Earth & environment · OECD FoS 1.5 · China MoE 0708 地球物理学 / 0709 地质学 / 0707 海洋科学
Earth sciences (seismology, geology, oceans)
地球科学（地震/地质/海洋）
L2Local breakthroughs → Steady Score 2.20 · Last reviewed 2026-10-22 · Next review 2027-01-19
BIGGEST BREAKTHROUGH 2020-08-07 AI · majorField · major
EQTransformer detection & picking (Nat. Commun.) ↗
Joint detection and picking; standard tool
Why this level
Deep-learning phase picking (PhaseNet, EQTransformer) is standard for building catalogues; in 2026 AI gives aftershock risk in seconds, but individual large-quake prediction remains unreliable.
S = 0.30·method + 0.35·discovery + 0.25·adoption + 0.10·education (education scores 0 when there is no evidence; if empty, the other weights are renormalised)
Sources
- doi.org/10.1093/gji/ggy423
- nature.com/articles/s41467-020-17591-w
- bgs.ac.uk/news/new-research-shows-ai-earthquake-tools-forecast-aftershock-risk-in-seconds
Composite 1.5–2.49: at least one peer-reviewed/officially verified AI-key result, or one dimension at 3; routine workflow largely unchanged. All levels →
Research directions
Phase picking & cataloguesAftershock forecastingGenerative ground motionOcean forecasting
Papers
- Global seamount detection using machine learning
AI: trained on ship/gravity-derived seamounts with GEBCO 2025 input. Field: updates the global seamount census with height/area/volume, relevant to ocean circulation, ecology and intraplate volcanism
2026-10-01 · Earth and Planetary Science Letters
- Earthquake Aftershock Forecasting using Conditional Generative Models
Beats operational spatial forecasts on 2024–25 events
2026-07-27 · arXiv
- Upper-mantle earthquakes beneath East Antarctica
AI: transfer-enhanced deep learning detects events missed by conventional methods. Field: intraplate intermediate-depth quakes support lithosphere–asthenosphere models and may be more common than recognised
2026-05-28 · Science
- Earthquake transformer—an attentive deep-learning model for simultaneous earthquake detection and phase picking
Joint detection and picking; standard tool
2020-08-07 · Nature Communications
- PhaseNet: A Deep-Neural-Network-Based Seismic Arrival Time Picking Method
Picks beat rule-based methods; widely used
2018-10-13 · Geophysical Journal International
- Deep learning of aftershock patterns following large earthquakes
High-profile, but later shown matched by simple baselines
2018-08-29 · Nature
- Physics-based deep learning for non-stationary adaptive multiple suppression
Physics-based deep learning improves adaptive subtraction of surface-related multiples in marine seismic processing.
2026-10-13 · arXiv · Chao Li, Yangkang Chen
- Event-Centric Earthquake Monitoring with a Multi-Station Transformer
EQNet2, a multi-station transformer, jointly predicts arrivals, polarities, location, magnitude and event association.
2026-10-12 · arXiv · Haoyu Wang, Weiqiang Zhu
- Fault-conditioned Seismic Image Generation using Denoising Diffusion Probabilistic Modeling and Neural Style Transfer
A DDPM framework generates realistic fault-conditioned seismic images to augment training data.
2026-10-07 · arXiv · Tolulope Agbaje, Nam Pham, Altay Sansal, Yangkang Chen
- An AI-assisted conditioning and geological interpretation workflow for usage in implicit geological modeling
An ML toolkit speeds interpretation of onshore seismic data (300–3,500 m) for implicit geological modelling.
2026-10-07 · arXiv · Stefan Carpentier, Jan Diederik van Wees, Eva de Boever, Jan Niederau
- Real-world application of deep learning in large-scale seismic interference attenuation: A case study in the Camie field of Angola
Field case applying a DNN workflow to attenuate seismic interference over 345+ km² offshore Angola.
2026-10-07 · arXiv · Jing Sun, Song Hou
- Multi-Modal Surface Wave Inversion using Physics Guided Variational Autoencoders
A physics-guided multimodal VAE for scalable near-surface surface-wave inversion.
2026-10-01 · arXiv · Sujith Swaminadhan, Andrey Bakulin, Sergey Fomel, Yangkang Chen
- GeoFWI3D: Large-scale 3D Velocity Model Dataset for Deep Learning-assisted Seismic Imaging
An open benchmark of geologically plausible 3D velocity models for deep-learning-assisted seismic imaging and FWI.
2026-10-01 · arXiv · Sujith Swaminadhan, Yiran Shen, Chao Li, Kai Gao
- Elasticity of polycrystalline davemaoite constrains its grain size in the lower mantle
Million-atom machine-learning MD reconciles davemaoite’s shear modulus and constrains its grain size in the lower mantle.
2026-09-30 · arXiv · Peiyu Zhang, Liang Yuan, Rong Huang, Yu Ye
- From arXiv, bioRxiv, medRxiv and OpenAlex (last ~30 days); summaries written by AIMap from the abstract.
Milestones 8
Big glowing nodes are leaps: AI leap cool · Field leap warm · double ring = both
-
Paper
Global seamount detection using machine learning: 1.8–2.7× more seamounts than prior ground truth, 28,434 rising ≥1 km (EPSL)AI: trained on ship/gravity-derived seamounts with GEBCO 2025 input. Field: updates the global seamount census with height/area/volume, relevant to ocean circulation, ecology and intraplate volcanism
Earth and Planetary Science Letters -
Paper
Berkeley Lab CGM-GM generative ground-motion simulationGenerative ground-motion simulation
Berkeley Lab -
Paper
QuakeGen: conditional generative aftershock forecasting beats USGS operational spatial forecastsBeats operational spatial forecasts on 2024–25 events
arXivPreprint -
Discovery
Deep learning reveals upper-mantle intraplate earthquakes beneath East Antarctica (Science)AI: transfer-enhanced deep learning detects events missed by conventional methods. Field: intraplate intermediate-depth quakes support lithosphere–asthenosphere models and may be more common than recognised
Science -
Paper
BGS: AI tools forecast aftershock risk in secondsComparable to ETAS, in seconds
British Geological Survey -
System
EQTransformer detection & picking (Nat. Commun.)Joint detection and picking; standard tool
Nature Communications -
System
PhaseNet deep-learning phase picking (GJI)Picks beat rule-based methods; widely used
Geophysical Journal International -
Paper
Deep learning of aftershock patterns following large earthquakes (Nature)High-profile, but later shown matched by simple baselines
NatureContested
No milestones match this filter.
More papers & repos 6
| Type | Title | Date |
|---|---|---|
| Paper | Earthquake Aftershock Forecasting using Conditional Generative Models Preprint | 2026-07-27 |
| Paper | Earthquake transformer—an attentive deep-learning model for simultaneous earthquake detection and phase picking | 2020-08-07 |
| Paper | PhaseNet: A Deep-Neural-Network-Based Seismic Arrival Time Picking Method | 2018-10-13 |
| Paper | Deep learning of aftershock patterns following large earthquakes | 2018-08-29 |
| Repo | AI4EPS/PhaseNet To be verified | — |
| Repo | smousavi05/EQTransformer To be verified | — |
Latest news 3
- BridgeCast: Bridging Ocean Wave Forecasts to Reanalysis via Flow Matching with Exogenous VariablesarXiv physics.ao-phenPaperPreprint
- New forecasting research could help millions living in earthquake-prone regionsUniversity of BristolenTo be verified