San Francisco Estuary and Watershed Science

We investigate the fidelity of the Delta Simulation Model-2 (DSM2), a one-dimensional branched network hydrodynamics solver, which is used to model water quality and ecology in the Sacramento–San Joaquin Delta estuary. We find that while DSM2 reproduces the total flows well, it does not accurately represent the harmonic components of the tides and tidal modulation of subtidal flow. The inaccurate representation of tidal dynamics affects prediction of subtidal flows, flow splits at key junctions, and salinity. These deviations are the result of coarse spatial and temporal representation of tides as well as unrepresented estuarine physical processes. We propose and evaluate two types of schemes intended to improve fidelity: modifying the model domain and specifying fine grid and boundary conditions, and incorporating and parameterizing more complex physical processes into the 1-D model. We also develop a comprehensive protocol to evaluate the model in which we assess the fidelity of model results. In this protocol, we also include a decomposition of the model error into a systematic component because of model representation, and an unsystematic component, which includes errors from both unmodeled physical processes and data precision. Our analysis reveals that these recommendations would be effective provided they can be incorporated with model recalibration. Both our proposed schemes and the model evaluation process will be useful in analyzing models of networked surface water systems such as the Delta in which the distribution of observations is spatially inhomogeneous.