182 Expanding Geology’s Horizons: Geoinformatics, Open Science, and Open Data (Posters)

This session highlights the development of new methodologies for collecting, sharing, and analyzing geological datasets and the advancement of FAIR (findable, accessible, interoperable, and reusable) data practices.

Tuesday, 21 October 2025
8:00 AM - 5:30 PM
HBGCC, Hall 1

Exfoliation Dome Distribution: Leveraging crowdsourced data and AI for geomorphic feature detection.

Paper number: 182-0
Program: Technical Sessions
Time: 8:00 AM – 8:00 AM

Authors

Abstract

Exfoliation domes and their associated surface parallel sheet joints are geomorphically significant features, yet they remain underrepresented in geologic datasets, limiting landscape-scale analyses. A spatially-informed inventory of exfoliation domes could advance geomorphological research, understanding of dome characteristics, and the processes that lead to sheet jointing. Manually identifying domes over broad geographical regions is, however, time-intensive and impractical. Automating their locations using topographic data or geologic data - as has been done for features such as river terraces - would preclude identifying domes that do not fit known examples.  The goal of this work was to develop a semi-automated workflow to locate exfoliation domes in the landscape. 

In the southeastern US, exfoliation domes occur at a variety of scales (often called ‘balds’ or ‘barrows’) beyond the obvious localities (i.e., Stone Mountain, GA). These are clearly visible features characterized as broad, bare rock surfaces with common surface-parallel sheet joints. Thus, we set out to identify exfoliation domes across a region known to have such domes (North Carolina) by integrating geotagged social media images, AI filtering, and spatial analysis in a GIS environment. We surveyed over 20 platforms with publicly available geotagged content; Google Maps, Flickr, and OpenStreetMap (OSM) emerged as the most promising sources. A custom Python script was developed to mine images and metadata based on a set of curated keywords. To assess dome relevance, we used a CLIP-based image-text model developed by OpenAI to convert images into numeric embeddings, which were then classified using a logistic regression model trained on a set of images. The workflow with the best results filtered thousands of crowdsourced photos to identify relevant imagery, with roughly 75% classification accuracy that identified most domes. Inaccuracies stemmed primarily from overidentification, resulting in false positives, rather than missed features. 

In ongoing work, we are analyzing the validated results (36 total domes) by comparing their topography, soil, and geologic information to explore whether exfoliation domes can be spatially predicted using terrain attributes derived from GIS-based raster analysis.

Beyond this study, our workflow illustrates how crowdsourced data can expand the scope of geomorphic research and make regional-scale data collection more thorough and efficient. Our process is reproducible, adaptable to other regions, and highlights the role of public data in accelerating scientific discovery in the Earth Sciences. 

Geological Society of America Abstracts with Programs. Vol. 57, No. 6

doi: 10.1130/abs/2025AM-412920

© Copyright 2025 The Geological Society of America (GSA), all rights reserved. Permission is hereby granted to the author(s) of this abstract to reproduce and distribute it freely, for noncommercial purposes. Permission is hereby granted to any individual scientist to download a single copy of this electronic file and reproduce up to 20 paper copies for noncommercial purposes advancing science and education, including classroom use, providing all reproductions include the complete content shown here, including the author information. All other forms of reproduction and/or transmittal are prohibited without written permission from GSA Copyright Permissions.