Multiresponse modeling of variably saturated flow and isotope tracer transport for a hillslope experiment at the Landscape Evolution Observatory
- 1Institut National de la Recherche Scientifique, Centre Eau Terre Environnement (INRS-ETE), Université du Québec, Quebec City, Canada
- 2Department of Mathematics, University of Padua, Padua, Italy
- 3Department of Geosciences, Georgia State University, Atlanta, USA
- 4Laboratory of Ecohydrology, École Polytechnique Fédérale de Lausanne, Lausanne, Switzerland
- 5Biosphere 2 – Earth Science, University of Arizona, Tucson, USA
- 6Department of Hydrology and Water Resources, University of Arizona, Tucson, USA
Abstract. This paper explores the challenges of model parameterization and process representation when simulating multiple hydrologic responses from a highly controlled unsaturated flow and transport experiment with a physically based model. The experiment, conducted at the Landscape Evolution Observatory (LEO), involved alternate injections of water and deuterium-enriched water into an initially very dry hillslope. The multivariate observations included point measures of water content and tracer concentration in the soil, total storage within the hillslope, and integrated fluxes of water and tracer through the seepage face. The simulations were performed with a three-dimensional finite element model that solves the Richards and advection–dispersion equations. Integrated flow, integrated transport, distributed flow, and distributed transport responses were successively analyzed, with parameterization choices at each step supported by standard model performance metrics. In the first steps of our analysis, where seepage face flow, water storage, and average concentration at the seepage face were the target responses, an adequate match between measured and simulated variables was obtained using a simple parameterization consistent with that from a prior flow-only experiment at LEO. When passing to the distributed responses, it was necessary to introduce complexity to additional soil hydraulic parameters to obtain an adequate match for the point-scale flow response. This also improved the match against point measures of tracer concentration, although model performance here was considerably poorer. This suggests that still greater complexity is needed in the model parameterization, or that there may be gaps in process representation for simulating solute transport phenomena in very dry soils.