Advancing UAV-Based Remote Sensing and Machine Learning for Ecological Monitoring, Habitat Disturbance Detection, and Forensic Science in Arid Landscapes
Loading...
Authors
Best, Jeffery
Issue Date
2025
Type
Dissertation
Language
en_US
Keywords
conservation , desert environments , drones , machine learning , remote sensing , Unmanned aerial vehicles
Alternative Title
Abstract
The advances in unmanned aerial vehicle (UAV) technologies have the potential to revolutionize remote sensing. By integrating high-resolution low-altitude remote sensing data captured via UAVs with sophisticated machine learning algorithms, researchers can now analyze habitat disturbances, vegetation cover, and even the detection of threatened and endangered species with unprecedented precision. This synergy not only facilitates assessments of vegetation cover and health but also enhances the detection of anthropogenic impacts such as land degradation caused by off-highway vehicle (OHV) activity. Furthermore, these advancements hold significant promise for forensic science applications to assist in search and recovery, cold-case investigations, and disaster response. The deployment of UAV-based systems equipped with multispectral capabilities allows for comprehensive spatial analysis that can substantially improve management decisions and reduce workloads. We deployed multiple drone and camera systems at study sites within the arid deserts of the western United States to demonstrate low-cost, user-defined, scientific, and civil applications. In chapter one we use a UAV-based multispectral camera to classify vegetation with a goal of detecting individual plants of the endangered Peirson’s milkvetch (Astragalus magdalenae var. peirsonii) (PMV) within the Algodones Dunes, California. In chapter two, we use UAV imagery and innovative computer vision analyses to detect OHV tracks that disrupt fragile dune ecosystems; such modeling is crucial for understanding human impact on arid biomes and informing management practices. Finally, chapter three focuses on forensic applications, as we apply multi-spectral analyses via a UAV collection platform to detect surface skeletal remains in the Great Basin Desert—illustrating how low-cost aerial methods can serve multifaceted purposes in both conservation efforts and forensic investigations amid desert environments.
Our American deserts have often been ignored or simply perceived as obstacles enroute to more hospitable regions in riparian zones or along the coast, however, modern improvements in living conditions and immense growth in urban populations have increased interaction between humans and the desert landscapes of the west. Many more people enjoy the benefits of living and working in the American deserts than ever before, and this contact continues to place a heavy burden on fragile ecosystems that include many endemic species unique to the Sonoran, Mojave, and Great Basin Deserts. Our research, focused on arid landscapes, emphasizes the challenges unique to our deserts which require specific remote sensing solutions. Many of the desert locations in this study are hampered by direct access and the difficult nature of the rugged terrain creating a need for continuing development in drone technology to facilitate research in these regions. Reduced levels of moisture and high heat levels in summer require distinct remote sensing solutions as well as opportunities to conduct research not typical in other regions, such as our skeletal surface remains detection. Finally, the deserts decades-to-centuries long recovery cycle from fire, ground clearing, and disturbance heightens the growing awareness of the threat to these fragile ecosystems and the need for continuing research. Each project in this dissertation underscores both an interactive and automated approach to studying desert landscapes through user-oriented technologies tailored in the pursuit of advancing conservation and scientific efforts within arid landscapes.