Ensemble Meteogram
One forecast is a guess. Thirty-one forecasts, each started from slightly different conditions, show how sure the models are. Pick any spot in Pennsylvania and see the full spread for temperature, rain, snow, wind and clouds, plus the odds for each day.
Temperature °F · 2 m
Precipitation inches · accumulated from now
Snow inches · 10:1 ratio · accumulated
Wind & clouds sustained 10 m wind, mph · total cloud cover
Pattern sea-level pressure (hPa) and 850 mb temperature (°C)
Day by day local days · share of the 31 members
Highs and lows come from 6-hourly model output (2 AM, 8 AM, 2 PM, 8 PM EDT), so true extremes run a degree or two beyond what is shown. Snow uses the model's precipitation-type flag and a 10:1 ratio.
How to read this
The bands are the forecast. The dark band holds the middle half of the members, the light band holds 80 percent of them. A narrow band means the members agree; a band that fans out a week ahead means the pattern is genuinely uncertain, and any single-number forecast for that day is a coin flip dressed up.
Probabilities are member counts. "40% chance of 0.25 inch" means 12 or 13 of 31 members produce at least that much at this grid point that day. The GEFS is somewhat under-dispersive, so real-world odds for extremes are a little wider than it shows.
Grid-point caveat. Each point is a 0.5° box, roughly 35 by 25 miles. Valleys, ridges and lake-effect bands are smoothed out. Use it for timing and odds, and let the NWS point forecast handle local detail.
Compare with previous run overlays the prior cycle's median. When successive runs jump around, confidence should stay low even if the current run looks tight.
This page is the CPA Weather Lab take on the prob_meteogram idea: a pipeline on the lab server pulls the 31 GEFS members from NOAA NOMADS for a Pennsylvania box four times a day (Herbie fetches from the NOAA archive on AWS as a fallback) and publishes one small file per grid point.