How do we efficiently generate high-resolution hydraulic models at large numbers of riverine reaches?

Abstract

In support of efforts to quantify relationships between juvenile salmonid habitat and population dynamics in the Pacific Northwest, over 2200 hydraulic models were generated at more than 900 individual reaches with unique bathymetry. Hydraulic models generated two dimensional field estimates of depth and velocity for each survey, providing a key linkage used to relate bathymetry and habitat data to juvenile salmonid population dynamics. Generating more than 2200 hydraulic models required development of an automated process to generate input files specifying bathymetry, computational grids, and boundary conditions for the Delft3D Flow software (which we run in 2D, and hereafter refer to as “Delft Flow” for clarity), enabling batch-processing of large numbers of hydraulic models, which is the novel advancement we present here. Hydraulic model inputs included digital elevation models (DEM) from topographic surveys, estimates of surface roughness based on pebble size distributions, and discharge. Outputs included velocity vector and depth fields estimated on a rectilinear grid of 10 cm spacing between grid points. Modeled velocities and depths were in reasonable agreement with field-collected velocities and depths. Certain topographic features, such as undercut banks and porous structures not represented in the DEM, resulted in modeled values that failed to reflect accurate velocities but were explained by the presence of these features. By utilizing a rectilinear grid, scaling grid spacing to computational resource limitations, leveraging a cloud computing system, and selecting simplified rules for discharge distribution for boundary conditions and model run times, we were able to successfully automate the hydraulic modeling process. Overall, automation of hydraulic model generation met precision and accuracy needs of habitat condition models, lowered labor costs, and standardized the modeling workflow, and enabled high survey volume processing needs.