Numerical dispersion of solute transport in an integrated surface–subsurface hydrological model

Integrated surface–subsurface hydrological models (ISSHMs) are increasingly being used for the assessment of contaminant transport in the environment, in addition to their more common use in water flow applications. However, the subsurface solute transport solvers in these models are prone to numerical dispersion errors. Numerical dispersion is a well-known issue in groundwater modeling, but its impacts on the results of ISSHM simulations are still poorly understood. In this study, the CATchment HYdrology (CATHY) model is used to assess the potential impacts of numerical dispersion on the simulation of coupled surface–subsurface solute transport. We first simulate the subsurface transport of a nonreactive tracer in two soil column test cases (1D and 3D) with known analytical solutions. The subsurface solute transport solver in CATHY adopts a computationally efficient time-splitting technique whereby the advection component of the governing equation is solved on elements and the hydrodynamic dispersion component is solved on nodes. Comparison between simulation results and analytical solutions with different mesh discretizations and different values for the hydrodynamic dispersion coefficients allows for accurate quantification of the numerical dispersion error and yields insights into the parameters and other factors that control it. It is shown that, taken alone, the advection and dispersion solvers are very robust, but their combination can result in significant numerical dispersion, stemming from the exchange of concentration information from elements to nodes and vice versa in the time-splitting procedure. The tests also show that these errors can be kept under control by ensuring that the grid Péclet number is in the range 0.5-1.0 or smaller. We then apply CATHY in a third test case involving two synthetic hillslopes (concave and convex) in fully coupled surface–subsurface mode, in order to examine the impact of this subsurface numerical dispersion on simulated streamflow hydrographs, in particular with reference to pre-event water contributions to runoff. Here as well the results show that the effect of numerical dispersion can be controlled by keeping the grid Péclet number sufficiently small. This work provides a new set of benchmark test cases for integrated surface–subsurface hydrological models, extending to solute transport the flow-only suite of benchmarks recently published in two intercomparison studies.