- Recommended articles
- August 17, 2025
[Frontier Letter] Double difference earthquake location with graph neural networks
Double difference earthquake relocation is an essential component of many earthquake catalog development workflows. This technique enables locating earthquake catalogs with high precision, which helps to more accurately reveal underlying fault structures and seismic processes. Traditional methods rely on linearized gradient descent optimization, however these approaches are expensive for large catalogs. Here, we propose a Graph Neural Network (GNN)-based earthquake double-difference relocation framework, Graph Double Difference (GraphDD), that can relocate very large catalogs and minimize the double-difference residuals through a supervised learning process. In several real world and synthetic tests, the model obtains high resolution locations, suggesting this approach is a promising direction for earthquake relocation.
- McBrearty and Beroza (2025): [Frontier Letter] Double difference earthquake location with graph neural networks. Earth Planets Space 77:127, https://doi.org/10.1186/s40623-025-02251-4
![[Frontier Letter] Double difference earthquake location with graph neural networks](https://www.earth-planets-space.org/wp-content/uploads/2025/08/40623_2025_2251_Figa_HTML-1-80x40.webp)
![[Frontier Letter] Balancing of geodetic and seismic moment rates and its implications for probabilistic seismic hazard analysis in Taiwan](https://www.earth-planets-space.org/wp-content/uploads/2025/07/Unknown-1-80x24.png)
![[Frontier Letter] A new paleomagnetic constraint on the duration of a twin caldera-forming eruption sequence: a potential solution to an enduring problem in decoding >VEI 7 eruption timescales](https://www.earth-planets-space.org/wp-content/uploads/2025/05/40623_2025_2179_Figa_HTML-75x40.webp)
