Computational optimization in simulating velocities and water-surface elevations for habitat–flow functions in low-slope rivers

Abstract

ABSTRACTThe impact of varying computational mesh discretization on the accuracy of simulating velocities and water-surface elevations was investigated using the River2D model and data from 10 study reaches in three low-slope (<0.2%) Canadian rivers. A wide range of computational aspects were examined, including node spacing (1.35 to 60 m), number of mesh nodes (1250 to 58,000) and domain widths (60 to 800 m). Computed values for average cross-sectional velocities and water-surface profiles were compared with corresponding field survey data. The statistical mean absolute and root-mean-square errors were used to evaluate the discrepancy between measured and simulated values due to the mesh discretization characteristics. The results showed that mesh design discretization had a pronounced effect on the precision of the velocity and water-surface elevation predictions. Although the ratio of node spacing to river reach width may be site specific, it was found that optimal values should be <0.022 to obtain the highest accuracy. Regression equations of optimal ratios of node spacing to river reach width and the corresponding minimum mesh resolution errors were estimated. An example is provided that illustrates how the choice of computational mesh properties can affect habitat characterization for fish.