Hydrodynamic and sediment transport model of San Francisco Bay, California, Nov-Dec 2014

A three-dimensional hydrodynamic and sediment transport model of San Pablo and Suisun Bays was constructed using the Delft3D4 (D3D) modeling suite (Deltares, 2021a) to simulate water levels, flow, waves, and suspended sediment for time period of Nov 1 to Dec 31, 2014. The San Pablo and Suisun Bay model was nested within an overall model for San Francisco Bay, also built using D3D. The overall model was modified from the published tide model for San Francisco Bay (Elias and others, 2013). The 2D tide model was extended to 3D with 10 vertical sigma layers (representing 2,3,5,8,13,20,20,20,7, and 2 percent of the total water depth, from near-bed to near-surface). A non-tidal residual based on water-level observations at NOAA site 9414290 (San Francisco) was added to the tidal harmonics of the published model to force the oceanic boundaries. These observations were low-pass filtered with a 66-hr cutoff to remove tidal fluctuations. The overall model was spun up from Jan 1 to Nov 1, 2014, with salinity at the river boundaries of 0 PSU and 33 PSU in the ocean. Wind forcings from the COAMPS model for Northern California (Naval Research Laboratory, 2020) were applied to the overall model, but the wave model (SWAN) was not turned on. Freshwater inputs through the rivers were derived from USGS Stations: the Sacramento River at Freeport (USGS gauge 11447650), the San Joaquin River at Stockton (USGS gauge 11304810), and the San Joaquin River at Vernalis (USGS gauge 11303500). Three sediment types were included in the overall model: mud (ws = 0.2 mm/s, ρ = 2650 kg/m3), flocs (ws = 1 mm/s, ρ = 1300 kg/m3), and sand (D50 = 0.2 mm, ρ = 2650 kg/m3). Sediment availability on the bed was prescribed based on the bathymetry: at depths shallower than 5 m, 1.05 m of mud was available on the bed, 2.55 m of flocs, and 0.05 m of sand. In deeper parts of the system, there was 0.05 m each of mud and flocs, and 1.1 m of sand. Sediment inputs at the rivers were set as constant values: 0.03 kg/ m3 for flocs, 0.02 kg/ m3 for mud, and 0 kg/ m3 for sand. Roughness in the model was prescribed using a Manning’s n value of 0.0225. Following the spin-up period, the overall model was run from Nov 1 to Dec 31, 2014 to prescribe the boundary conditions for the nested model.

The nested model was forced with Riemann boundary conditions computed from the overall model at the Richmond boundary. At the Sacramento and San Joaquin River boundaries, total discharge was set to that of the overall model. Salinity and sediment concentrations at the Richmond boundary were based on a spatial and temporal interpolation of values in the overall model to the nested grid. Salinity at the Sacramento and San Joaquin River boundaries was set to 0 ppt, and sediment was input as a linear profile with concentrations derived from the overall model. Freshwater was supplied at the Petaluma, Sonoma, and Napa Rivers, using observations of river outflow, Q, from USGS gauges (11459150, 11458500, and 11458000, respectively). Time- and space-varying wind conditions were applied to the nested runs based on COAMPS (see above). Waves were implemented through two-way coupling with the SWAN model (Deltares, 2021b) with Madsen bottom friction, assuming a roughness coefficient of 5 x 10-4 m. Sediment availability on the bed was initialized with a single bed layer. The model was allowed to spin up over 3 months; the evolved bed following this spin up period was used as the initial condition. The initial availability of flocs and sand is shown in the "Sand Fractions" folders of the model inputs.

The model was calibrated to minimize errors in water levels through the hydrodynamic roughness, using observational data from NOAA stations at San Francisco (station number 9414150), Richmond (9414863), Martinez-Amorco (9415102), and Port Chicago (9415144) (National Oceanic and Atmospheric Administration, 2020), using common tidal constituents (Pawlowicz and others, 2002). Using Manning's n of 0.0225 in the overall model, tidal propagation of the M2, K1, and O1 constituents through the north bay adequately matched the observed values through a 3-month period. Other tidal constituents, including S2 and N2 also followed the observations.

To evaluate model performance, we compared the observed water depth, velocity, salinity, wave conditions, and suspended sediment concentration at sites N1T, S1T, and M2T (Allen and others, 2019) with modeled values at the same location with the time lag, normalized unbiased root-mean-square-deviation (ubRMSDN), the normalized bias (BiasN), and amplitude ratio. Modeled water levels in the shallows were within 10-20 cm of the measured values. The amplitude ratio for water levels was close to 1 at each site in the shallows, and the time lag between observations and models was under 6 minutes. BiasN and ubRMSDN were both less than 0.5 for each location, indicating excellent performance for water level. Velocities in the east and north directions performed reasonably well between the observations and the model, with BiasN between -0.14 and 0.33, ubRMSDN between -0.52 and 0.33 (with an exception at 0.97, for site M2T in the east direction), and amplitude ratio between 0.68 and 1.34.

Performed minor edits to the metadata to correct typos. No data were changed