CAN DRONE IMAGERY BE USED TO PRODUCE 3D MODELS FOR EROSION ANALYSIS?: AN EXPERIMENT AT MCINTYRE’S BLUFFS, STERLING, NY

Focused erosion along the southern edge of Lake Ontario has dissected drumlins left behind as glaciers melted at the end of the Wisconsinan Stage of the Laurentide glaciation (~10 Ka). McIntyre’s Bluffs in Sterling, NY, contains several exposures of dissected drumlins whose eroded till, sand, and gravel produces steep faces, rivulets, and spires. The degree of erosion along these bluffs is dictated by wave action, precipitation, ice, and groundwater, and can vary considerably over distances of less than several hundred yards. While repeat visits to the bluffs indicate continued erosion at a rate noticeable with a simple visual inspection, a more exact and detailed method is necessary to quantify the rate of change. Although ground-based LiDAR has been shown to accurately determine modern erosion rates, LiDAR instrumentation can be difficult to use and data collection is very expensive. In this study, we demonstrate that imagery collected from a DJI Phantom drone can be used to create 3D models of a bluff face in order to document erosional changes at the bluff over time. Drone video and image data were collected at a portion of McIntyre’s Bluffs each month, from September to December 2018, and from January to March 2019. Waypoints were established to create a drone flight route along the face of the bluff. This ensured that all videos recorded over the course of the study were similar in total elapsed time, and that imagery was collected over the same path. Image frames were extracted from drone videos in AGI Photoscan and, in combination with individual photos collected from regular cameras, were used to generate a point cloud, a mesh, and finally a texturized 3D model of the bluff surface. 3D models created for each month of data collection were imported into ArcGIS, and differenced. Differenced images provide both a qualitative and quantitative measure of erosion rate on the bluff surface. This study has provided a proof-of-concept for easily and inexpensively generating 3D models over short time scales to track rapid erosional changes.