Research · AI project2022 – 2024.02

Seismic interpretation of large overseas CO2 storage sites considering plate-tectonic history (first-author paper)

Reconstructed how large overseas CO2 storage sites formed and interpreted their seismic images from a structural-geology perspective

Geophysical Computing Lab (undergraduate researcher) · Undergraduate researcher · first author

First-author paper ↗

Problem

Korea needs ~30 Mt/yr of domestic CO2 storage by 2050, but there was little systematic analysis of how large storage sites form geologically

My role

As first author, compare large CO2 storage sites in four regions from a plate-tectonic viewpoint (submitted 2023.07, accepted 2023.12)

What I did

Reconstructed the plate-tectonic history of four regions with GPlates (PALEOMAP PaleoAtlas)
Interpreted seismic images of two representative CCS projects per region — Quest & Weyburn-Midale, Northern Lights & PORTHOS, Gorgon & Moomba, Uthmaniyah & Qatar LNG — focusing on reservoir and cap rock
Short stay at Univ. of Calgary (2023.12–2024.02): processed gravity data from the Utah FORGE geothermal site

Results

First-author paper: Analysis on Seismic Interpretation for Overseas Large-scale CO2 Storage Considering Geological History Related to Plate Tectonics, Geophysics and Geophysical Exploration 27(1), 1–22 (2024)
Showed that stratigraphy in each seismic image reflects the region's tectonic history → a basis for rejecting geologically unrealistic velocity models when processing domestic data
1 second-author paper (CarbonNet CCS, Gippsland Basin, 2023) · 3 conference talks (2022–2023)

Decisions & trade-offs

What I chose, and what I gave up for it

Instead of going deep on one site, I compared eight projects in four regions with the same frame (plate-tectonic history → reservoir and caprock structure), prioritizing common patterns over per-site quantitative analysis.

Looking back · open questions

What I would change, and what is still open

Looking back: even when results look good numerically, 'it makes sense once you know how it formed' is the stronger evidence — the role domain constraints play in ML. It became the habit of comparing with independent models in my M.S. work.

Applying the proposed criterion (ruling out geologically unrealistic velocity models) to real domestic survey data remains open.

Photos & figures

2022 — conference badges (KSEG · K-CCUS)

2024.04 — Utah FORGE poster presentation