NOAA AIGFS Surface Forecasts

Description

AIGFS Surface is the surface-field product from NOAA/NCEP's operational Artificial Intelligence Global Forecast System. NOAA implemented AIGFS v1.0 on December 17, 2025, alongside AIGEFS and HGEFS, as part of the operational AI global model family described in NWS Service Change Notice 25-89.

The model is based on Google DeepMind's GraphCast architecture and was developed by NCEP with NOAA research laboratories and EPIC. NOAA's technical background describes the GraphCast lineage as a learned global forecast model using current and previous six-hour atmospheric states, with GDAS data providing real-time initial conditions.

GribStream mirrors this NOMADS-only GRIB2 feed into private S3 storage and indexes the GRIB offsets so the surface variables can be queried through the same /timeseries and /runs endpoints as GFS, GEFS, and GraphCast GFS.

What is in the surface product

The surface slice is intentionally compact: 10 m U/V wind components, 2 m temperature, mean sea-level pressure, and accumulated precipitation. It is useful for global point forecasts, first-pass operational dashboards, model comparison, and AI-vs-physics checks where you want the most common near-surface fields without pulling the pressure-level product.

Detail

• Archive window: Last 5 days
• Product inventory: NCEP AIGFS products
• Operational source: NOAA NOMADS AIGFS
• Operational notice: NWS SCN 25-89
• Technical background: NOAA/NCEP Office Note 521
• Distribution: NOMADS GRIB2 only; GribStream mirrors the files before serving API requests.
• Resolution: 0.25° global grid
• Cycles: 00, 06, 12, 18 UTC
• Forecast cadence and length: 6-hour steps to 384 hours
• Companion dataset: AIGFS Pressure

Parameters