GSA Annual Meeting in Indianapolis, Indiana, USA - 2018

Paper No. 172-22
Presentation Time: 9:00 AM-6:30 PM

QUANTITATIVE STRUCTURE-FROM-MOTION MODELING FOR MONITORING CHANGE IN COASTAL BOULDER DEPOSITS


NAGLE-MCNAUGHTON, Timothy P., Geosciences, Williams College, Williams College, Williamstown, MA 01267 and COX, Rónadh, Geosciences, Williams College, Williamstown, MA 01267

Coastal boulder deposits (CBD) are supratidal accumulations of wave-deposited boulders, occurring as isolated clasts weighing many tonnes or imbricated boulder ridges. Little is known about the rate at which CBD accumulate, whether they aggrade or erode over time, or how often new boulders are created and added to the deposits. We applied drone-based photogrammetry to monitor CBD evolution by building centimeter-resolution structure-from-motion (SfM) models in Agisoft PhotoScan and comparing a time-series of these models using the open source software CloudCompare. We applied this technique to boulder ridges on Inishmore, Aran Islands and on Valentia Island (west of Ireland).

Comparing a 2015 SfM model of the Inishmore site to a 2017 model revealed substantial change, despite the lack of large storms during that time interval. The seaward edge of the boulder ridge was extensively rearranged. In an area ~0.03km2 in size and ~8m above high water, CloudCompare detected displacement of ~200 individual boulders weighing 10s of kg to 28t, which moved distances of decimeters to >15m. We corroborated the CloudCompare results by manually comparing the SfM models, drone images, and field photographs.

SfM models of CBD on Valentia Island from July 2017 and November 2017, a period of zero change, served as a control for CloudCompare’s change-detection algorithms. CloudCompare correctly detected no reworking and no volume change between the two image sets, down to the ~3cm/pixel resolution of the models. This result confirms CloudCompare’s accuracy, and indicates that our Aran Island results did not include false-positive detections.

PhotoScan and CloudCompare together proved to be reliable and useful tools for detecting and quantifying CBD movement. Using CloudCompare to analyze PhotoScan models is more comprehensive and more objective than manual comparisons of models or photographs, and our testing on Valentia Island demonstrates the accuracy of the CloudCompare results. Together, these programs can facilitate systematic and detailed observations of CBD in the future.