The Spread
564570570570570576576576576576582582582582582582588588588588588588588 9.9

Widest gap 9.9 dam at 24.0N 122.5W · unshaded ground is where all 4 land within 2 dam of each other

Thin lines are the 500 mb height itself, every 6 dam — the per-cell median of the 4, so a pattern no single model forecast. Read the troughs and ridges off them; the shading is where that pattern is in doubt.

Shows where the models disagree about the 500 mb pattern at 18Z 6 Sep — the day you choose, always at the same hour so that successive days compare.

Methodology

Across-model 500 hPa geopotential height spread from GFS, ECMWF, ECMWF AIFS, GEM — four models and three centres, named because a roster is a fact about a chart: two of these are ECMWF's, and where those two agree they agree partly for family reasons. Raw model output on every side.

Heights are in decametres (tens of metres), the unit every 500 mb chart uses: 582 is 5,820 m. The bands were cut against what this field actually does — over twenty valid times in early September the models sat within 2 dam of each other on two thirds of the frame, and past 8 dam on about one cell in eighty — so the scale is the disagreement's own, not a round guess.

The pattern at 500 mb steers the weather under it, which is why a forecaster looks here first: two models that agree on tomorrow's temperature and disagree by 10 dam on where the trough sits are not agreeing about the day after. Spread grows with lead — a day-1 map is mostly unshaded, and that is the picture.

The widest gap is read from the unsmoothed grid, because smoothing fills extremes, and from the drawn frame only. Valid at 18Z and the same hour every day so successive maps compare — the temperature face's hour, kept so a reader switching between them compares one moment.

Shaded ground is a disagreement rather than an error — none of these models is the truth here, and the map cannot say which is right. For who has actually been closest, the model leaderboards are the place.

Every map on this page is drawn from model output we store ourselves, not from anyone's picture of it. The per-city read of the same day — which cities the models are fighting about, and what the discussion makes of it — is on the national page.