Evaluation of high-resolution, true-colour, aerial imagery for mapping bathymetry in a clear-water river without ground-based depth measurements

This study evaluates the potential of the hydraulically assisted bathymetry (HAB-2) model coupled with true-colour, three-band, aerial imagery to map bathymetry in the clear-water McKenzie River, Oregon, USA, in the absence of ground-based depth measurements. It is the most rigorous test of the HAB-2 model to date. Correlation-coefficient (r(2)) values for sonar depths versus modelled depths are 0.40 for 2007, 10 cm resolution imagery. Overall depth trends follow those of sonar data, except in areas where there are shadows, riffles or obstructions that block the view of the river (e. g. overhanging trees, bridges). Low-pass filtering of the image to remove film granularity does little to improve the results, although an Olympic filter improves the r(2) value from 0.40 to 0.48. The moderate fit of the model results to sonar data in 2007 may also result from the 28-39 day gap between sonar and image acquisition, during which time, the discharge changed. HAB-2 depth estimates for the 2008 0.5 m imagery fit the depth measurements more closely (r(2) 0.89). The better fit may reflect the collection of ground and image data at approximately the same time and discharge, as well as coarser spatial resolution, which created less sensitivity to changes in substrate size and colour. The results suggest that the best depth-estimate results for the HAB-2 model are for depths ranging from 0.25 and 1.5 m. Use of digital imagery collected with digital cameras should also improve accuracies. Results indicate that the HAB-2 is useful for characterizing the approximate depths throughout the river channel if one avoids shadows, riffles and obstructions.