Evans, Alexandra
Submitted to the University of New Hampshire in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy in Earth and Environmental Sciences.
DISSERTATION ABSTRACT:
Administrative, technical, and financial barriers often prevent sufficient collection of data for stream restoration project evaluation (Roni & Beechie, 2012; Bernhardt et al. 2005; NRC 1992). This lack of evaluation and understanding of restoration impacts can lead to the same misinformed strategies being repeated across restoration sites (Sommerville & Pruitt, 2004). This is particularly relevant for dam management, as the river ecosystem response of dam management strategies, like removal, are not fully understood due to minimal pre-/post-removal studies (Foley et al. 2017; Hart et al. 2002; Poff & Hart, 2002). Stream restoration practitioners need ecological assessment approaches that are affordable, repeatable, objective, and logistically feasible to develop science-based restoration techniques. This dissertation demonstrates how ecological evaluation workflows that use consumer-grade drone imagery coupled with structure-from-motion (SfM) photogrammetry provide an affordable, feasible, adaptable, and flexible solution to many of the challenges facing stream restoration, improving the frequency, accuracy, precision, and coverage of evaluations and improving our knowledge of ecological impacts from restoration practices. The first chapter explores how illustrative drone products (videos, orthomosaics, and 3D models), made using a simple structure-from-motion photogrammetry workflow, can be coupled with a visual stream ecological assessment protocol to provide a remote visual evaluation and ecological condition archive approach. The second chapter explores the intersection of drones, close-range remote sensing, and data science. The work demonstrates how hybrid feature sets derived from the drone RGB orthomosaics and digital surface models can be used to accurately map and quantify riparian vegetation structure via machine learning algorithms for conventional classification tasks that focus on classifying a single site, providing sufficiently accurate results with workflows amenable to use by restoration practitioners. The third chapter demonstrates how drone workflows for mapping and quantifying riparian vegetation structure as well as erosion and deposition throughout fluvial environments can be used to evaluate and monitor changes pre-/post-dam removal, using the Sawyer Mill dam removal project in Dover, NH, USA as a case study. The limitations and efficacies of the drone approaches vs. conventional ecological approaches are compared, and the study demonstrates how the drone approaches leverage the drone’s aerial perspective to provide holistic ecological evaluation data at a landscape scale.