Author ORCID Identifier

Year of Publication


Degree Name

Doctor of Philosophy (PhD)

Document Type

Doctoral Dissertation


Arts and Sciences


Earth and Environmental Sciences (Geology)

First Advisor

Dr. William C. Haneberg


This dissertation uses digital terrain modeling and computational methods to yield insight into three topics: 1) evaluating the influence of glacial topography on fluvial sediment transport in the Teton Range, WY, 2) integrating regional airborne lidar, UAV lidar, and structure from motion photogrammetry to characterize decadal-scale movement of slow-moving landslides in northern Kentucky, and 3) applying machine learning methods to surficial geologic mapping.

The role of topography as a boundary condition that controls the efficiency of fluvial erosion in the Teton Range, Wyoming, was investigated by using existing lidar data to delineate surficial geologic units, geometrically reconstruct the depth to bedrock, and estimate the sediment volume and sediment production rate in two catchments. This data was coupled with seismic reflection data in the bay into which these catchments drain. We found that while the sediment production rate of 0.17 ± 0.02 mm/yr is similar to the uplift rate of the Teton Range, only about 2.6% of the post-glacial sediment has been transported out of the catchments, and the denudation rate is just 0.004 ± 0.001 mm/yr. We conclude that once the topography has been altered by glaciers, which flatten the valley bottom and steepen the valley walls, rivers are incapable of evacuating the sediment effectively. Sediment will be trapped in the valleys until the next glacial advance, or until uplift steepens the system such that rivers can once again become efficient.

Repeat digital terrain surveys can be used to quantify changes to the Earth’s surface. Challenges include determining the threshold of change that can be detected when combining topographic data acquired by different platforms and of varying quality. To quantify the threshold of detectible elevation change in a slow-moving colluvial landslide in northern Kentucky over 14 years using county-wide lidar, uncrewed aerial vehicles (UAV) structure from motion surveys (SfM) and a UAV lidar survey, we used the statistics of noise from elevation difference maps in areas outside of the landslide. We found that the threshold of detectable elevation change ranges from 0.05 to 0.20 m, depending on the survey combination, and that detectable change in the landslide was found between all surveys, including those separated by only 2 weeks.

For most users, geologic maps may convey a level of certainty which obscures the decisions and interpretations made by the mapper. The combination of machine learning and digital terrain data provides a new method for producing geologic maps which can also convey and preserve the underlying uncertainty. We test the performance of machine learning methods to accurately map the surficial geology of two quadrangles in Kentucky using 31 variables derived from lidar data, including surface roughness, slope, topographic position, and residual topography. The performance of eight machine learning methods were compared, and the importance of each variable was measured. The classifier with the highest accuracy using just the most important variables was used to produce surficial geologic maps in 6 areas, with resulting accuracies ranging from 0.795 to 0.931. The uncertainty resulting from the machine learning process is conveyed using gradations of color, which can be modified depending on the needs of the map user.

Digital Object Identifier (DOI)

Funding Information

This study was supported by the National Science Foundation Grant (NSF-EAR 1932808) in 2020, the University of Kentucky Earth and Environmental Science DepartmentFerm Fund in 2019, and the University of Kentucky Earth and Environmental Science Department Overcash Field Fund in 2018.