Toward Data-Informed Process-Based Models of Watershed Function


Sergi Molins1* (, Kate Maher3, Zexuan Xu1, Dipankar Dwivedi1, Tristan Babey2, Lijing Wang2, J. David Moulton3


1Lawrence Berkeley National Laboratory, Berkeley, CA; 2Stanford University, Stanford, CA; 3Los Alamos National Laboratory, Los Alamos, NM



Aggregate measures of mountainous watershed responses, such as concentration–discharge (C-Q) relationships, are available from point measurements at gauged streams. However, this response is the result of the complex coupling between hydrologic, biogeochemical, and ecological processes across the different watershed subsystems, including hillslopes and floodplains, and across scales. Spatially resolved (or distributed) mechanistic models play an important role in quantitatively understanding and predicting this response. For an accurate system representation, models must consider the gamut of processes affecting energy and mass balances. At the same time, process-rich models must be able to incorporate the heterogeneous datasets derived from field campaigns, as well as geological and machine learning models, at different scales of observation. In turn, model insights will then be used to inform the need for additional observations in a classical ModEx loop.

This presentation illustrates how to use workflow and simulation tools in the IDEAS-Watersheds software ecosystem to incorporate a variety of datasets and perform simulations of integrated hydrology and reactive transport. Specifically, these tools are applied to the Lower Triangle Region (LTR) use case in partnership with the Watershed Function science focus area (SFA), a headwater-dominated catchment of the East River, Colorado, that comprises hillslope and floodplain subsystems.

Watershed Workflow is used to mesh the simulation domain with variable resolution following the streams in LTR, and to process datasets of vegetation cover, organic carbon, and subsurface geological materials to assign heterogeneous properties to the mesh at the appropriate resolution. Simulations are performed with the Advanced Terrestrial Simulator (ATS) and constrained by integrated C-Q data both at the upstream and downstream.

At finer scales, floodplain architectures dictate exchange between streams and surrounding groundwater systems. Strong vertical layering of fine-grained sediments over coarse-grained units can create strong redox gradients, where anoxic conditions in fine-grained units exchange with underlying oxic gravel aquifers. The degree of exchange can mobilize colloids and release metals into the stream. Expanded treatments of carbon, iron (Fe), and sulfur redox cycling, with an emphasis on Fe-colloid formation and mobilization, are being developed in partnership with SFA work through the SLAC National Accelerator Laboratory at the Slate River, Colorado. These improved fine-scale process models form the basis upon which process complexity is being expanded in watershed models.