Accessed online March 5, 2025
4th Resource Publication
MACHINE LEARNING LINEAMENT CASE STUDY: THE AFAR TRIANGLE
Dan Kalmanovitch (1,2), Phil Harms (1,2) and Trent Dinn (3),
(1) 4th Resource Corp., (2) GEOSEIS Inc., (3) Gno-Sys Technology Ltd.
Calgary, CANADA
dan.kalmanovitch@4thresource.ca
ABSTRACT
This case study presents the results from an algorithm developed to use a machine learning classifier to detect surface structural lineaments, as expressed by topography, within the Afar triangle of Ethiopia. The Afar triangle is a region with active continental rifting and high geothermal resource potential. Surface lineaments support characterization of deeper faults and fractures which may be associated with geothermal resources. Tectonically significant faults typically play a critical role in facilitating deep fault circulation of hot geothermal fluids.
To utilize machine learning for a lineament detection application, an algorithm was developed to (1) prepare geospatial data as training data, (2) implement a machine learning classifier to detect lineaments, and (3) convert the results into useful formats.
The predictions from a machine learning model are, for most cases, evaluated against truth datasets to assess predictive accuracy. However, for this case study, the truth lineament datasets themselves are being evaluated as well. If the truth lineaments are pretty good in following obvious intuitive patterns in the surface topography, the model will yield adequate predictions which can be visually evaluated on maps.
Once intuitive lineaments are achieved, this approach allows for interpretations over larger areas that may be impractical to interpret manually. The predicted lineaments provide a useable dataset for a geoscientist to focus in and analyse lineaments along azimuths prone to known, favourable stress directions or for any geological application.
Proceedings of the 10th African Rift Geothermal Conference, Dar es Sallam, Tanzania, 13-15 November 2024