The storm surge is now being asked to file its own forecast
the storm surge is now being asked to file its own forecast: a peak aware graph transformer predicts station level surge from atmospheric forcing with a strategy built around rare extremes, a spatio temporal graph network cuts forecast error by over 70 percent at 48 hours compared to an ocean model baseline, and a transformer hybrid fuses 583 inputs of forcing, model surge, and tide gauge readings to nowcast the southern north sea.
the coast now files its own water level.
Context
Storm surge forecasting is a trade between accuracy and cost. High-fidelity hydrodynamic models such as ADCIRC are accurate but expensive, which limits how many scenarios you can run. All three results use machine learning to get near that quality much faster, and each attacks a different weakness.
PACT is built around rare extremes, because the peak is what floods a town and what ordinary regression underfits. StormNet is a graph network that cuts forecast error by over 70 percent at 48 hours against a baseline model, per its summary. The third paper nowcasts 2 to 12 hours ahead at 22 tide gauges in the southern North Sea, combining gauge observations, ERA5 wind and pressure, and output from the operational COHERENS model.
The first two are emulators and correctors of numerical models; the third is a hybrid that keeps the physical model in the loop as an input. For operational use the hybrid design is easier to trust, because a bad input is visible and the physics still anchors the forecast.
Related work
- StormNet code (NoujoudNader/SormNet) ↗Public repository for the graph network.
Watch next
- Skill on the rare, highest surges rather than on average error.
- Whether these models hold up on storms unlike the ones in their training sets.
Sources
- PACT: Peak-Aware Cross-Attention Graph Transformers for Efficient Storm-Surge Emulation (arXiv)arxiv.org
- StormNet: Improving storm surge predictions with a GNN (arXiv)arxiv.org
- A hybrid machine-learning framework combining hydrodynamic simulation and tide-gauge data for nowcasting (EGUsphere preprint)egusphere.copernicus.org
Provenance
The note above is reproduced unedited from the original post, first published on Threads on 4 October 2026 at 14:48 IST. Sources are the papers and datasets the note draws on.
View the original post ↗Embed this note
More notes
The air is now being asked to keep its own ledger
the air is now being asked to keep its own ledger: ecmwf’s aifs compo becomes the first ai model to forecast atmospheric composition globally every three hours, cleanair simulates 365 days of pm2.5 over china in ten seconds, and a unified framework maps six pollutants at one kilometer across the whole country. the air now files its own composition report.
read the note →The current is now being asked to draw its own map
the current is now being asked to draw its own map: china’s langya 2.0 predicts six ocean phenomena including internal waves and mesoscale eddies, a deep net called wenhai resolves eddies globally with air sea flux formulas built in, and scripps infers surface currents from the way temperature patterns deform in satellite images. the ocean now files its own circulation report.
read the note →The soil is now being asked to report its own carbon
the soil is now being asked to report its own carbon: a nix color sensor paired with generative data augmentation predicts soil organic carbon without a lab, random forest drives 74 percent of soil health mapping studies, and sentinel 2 tracks five year carbon change across france and italy from 922 samples. the dirt now files its own carbon account.
read the note →