been attempted to incorporate additional chemical and hydrologic information gained in the last 6 years since its original construction. The first stage of this effort has been an analysis of the hydrologic submodel with the goal of extending it to predict concentrations of a conservative tracer in streamwater. An objective calibration of the model indicated that the model is overparameterized. Only one passive store is identifiabile, not two as currently contained in the model and the routing between the two reservoirs is not determined by the data. Inclusion of the conservative tracer improved the identifiability of the dimensional parameters, but had little effect on the rate or routing parameters. If the hydrologic structure is to be determined from the hydrograph and conservative tracer alone, it must be simplified to eliminate unidentifiable parameters. The validity of using more complex rainfall‐runoff models in hydrochemical models which seek to test chemical mechanisms is called into question by this analysis.

Flow, mixing, and displacement in using a data‐based hydrochemical model to predict conservative tracer data

“…For example, Weiler et al [2003] identified different transfer functions for the two studied events examined related to distinct stormflow generating processes/sources; Iorgulescu et al [2005] identified a steep, almost threshold-like, increase in the soil water contribution with catchment wetness and similar shapes of the transfer function for the soil water and direct precipitation components. Another interesting inference from these studies is that multiresponse discharge and tracer data may support a higher-dimensional model parameterization than previously thought [e.g., Hooper et al, 1988].…”

“…Such water must be displaced from storage during the response to rainfall, implying that celerities associated with the propagation of the effects of the input are faster than the velocities of flow of that input. There is already an established literature demonstrating the use of tracer data response in the framework of conceptual catchment models [e.g., Hooper et al, 1988;Robson et al, 1992;Barnes and Bonnell, 1996;Kuczera and Mroczkowski, 1998;Uhlenbrook and Leibundgut, 2002]. More recently, Seibert and McDonnell [2002] adopted a flexible approach based on catchment-specific conceptual modeling using a set of interconnected boxes and ''soft'' data integration, though they assume that the effective mixing volumes for both runoff and tracer responses are the same.…”

Bayesian multiresponse calibration of TOPMODEL: Application to the Haute‐Mentue catchment, Switzerland

“…Hooper et al [1988] developed and tested a multisignal optimization methodol-ogy of the Birkenes hydrochemical model in order to better assess the identifiability of the model parameters. They calibrated the model to runoff and tracer data and used simple and weighted least squares objective functions and a gradient search optimization technique in order to estimate model parameters.…”

On the dialog between experimentalist and modeler in catchment hydrology

“…We view this use of isotopic data as one of the first formal attempts to include isotope-based hydrograph separation results into a model exercise. While a few studies in the past have used stream isotope concentrations through time for model calibration and testing [Hooper et al, 1988;Seibert et al, 2002a], there are no studies we are aware of that make use of computed new water ratios. O-18 time series could be used directly (i.e., as hard data), but often the observed signal is weak and observed time series are short and discontinuous (for review see Buttle [1994]).…”

“…Few studies to date have used isotope data in model calibration-despite the now common use of this in watershed analysis ]. Hooper et al [1988] instance, snow cover information may be used. In cases where the expansion and contraction of surface-saturated areas is important (and considered in the model), knowledge of the maximal portion of the catchment that might become saturated can be used.…”