Evaluating the Ecological Impacts of Dam Removal Using Drones, Structure-from-Motion, and Machine Learning

Authors: Alexandra Evans*, University of New Hampshire, Kevin Gardner, University of Louisville, Scott Greenwood, University of New Hampshire, Brett Still, University of Rhode Island, Denise Burchsted, Keene State College
Topics: Environmental Science, Remote Sensing, UAS / UAV
Keywords: drone, UAV, sUAS, UAS, stream, river, restoration, dam removal, quantitative methods, ecological evaluation, structure-from-motion, method development, close-range remote sensing, machine learning, deep learning
Session Type: Paper
Presentation File: No File Uploaded


Many stream restoration projects’ outcomes are not evaluated (Bernhardt et al. 2005; Roni & Beechie 2013; Nilsson et al. 2016), despite having conventional ecological assessment methods available. Stream restoration professionals need ecological assessment approaches that are affordable, repeatable, objective, and efficient to develop science-based restoration techniques and better understand how to improve the ecological health of our waterways. This work develops stream ecological assessment approaches using small unmanned aerial systems, structure-from-motion (SfM) photogrammetry, and machine learning methods to increase our understanding of restoration impacts and improve the stream professional’s toolkit. Specifically, this research is creating a suite of models in GIS for analyzing drone orthomosaics and digital surface models for various stream ecological metrics. Ecological metrics such as vegetation quantity, vegetation quality, habitat complexity, substrate, and other metrics analogous to those used in conventional stream ecological assessment protocols are measured in GIS and compared to conventional field data. This toolbox will provide stream ecologists, restoration practitioners, and other stream professionals an approach to assessing the ecological condition of their sites in an efficient, repeatable, objective, quantitative, and transparent way. The Sawyer Mill dam removals (Dover, NH) provide an example of how this drone-based analytical workflow can be applied to stream restoration projects to evaluate changes over time. In addition, the changing geomorphology at the dam removal site was evaluated using the digital surface models produced via SfM and compared to conventional cross-sections and unconventional transects throughout the site.

To access contact information login