Jump to content

SACRUS

Members
 View Profile  See their activity

  • Posts

    14,635

  • Joined

  • Last visited

 Content Type 

Posts posted by SACRUS

    Extreme Cold, Snow & Sleet: SECS 1/25 - 1/26

    in New York City Metro

    Posted

    3 minutes ago, EW9616 said:

    I’ve seen some TV Mets referring and showing the Graf model that shows little mixing. Anyone know what model that is? Is correlated with one of the main US model?


    .

    GRAF refers to an AI-based global weather forecasting model developed by NOAA (GSL) and collaborators, and the name stands for:

    GRAF = Global Regression AI Forecast model

    It’s part of the new generation of machine-learning weather models, similar in spirit to GraphCast or Pangu, but developed inside the NOAA ecosystem.

    What is the GRAF model?

    GRAF is a pure machine-learning global forecast model that:

    • Learns atmospheric evolution from historical reanalysis data

    • Produces global forecasts without explicitly solving physical equations

    • Uses regression-based deep learning to predict future atmospheric states

    Think of it as:

    “AI learning how the atmosphere usually evolves, then extrapolating forward.”

    Key characteristics

    • Global coverage

    • AI / ML-based (no traditional physics core)

    • Predicts large-scale fields (e.g., 500 mb heights, winds, temps, MSLP)

    • Extremely fast compared to physics models

    • Designed primarily for pattern and flow prediction

    What GRAF is good at

    • Large-scale synoptic pattern recognition

    • Jet stream placement

    • Ridge / trough evolution

    • Medium-range guidance (Days 3–7)

    This makes it useful for:

    • Pattern forecasting

    • Ensemble support

    • Early signal detection

    What GRAF is NOT good at

    • Precipitation amounts

    • Snowfall totals

    • Precipitation type

    • Mesoscale features (banding, fronts)

    • Boundary-layer processes

    • Like 1
×
×
  • Create New...