UC Merced

River confluences present a challenging environment for both data collection and hydrodynamics and advection-dispersion modeling. The Merced River-San Joaquin River provides a site with distinct water salinity signatures where advection-dispersion modeling can be readily tested. This work describes the application of a robotic delivery system for water velocity and specific conductivity measurements whose infrastructure is sufficiently agile to enable analysis of several cross-sections along the confluence within a week-long study. The volumetric water flow estimates from the cross-sectional water velocity field were comparable (within 10%) to those recorded at a nearby gaging station. With the river bed elevation and water surface elevation determined by echo-sounding and surveying, a 2-D, finite element hydrodynamic model is parameterized with additional fitting parameters including bed roughness coefficients and eddy viscosities. Various parameter adjustment scenarios were undertaken to calibrate the RMA2 hydrodynamic model, and the resulting steady velocity field was then used as input for the advection dispersion model RMA4. The best results in RMA2 were automatically assigning the roughness coefficient by depth and a single, isotropic eddy viscosity value of 35 Pa-s for the entire model domain. This approach resulting in an average absolute percent difference (AAPD) between the modeled and observed lateral velocity profile for the furthest downstream crosssection of 31.87%, a -2.47% difference between modeled and observed water surface elevation, and a best qualitative shape-fit agreement between the modeled and observed lateral velocity profile. These results could be improved by quicker elevation data collection methods that are more closely synchronized with the times of the velocity/flow data collection. The estimated eddy viscosity value is close to the theoretical value (34.69 Pa-s) estimated given the downstream cross-sectional dimensions and flow conditions. For the advection-dispersion modeling, the longitudinal dispersion coefficient of 0.1 m2/s and lateral dispersion coefficient of 0.01 m2/s assigned to the entire model domain generated the best fit, resulting in a 7.78% AAPD between the modeled and observed lateral concentration (salinity) profile for the furthest downstream cross-section, and a best qualitative fit shape-fit agreement between the modeled and observed lateral concentration profile. These fitted values are within the range of theoretically estimated longitudinal (0.06 to 0.13 m2/s) and are similar lateral dispersion coefficients (0.014 to 0.015 m2/s) based on the upstream San Joaquin River cross-sectional dimensions and flow conditions. This work demonstrates that high-resolution data collection coupled with appropriate model settings can provide reasonable 2-D hydrodynamic and advection-dispersion simulations in complex river environments, such as confluence zones. The models developed here are useful for informing reservoir operation and water quality management decisions in the SJR basin.