Improving Parameter Inferences in a Multiscale Model for Transport in Stream Corridors

Abstract:

Physical processes in the hyporheic zone have significant impacts on observed breakthrough curves (BTCs) in the form of spreading, retardation, and development of a long tail, which are encoded in tracer test results. We explored ways to extract information from these BTC features in order to improve the quality of the process inferences from the conservative tracer data. Parameter-estimation experiments were performed to systematically investigate the effects of tracer-test design and inverse problem setup on the reliability of the estimated parameters. A newly proposed multiscale model (Painter S. L., WRR, 54-10, (2018),7216-7230) with the subgrid representation of hyporheic zone processes in the Lagrangian framework implemented in Advanced Terrestrial Simulator (Coon et al., 2019, DOI: 10.11578/dc.20190911.1) was used for inverse modeling. For the first time, shape-free travel time distribution was estimated together with other parameters of the stream corridor model using Markov Chain Monte Carlo technique unraveling the parameter interactions and tradeoffs in process representations. We found that the development of the BTC tail is a strong signal containing key information about the hyporheic-zone processes. Hence, a longer reach to allow development of a prominent tail, measurement duration adequate to capture a significant part of the observed tail, and choice of a source without a significant tail to capture the tail development were found to reduce the parameter-ambiguity significantly. These insights were used to efficiently utilize the observations from multiple locations in the reach by jointly fitting all BTCs with a single set of hyporheic zone parameters and section-specific channel areas and dispersion coefficients. This allow us to honor the known heterogeneity in the stream while reconciling the spatial scale at which hyporheic zone processes manifest in the field with the inverse model. The obtained parameters provided accurate prediction potential throughout the reach length along with an improved representation of the heterogeneous field conditions. This study offers guidance to designing efficient tracer tests and setting up inverse problems for reliable parameter estimates and process inferences.