Content area

Abstract

In wetland resource management, drones, also known as uncrewed vehicle systems, offer unique advantages for monitoring and ensuring regulatory compliance at wetland mitigation sites. However, before introducing drone monitoring to evaluate performance standards on these sites, agencies need to understand the fundamental processes of working with drone-derived data, how these data compare to that collected via traditional field methods, and the respective advantages and disadvantages of drone-based monitoring. This study evaluated two critical components of the classification process: classification method and machine learning algorithm. I assessed overall accuracy and woody vegetation class accuracy at six mitigation sites in western Washington. These sites encompassed wetland, buffer, and riparian zones, and ranged in age from 3 to 10 years. Results from four classification trials indicated that object-based classifications consistently outperformed pixel-based approaches. The choice of machine learning algorithms, whether support vector machines or random trees, had no significant impact on accuracy metrics. I then compared woody cover measurements from drone- and field-derived data and analyzed results at the mitigation zone and sample plot scales. At the zone scale, I represented drone cover as the percentage of woody cover across the entire zone and field cover as the average woody cover of all plots. At the plot scale, I expressed drone cover as the percentage of woody cover within each plot, and field cover as the raw data collected in each plot. Differences in mean woody cover estimates between methods ranged from marginally significant to significant at the zone and plot scales. While correlations were significantly positive at the zone scale, they were weaker and more variable at the plot scale. These findings suggest that drone methods underestimate woody cover, and further refinement is needed to improve agreement between methods at the plot level. This study highlights the potential for drones to help site managers meet monitoring demands, particularly when resources are limited. A key drawback is the upfront time investment required to learn processing and analysis techniques. Overall, these findings contribute to developing guidelines for government agencies to adopt drone-based monitoring using off-the-shelf equipment and user-friendly methods that prioritize practical implementation, affordability, and accessibility.

Details

1010268
Title
Evaluating Uncrewed Vehicle Systems (UVS) for Regulatory Monitoring of Compensatory Mitigation Sites
Number of pages
152
Publication year
2025
Degree date
2025
School code
0250
Source
MAI 86/10(E), Masters Abstracts International
ISBN
9798310397279
Committee member
Moskal, L. Monika; Halabisky, Meghan
University/institution
University of Washington
Department
Environmental and Forest Sciences
University location
United States -- Washington
Degree
M.S.
Source type
Dissertation or Thesis
Language
English
Document type
Dissertation/Thesis
Dissertation/thesis number
31847611
ProQuest document ID
3193536410
Document URL
https://www.proquest.com/dissertations-theses/evaluating-uncrewed-vehicle-systems-uvs/docview/3193536410/se-2?accountid=208611
Copyright
Database copyright ProQuest LLC; ProQuest does not claim copyright in the individual underlying works.
Database
ProQuest One Academic