QUANTIFYING BANK EROSION USING STRUCTURE FROM MOTION PHOTOGRAMMETRY

Fine-grained sediment is a major source of pollution in the Chesapeake Bay and its tributaries due to its ability to attenuate light, smother habitat, and sorb and transport nutrients such as phosphorus and particulate nitrogen. Piedmont streams in the Mid-Atlantic region of the United States are frequently characterized by incised channels with steep and highly erodible banks of legacy sediment that can contribute to high sediment loads in these watersheds.

The quantification of bank erosion is essential to producing accurate sediment budgets, which are proving useful to inform management strategies and the remediation of degraded fluvial systems. A number of methods are outlined in the literature to measure bank erosion at various spatial scales. This investigation compares the use of traditional bank pin measurements and structure-from-motion photogrammetric techniques to quantify streambank erosion in the agricultural Little Catoctin Creek in Central Maryland, USA.

Bank pin measurements, representing only a single point in space, were found to vary based on the subjective initial placement of bank pin locations and completely missed nearby large bank failures. Photogrammetric techniques using Structure from Motion, in contrast, were able to capture a more spatially-complete profile of a streambank and capture changes at a scale that a single bank pin would be unable to detect. Using a Nikon D810 camera and AgiSoft Photoscan software, bank scans were able to reconstruct banks with a RMSE as low as 0.1mm and repeat scan alignment resolution of <2mm. However, photogrammetry exhibited some coverage gaps in areas of high vegetation density during summer months.

This spatial data, when coupled with bank sediment bulk density, may be used to accurately quantify volumetric change as well as sediment loads originating from bank erosion. This is a powerful, emerging tool for the generation of sediment budgets, and may provide researchers with higher resolution and spatial coverage in comparison to traditional methods.