Welcome to SeismoScan
SeismoScan is a senior capstone project focused on developing a machine learning system designed to automatically detect active submarine faults from bathymetry data. Traditional fault-mapping techniques require geologists to manually trace faults over extremely large regions, which is time-intensive and difficult to scale.
Our goal is to streamline this process by creating an automated hybrid machine learning pipeline that highlights fault-like structures, groups related features, and exports meaningful geological outputs for further study.
Why This Project Matters
Mapping faults on the seafloor is essential for understanding major Earth processes such as:
- Earthquakes
- Tsunamis
- Tectonic plate movement
- The deep water cycle
As more high-resolution bathymetry becomes available, automated analysis tools become increasingly important for supporting geological research.
Project Overview
SeismoScan uses a hybrid machine learning approach that combines:
- Convolutional Neural Networks for spatial feature detection
- Clustering algorithms to group continuous structures
- Random Forest classification for verification
By integrating these methods, we aim to create a reliable and scalable solution for identifying fault-like regions in seafloor topography.
Bathymetry & Fault Detection
Bathymetry — the measurement of seafloor depth — reveals natural ridges, valleys, and linear structures. Many of these features are related to tectonic plate activity. SeismoScan analyzes bathymetric grids, extracts spatial patterns, and identifies abrupt depth changes that may correspond to fault structures.
Meet the Team
SeismoScan is developed by:
- Noah Carges
- Alyssa Sombrero
- Thomas Rotchford
- Brady Wisniewski
Visit the Team Page for individual roles and photos.
Sponsors & Mentor
- Dr. Donna Shillington
- Dr. James Gaherty
- Dr. Christine Regalla
Mentor: Scott LaRocca