GSA Annual Meeting in Indianapolis, Indiana, USA - 2018

Paper No. 155-10
Presentation Time: 11:00 AM

THE OPPORTUNITIES AND CHALLENGES OF MODELLING WATERSHED-SCALE RIVER MORPHODYNAMICS WITH SFM PHOTOGRAMMETRY


BRASINGTON, James, Te Waiora Institute for Freshwater Management, University of Waikato, Private Bag 3105, Hamilton, 3240, New Zealand

The emergence of Structure-from-Motion (SfM) photogrammetry has revolutionized the availability of high resolution terrain models, democratizing datasets that were historically the preserve of national mapping agencies. This approach employs a non-linear bundle adjustment to solve simultaneously camera position, pose, distortion and 3D coordinates using imagery from highly redundant camera networks. Unlikely conventional aerial photogrammetry, SfM is can be applied to imagery from low-cost, non-metric cameras captured from a wide range of terrestrial and aerial platforms, enabling flexible deployments tailored to specific science requirements. To date, most applications have focused on relatively small spatial scales (0.1-1 km2), where logistics permit dense ground control and close-range photography. Broad-scale or regional mapping with SfM remains comparatively rare and it is unclear whether watershed scale projects that extend over 10-100 km2 can achieve similar or even improved results in terms of survey density, accuracy and precision than those more commonly obtained with airborne laser ranging.

In this presentation we provide a systematic assessment of the quality of SfM terrain products derived for a c. 100 km2 corridor of the braided Dart River in the Southern Alps of New Zealand. Two surveys are presented, the first following a large headwater landslide in 2014 and a further survey one year later, with the aim of using DEM differencing techniques to model channel response. The results demonstrate the challenges of developing wide-area terrain models using SfM. In particular, reconstructions were found to exhibit a complex pattern of systematic errors associated with compensation between the inferred interior and exterior camera geometry. Refinements to the survey strategy show, however, that these effects may be partially mitigated using a convergent camera network that incorporates both nadir and oblique imagery from a single camera rig. The results presented are used to synthesize a series of recommendations that aim to guide future regional surveys based on SfM photogrammetry.