FLYING UNDER THE CANOPY: SUCCESSES AND FAILURES IN CAPTURING A RETREATED WATERFALL WITH DRONES AND HANDHELD SFM PHOTOGRAMMETRY

We mapped a retreated waterfall under forest canopy in central New York using imagery derived from drones and handheld cameras. Recent advances in structure from motion (SfM) photogrammetry have enabled sub-meter scale process-based studies of geomorphic features. Most published studies have utilized more expensive cameras and drones for data collection. We chose to explore the capability of lower end systems. These include three drones of different cost levels, as well as handheld cameras. The imagery data was converted into 3D point clouds and aligned to surveyed ground control points. Drones proved to be a much quicker data collection device than handheld cameras, but had a few drawbacks. Lighting and temperature were the biggest issues confronting photographic data collection. Fortunately, the gorge was mostly shaded, but bright sun in places caused reflections and saturated pixel. Drones required delicate control to weave around tree cover. Low light and slow camera responses often resulted in blurry images, even in the higher resolution cameras. With the low temperatures, the higher the resolution of the camera on the drone, the slower the drone moved in order to capture a clear snapshot. Nonetheless, data collected with all three drones compared favorably with 1 meter lidar elevation data from a state survey in 2006. The hand cameras produced the best models, but was very time consuming and dangerous in comparison to the drones. Although handheld cameras produced the best models, drones had the capability to capture images from many different angles and showed features which are not visible from ground level, and they were an easier platform from which to collect imagery. Overall, we found the lower resolution of drone imagery was offset by flexible aerial perspectives. Low end drones can produce 3D models with a moderate level of simplicity, and at a resolution (10s to 1000s of points per square meter) useful for process-based studies of geomorphic features.