The surge is now being asked to announce its own height
the surge is now being asked to announce its own height: a diffusion framework called surgegen synthesizes realistic storm surge scenarios from hypothetical storms in continuous parameter space, a graph transformer reads atmospheric forcing patches to hit station peaks, and a hybrid explainable system forecasts the bohai sea at medium to long range.
the coast now files its own tide forecast.
Context
SurgeGen (arXiv, September 2026) is a two-stage framework. A baseline model gives a coarse storm surge height estimate, which then conditions a diffusion model that generates refined surge scenarios for hypothetical storms with parameters in a continuous space. The authors say it produces realistic and diverse scenarios both within and outside the training distribution.
PACT (arXiv, May 2026) is a peak-aware cross-attention graph transformer. It treats each atmospheric forcing patch as a graph, encodes it with GraphSAGE, and uses a learned station query to predict station-level surge. It beats a spatio-temporal graph neural network baseline on RMSE and MAE across US Northeast tide gauges, needs about 3.5 seconds for a winter surge trajectory for one year, and degrades markedly when moving from reanalysis to climate-model forcings.
The Bohai Sea system (Yantai Institute of Coastal Zone Research, CAS, July 2026) uses three tide gauge stations (Tanggu, Huanghua, Weifang), 15 typhoons from 2000 to 2022 and ADCIRC simulations for training data. A CNN-BiLSTM-Attention model handles 24 hour multi-step forecasts and a standalone BiLSTM handles 24 to 72 hour single-step forecasts, with interpretability analysis of drivers.
The note's 'graph transformer reads atmospheric forcing patches to hit station peaks' matches PACT, but PACT claims better RMSE and MAE and better peak fidelity, not a hit rate on peaks, and it loses accuracy on climate-model forcings. SurgeGen produces scenario ensembles for hypothetical storms and is not a forecast of a particular storm. The Bohai system is described by the institute as hybrid and interpretable, but its page gives no error figures in the part retrieved, and its training data are ADCIRC simulations of 15 typhoons, a small set. The 'medium to long range' in the note is 24 to 72 hours in the source.
Related work
- SurgeGen: A Hybrid Generative Diffusion Framework for Storm Surge Scenario Synthesis (arXiv) ↗Source for the two-stage diffusion scenario generator.
- PACT: Peak-Aware Cross-Attention Graph Transformers for Efficient Storm-Surge Emulation (arXiv) ↗Source for the graph transformer and station peak results.
- Yantai Institute of Coastal Zone Research Achieves New Breakthrough in Medium- and Long-Term Intelligent Forecasting and Interpretability of Typhoon Storm Surge (CAS) ↗Source for the Bohai Sea hybrid forecast and its lead times.
Watch next
- Find the published error numbers for the Bohai system and check how it does on storms outside the 15 typhoon set. Test SurgeGen scenarios against real surge records.
Sources
- SurgeGen: A Hybrid Generative Diffusion Framework for Storm Surge Scenario Synthesis (arXiv)arxiv.org
- PACT: Peak-Aware Cross-Attention Graph Transformers for Efficient Storm-Surge Emulation (arXiv)arxiv.org
- Yantai Institute of Coastal Zone Research Achieves New Breakthrough in Medium- and Long-Term Intelligent Forecasting and Interpretability of Typhoon Storm Surge (CAS)english.yic.cas.cn
Provenance
The note above is reproduced unedited from the original post, first published on Threads on 4 October 2026 at 01:47 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 →