Remote survey of large-scale braided, gravel-bed rivers using digital photogrammetry and image analysis

Abstract

The use of conventional survey methods to monitor large, gravel river beds has traditionally led to a reliance on repeat measurements of cross-sections which, unless very closely spaced, may give unreliable information about three-dimensional channel morphology and morphological change. Provided certain technological limitations can be overcome, remote survey techniques, such as digital photogrammetry and airborne laser scanning, remove the spatial and temporal constraints typically associated with ground-based surveys, allowing high spatial resolution, distributed, elevation mapping of gravel river beds. This paper develops the use of digital photogrammetry for the survey of a 3.3 km reach of the braided Waimakariri River, New Zealand, which, when combined with image analysis of water colour to infer water depth, provides a Digital Elevation Model (DEM) of the entire river bed. Central to the successful application of this method is DEM post-processing. Errors take two forms: (i) individual point errors associated with incorrect stereo-matching during automated data collection; and (ii) spatially-variable systematic errors that are associated with uncertainties in sensor position and orientation as determined during the bundle adjustment. An automated post-processing procedure is developed to deal with individual point errors and this improves DEM surface quality in terms of accuracy, precision and internal reliability. Systematic errors in the final DEM surface were reduced by applying a simple correction based on surveyed photo-control point elevations.