AI for Ocean Currents & Tsunami Early-Warning

AI for Ocean Currents & Tsunami Early-Warning is a 3-day, hands-on workshop that blends coastal physics with modern ML to turn ocean data into actionable alerts. You’ll clean and grid buoy/HF-radar/altimetry data, run data assimilation, train compact models (PINNs, FNOs, encoder–decoders) for short-term forecasts, and quantify uncertainty with ensembles and conformal prediction. By the end, you’ll detect tsunami-like anomalies, estimate ETAs with confidence bands, and wire a mini operational pipeline—from ingest and QC to calibrated alerts and a one-page situation report.

Equip participants to build an end-to-end, physics-aware AI workflow that converts ocean observations into reliable, uncertainty-calibrated forecasts and tsunami early-warning alerts—covering data ingest/QC, data assimilation, ML forecasting (PINNs/FNOs/transformers), and operational alert design.

• Grasp core coastal/tsunami physics (shallow-water, bathymetry).

• Ingest, QC, align, and grid buoy/HF-radar/altimetry data.

• Benchmark baselines vs ML (PINNs, FNOs, encoder–decoders/SSMs).

• Apply lightweight data assimilation and read analysis increments.

• Produce short-horizon forecasts with calibrated uncertainty.

• Detect tsunami-like anomalies and estimate ETA with bands.

• Evaluate skill (CRPS, PR/ROC, cost–loss) and set alert thresholds.

• Wire an ops pipeline and generate a one-page situation report.

• Coastal & ocean modelers, met-ocean analysts, and early-warning/DRM teams

• Data scientists/ML engineers in public agencies, NGOs, or blue-economy startups

• Graduate/advanced undergraduate students in Ocean/Climate/EE/CS/ME

• Researchers working with buoy, HF-radar, satellite altimetry, or tide-gauge data

• QC, align, and grid buoy/HF-radar/altimetry data (NetCDF/Zarr).

• Train baselines + physics-aware ML (PINNs, FNOs, encoder–decoders).

• Assimilate observations (EnKF intuition) and visualize increments.

• Deliver +6 h surface-current & +24 h buoy forecasts with calibrated uncertainty.

• Detect tsunami-like anomalies and estimate ETA with confidence bands.

• Evaluate skill (CRPS, PR/ROC, cost–loss) and set tiered alert thresholds.

• Ship a minimal ops pipeline (ingest→QC→DA→forecast→UQ→publish) and a one-page situation report.