Evaluating ecohydraulic model sensitivity to remotely-sensed river bathymetry

Ecohydraulic models are fundamental tools for understanding species-habitat interactions across scales, and are widely used to inform water management decisions under current and future environmental conditions. Advances in remote sensing have enabled continuous, high-resolution mapping of river bathymetry, which is an essential input for ecohydraulic model development. Despite progress in remote sensing techniques, little is known about how errors in remotely-sensed river bathymetry propagate through ecohydraulic model predictions. In this study, we used a multispectral satellite image and hyperspectral imagery acquired from manned and unmanned aircraft to derive river bathymetry, and fused these data with near-infrared LiDAR data to generate three different hybrid, digital elevation models (DEMs). The DEMs were used to build three versions of a 2D flow model (Delft3D) for a 1.6-km reach of a large (> 100 m wide), gravel-bedded river. We evaluated the model performance of each bathymetry source through a comparison of measured vs. predicted hydraulics, and spawning, rearing and holding habitat for Pacific salmon. We found that model depth and velocity errors were less than 20% of the measured, reach-averaged values. Non-dimensional spawning and rearing habitat indices, with values between zero and one, had errors less than 0.1 for all sensors. Spatial patterns in modeled adult holding habitat were similar between each bathymetry, though peak, predicted metabolic costs varied between sensors, due to the nonlinear relationship between fish energy expenditure and water velocity. Our results indicate that bioenergetic calculations related to holding costs were more sensitive to model input bathymetry, and these errors could impact predictions of total energy expenditure from migration to spawning life stages. Conversely, hydraulic variables and spawning and rearing habitat metrics were less sensitive to bathymetric source, with predicted errors comparable to previously published 2D modeling studies developed using ground survey and bathymetric (green) LiDAR. Each of the image-derived bathymetric data sets we evaluated had advantages and limitations for parameterizing 2D models, which should be considered in potential modeling applications on large rivers.