Numerical Modelling of Braided Rivers with Structure-from-Motion-Derived Terrain Models

摘要

The development of three-dimensional reconstructions of channel morphology has historically been limited by the high costs of geospatial data collection and software modelling. Advances in image processing, sensor technology and portable remote-sensing platforms, however, now offer the opportunity to derive survey quality terrain models at significantly reduced cost and without traditional deployment and logistical constraints. There is a pressing need to establish whether new geospatial technologies such as structure-from-motion photogrammetry can be used to deliver topographic data products that are suitable for higher-dimensional hydrodynamic modelling. To address this question, we evaluate the results of simulations using Delft3D that were designed to model distributed, depth-averaged flows in a wide, shallow, gravel-bed braided river. The topography for these simulations was derived from digital elevation models (DEMs) generated using structure-from-motion and optical bathymetric mapping of two linked reaches of the Ahuriri River, New Zealand. The DEM quality achieved vertical surface errors of 0.10m in non-vegetated areas and 0.20m in inundated areas. Simulations with 1.5m and 2.5m resolution grids for low-flow, medium-flow and high-flow conditions were calibrated and tested against field real-time kinematic-global navigation satellite system observations. Results revealed that modelled depth errors were comparable to the DEM uncertainty, while simulated and observed inundation patterns achieve a maximum of 81% agreement. Given the complexity of the braided network and shallow flow depths, these simulations provide a powerful demonstration of the suitability of these terrain models for hydrodynamic applications. Copyright (c) 2015 John Wiley & Sons, Ltd.