Date of Award
Fall 2023
Document Type
Open Access Dissertation
Department
Geography
First Advisor
Scott White
Abstract
Locating hydrothermal chimney sites unlocks research into their correlations with geology, lithospheric cooling, and deep-sea biogeography at seafloor spreading ridges. High-resolution bathymetry and sidescan sonar collected by Automated Underwater Vehicles allows for chimneys, only a few meters wide and tall, and fissures, only a few meters wide, to be resolved across large areas (>100 km2). We developed a Chimney Identification Tool (CIT) that utilizes a Convolutional Neural Network (CNN), a Machine-Learning model able to classify based on shapes and textures, to identify chimneys in 1m-gridded bathymetry. We then utilized the CIT to identify many potential off-axis chimney structures at the East Pacific Rise from 9° 43’ to 9° 57’N. This served as the first CIT classification from an area that was not used in model training, and the discovery of abundant off-axis chimneys has important implications for how fluid flow, water-rock reactions, and biological communities develop over a zone wider than the typical (SEPR) 16° 30’-18°S. This provided a quantitative measure of fissure density to pair with water-column sensor data and geologic interpretations to investigate changes in geologic environment along the SEPR and how they relate to hydrothermal venting.
Rights
© 2024, Isaac Keohane
Recommended Citation
Keohane, I.(2023). Adapting Deep Learning Techniques for Geologic Investigations of Hydrothermal Venting at Seafloor Spreading Ridges Using AUV Surveys. (Doctoral dissertation). Retrieved from https://scholarcommons.sc.edu/etd/7625