Crowd-sourced SfM: Best practices for high resolution monitoring of coastal cliffs and bluffs

Digital oblique photos of an approximately 2.0 km alongshore reach of seaward-facing coastal bluff faces on the Strait of Juan de Fuca, Washington State, were collected at least quarterly between 2016 and 2022. In total over 3900 photos were collected over 38 separate surveys. The photos were collected to support digital surface reconstructions using Structure-from-Motion (SfM) photogrammetry, and specifically to assess the quality of surfaces generated with photos collected using relatively simple techniques that could be accessible to community science programs. A secondary goal was to use the photos, and the digital surfaces generated with them, to evaluate patterns and rates of erosion on the bluff face. 

Methods
Photos were collected in one of two ways. The first method was with a point-and-shoot digital camera hand-held by a photographer on a boat, approximately 100-200 meters from the toe of the coastal bluff. Individual photos were manually collected approximately every 2 seconds as the boat moved along the shoreline at speeds of approximately 2-3 knots, with the camera facing straight towards the bluff face. The second method was with a point-and-shoot digital camera, collected by a photographer standing at the base of the beach, around the mean low water elevation contour, at a distance of approximately 30-50 meters from the toe of the coastal bluff. For photos collected on the beach, photos were collected approximately every 10-15 meters in alongshore distance along the reach. Of the 38 photo sets, 3 were collected with no spatial information associated with each photo and 13 were collected with camera position supplied only by the cameras internal GNSS sensor. For the remaining 22 photo sets a ProMark 200 GNSS system, operating in Real-Time Kinematic mode (RTK-DGPS) was used to assign a high-accuracy position and elevation to the camera. For 7 of the photo sets camera positions are only know from the GNSS systems, whereas 15 of the photo sets include positions from both the camera's internal GNSS sensor and from the GNSS system. For those photos that were collected with high accuracy location and elevation information the camera was set on to a platform attached to a rover pole and positioned directly under the GNSS antennae, providing an estimated horizontal and vertical uncertainty for the camera position of approximately 15 cm. Three different handheld digital cameras were used during the project, a Canon D20 Powershot, a Fujifilm Finepix XP130, and a Motorola Moto G(6).