Date of Award


Document Type


Degree Name

Doctor of Philosophy (PhD)


Forestry and Environmental Conservation

Committee Chair/Advisor

Christopher Post

Committee Member

Elena Mikhailova

Committee Member

MZ Naser

Committee Member

Mark Schlautman


New technologies and applications of technologies are critical to protecting environmental and water resources. Students need to be aware of these new technologies so they can be prepared to utilize them in their future careers. One such technology is using Unmanned Aerial Vehicles (UAVs) because it can be applied to a wide variety of fields such as engineering, construction, wildlife biology, agriculture, and many more. Chapter two discusses the creation of an online teaching module to introduce students to using UAVs in natural resource research and evaluates how well students respond to the education module. Overall, student familiarity with UAVs and remote sensing increased after completing the online teaching module, and students had positive remarks about the introduction meaning this was an effective way to teach students about UAVs, despite not having access to a physical UAV, helping reduce the barriers to UAV education. The study also provides useful suggestions to educators for designing their own customized educational UAV content for future students.

Urban streams face increased threats due to climate change such as rising water temperatures, degraded water quality, and increased flooding events. Therefore, stream managers need innovative technologies and methodologies to study urban watersheds to combat these negative pressures.

Machine learning is one technique that can provide greater insight and help forecast both water quality and water quantity for urban streams. First, various machine learning algorithms including k-nearest neighbor, decision tree, random forest, and gradient boosting were used to predict dissolved oxygen levels to serve as an indicator of water quality in the urban stream, Hunnicutt Creek, in Clemson, South Carolina USA. These algorithms were applied to four different locations in the urban watershed, and random forest had the greatest prediction accuracy with NSE scores > 0.9 at three of the four sites and > 0.62 at the fourth site. Additionally evaluating the landcover of the sub watersheds that drain to each station location indicated that the varying amounts of impervious surface and vegetation impacted the water quality and resulting dissolved oxygen predictions.

The time-series machine learning algorithm, Prophet, was also used in the Hunnicutt Creek watershed to predict changes in water quantity to provide stream managers with information about the potential for flooding throughout the area. Machine learning was applied to five locations along Hunnicutt Creek to predict future hourly changes in water level, resulting in models with R2 values greater than 0.9 in all locations indicating that this algorithm can be successfully used in hydrological forecasting. These change in water level predictions were then translated into 3D visualizations of areas of the stream channel to model areas of high-water during precipitation events. This was done by using terrestrial LiDAR to capture high-resolution topography of the channel. Water volume calculations were also calculated to provide stream managers with more information about what areas are prone to higher risks of flooding. Areas with greater human influences such as culverts and greater amounts of surrounding impervious surfaces experience greater increases in water level and water volume and are more prone to flooding. Together, machine learning, 3D models, and volume calculations identify what areas of the watershed are likely to flood.

The methodologies presented in this research provide researchers with new ways to examine urban streams and to make resiliency plans to protect the watersheds from the negative impacts of climate change.

Author ORCID Identifier


Available for download on Saturday, August 31, 2024