Times are displayed in (UTC-05:00) Central Time (US & Canada) Change

61-3 Rapid Eruption Reconstruction via a Novel Image Segmentation Approach at the Valles Caldera

Session: Advancing Geologic Analysis with Digital Outcrops and Close-Range Remote Sensing Data (Posters)

Poster Booth No.: 70

Presenting Author:

Clare Brosnan

Authors:

Brosnan, Clare¹, Clarke, Amanda², Boyer, Ellie³, Arrowsmith, Ramon⁴, Chen, Zhiang⁵

(1) School of Earth and Space Exploration, Arizona State University, Tempe, AZ, USA, (2) School of Earth and Space Exploration, Arizona State University, Tempe, AZ, USA; Istituto Nazionale di Geofisica e Vulcanologia - Sezione di Pisa, Pisa, Italy, (3) Ira A. Fulton Schools of Engineering, Arizona State University, Tempe, AZ, USA, (4) School of Earth and Space Exploration, Arizona State University, Tempe, AZ, USA, (5) Department of Geological and Planetary Sciences, California institute of Technology, Pasadena, CA, USA,

Abstract:

Understanding the characteristics of volcanic eruptions provides insight into their hazard potential. However, this often relies on time-consuming sample collection and analysis, particularly in the case of eruptions outside of the historical record. Current standard procedure for classifying explosive eruptions via their fall deposits involves measuring unit thickness, bulk sampling clasts, iterative sieving , and taking measurements for the three principle axes of the 20 largest clasts at each site. Such observations may be required at tens of sites. Implementing this for the 170 potentially active volcanoes in the US (USGS Volcanic Hazards Program) and their various deposits would be impractical at mass scale or in a short amount of time at a single volcano experiencing unrest. This project explores the viability of image segmentation as a method of rapid data collection for tephra outcrop characterization.

We use the “Segment Anything Model” by Meta AI (SAM) to automatically segment individual clast and deposit boundaries in images of outcrops representing the two uppermost El Cajete pumice units of the Valles Caldera, New Mexico (~75ka). Clast volumes and unit thicknesses are estimated based on the 2-dimensional “masks” created by SAM to represent objects identified in images. This is done with the intent of finding three key pieces of information at each sample location: the maximum clast size, the total grain size distribution, and the total thickness of the unit in question. This information is then used to quantify the magnitude (total volume) and intensity (mass eruption rate) of the eruptions and to identify eruption style. To evaluate the success of this method we compare our results to those using traditional methods which characterize the events as plinian eruptions of 0.5 - 3 km3 with plume heights of ~25km above the vent, and mass eruption rates of nearly 108 kg/s.

Initial results show that SAM is capable of consistently identifying and accurately masking individual clasts in images of deposits, but the quality of results depends on outcrop characteristics (such as grain size), image quality, image preparation, and tunable SAM parameters. We seek to develop guidelines for quality image acquisition as well as customization recommendations for various outcrop characteristics and object identification goals.

Geological Society of America Abstracts with Program. Vol. 57, No. 6, 2025

doi: 10.1130/abs/2025AM-7623

Rapid Eruption Reconstruction via a Novel Image Segmentation Approach at the Valles Caldera

Description

Session Format: Poster

Presentation Date: 10/19/2025

Presentation Room: HBGCC, Hall 1

Poster Booth No.: 70

Author Availability: 3:30–5:30 p.m.

Back to Session