The COMPASS-GLM project has a long-term vision of improving the predictive understanding of the interdependent co-evolution of coastal regional processes and human systems using the Great Lakes region as a testbed. Toward that end, the project is constructing a regional Earth system model of the Great Lakes region with application-appropriate resolution and process detail in atmosphere, lake, and land models.

The land model in our Great Lakes regional model will couple the 3D integrated surface/surface hydrology model Amanzi-ATS to the E3SM Land Model (ELM). In the two-way coupled design, ELM represents surface biophysical processes and passes water and energy fluxes and transpiration demand to Amanzi-ATS, which then represents water and energy flows in 2D on the surface and in 3D in the subsurface. In this talk, we will show prototype simulations on a hillslope, compare to conventional ELM simulations that neglect lateral flow, and discuss near-term and long-term plans to build out this capability.

To address specific environmental issues associated with the Great Lakes, specifically degraded eutrophication and harmful algal blooms in the western basin of Lake Erie caused by nutrient runoff from agricultural lands, we have also extended Amanzi-ATS to represent hydrology and reactive transport in agricultural watersheds. The effect of subsurface tile drains, which are ubiquitous in the agricultural watersheds contributing to the western basin of Lake Erie, is represented through a physically based subgrid model that calculates flow to tile drains placed below the root zone and then redistributes that water to the nearest stream or agricultural ditches. A new mixed polyhedral meshing strategy is used to explicitly represent engineered streams and drainage ditches with coarser triangulated meshes between streams. Taken together, these new capabilities make it possible to capture the effects of artificial drainage while remaining tractable at river basin scales. Without calibration, high-resolution multiyear simulations of the Portage River Basin, OH, agree well with observed streamflow. Reduced need for calibration improves confidence in these simulations compared to existing models for agricultural lands, especially for projections in a nonstationary climate. We show that tile drains reduce peak discharge from large events and quantify how flow paths through the watershed are modified by drainage. Preliminary simulations of nutrient transport will also be shown.