7 Geologic mapping in the digital age: products, workflows, and methodologies (Posters)

This session will highlight geologic maps and their derivatives, showcasing products that leverage innovative digital methods and workflows, novel application of remote sensing data, and integration of advanced data processing techniques.

Sunday, 22 March 2026
8:00 AM - 12:00 PM
CCC, Ballroom C

Constructing Georeferenced HDPC of Butler Cave as Dataset to the Advance Machine Learning Applications in Cave and Karst Science

Paper number: 7-0
Program: Technical Sessions
Time: 8:00 AM – 8:00 AM

Authors

Abstract

Modern mapping with computer aid for karst features has provided the opportunity to develop more in-depth surveys than traditional mapping. Traditional mapping requires a broad understanding of caving techniques and time investment, unlike the modern techniques of LiDAR. Previous research proves that machine learning paired with LiDAR was successful in finding cave openings in wooded areas, supporting the idea of using a similar process in analyzing caves.

 

Butler Cave (BC), located in Bath County, VA, is home to many unique features within its eighteen-plus miles of passage including: maze passages, speleothem growth, bare cave passage, strata and faulting. The number of features make BC an ideal site for machine learning algorithms. Several applications including but not limited to enhance surveys of the general cave geometry, speleothem diversity, or to understand the general ecology. This study aims to apply collected georeferenced High-Definition Point Cloud’s (HDPC) of BC to train machine learning.

 

Initially, we constructed a georeferenced HDPC of more than 1.8 x 109 points using a handheld LiDAR-SLAM (Scanning Localization and Mapping) GeoSLAM horizon system. This HDPC extends for more than 1.5 km of cave passage mapped with more than 100 m of gradient. In addition, a georeferenced LiDAR survey of the surface has been paired with internal cave surveys for machine learning algorithms to refine current interpretations on how surface features such as hydrology, epikarst characteristics, or dolines could be affecting internal observations. Constructing an algorithm with the capability of analyzing data for specific features is possible by mass feeding it ‘raw’ (bare) HDPC’s, then HDPC’s that contain anomalies like speleothems. The algorithm will distinguish unique features based on the shape and dimensions of anomalies not apparent in bare HDPC’s. This allows the machine learning algorithm to sort through HPDC’s in a set containing unique features, or the possibility to highlight specific anomalous features compared to bare maps.

 

Digital mapping anomaly-sorting machine learning creates the ability to analyze karst features in-depth with the accuracy of HDPC’s. This opens opportunities for developing advanced tools for cave and karst sciences and mapping processes. Further works will include accurate speleothem identification, creating precise methods for cave mapping and generating higher HDPC quality.  Machine learning is merely the beginning, foreshadowing advances of future exploration, mapping, and karst geology analysis.

Geological Society of America Abstracts with Programs. Vol. 58, No. 2

doi: 10.1130/abs/2026NE-411135

© Copyright 2026 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.