CastCheck v0.3.1 · data through 2026-10-28

Daily verification · methodology v0.3.1

How far off was each weather model?

Over the last 365 days, across 23 U.S. stations, the most accurate raw sampled daily minimum forecast 1 day ahead is ECMWF IFS HRES: mean absolute error 2.12 °F [1.98, 2.30], bias −0.59 °F, on n = 87 scored days.

Scope: raw model output on the native 0.25° grid. This view scores the sampled daily minimum against the minimum of the four *observed* samples on the same day. Errors are computed in °C and shown in °F (methodology v0.3.1).

Window
the last 365 days
Variable
sampled daily minimum
Init
12Z
Interpolation
bilinear
Data through
2026-10-28
Updated
Next update

Leader, lead day 1

2.12°F

[1.98, 2.30] MAE

ECMWF IFS HRES

Persistence baseline

4.02°F

[3.60, 4.47]

the observation lead_day days earlier — the bar every model has to clear at its own lead

Scored days

87

23 stations · 12 systems

in the last 365 days, lead day 1

Systems ranked

11/12

next update

below n = 30: published, greyed, unranked

Mean absolute error with 95 % confidence intervals

Lead day 1 · the last 365 days · 23 stations pooled · 12Z · bilinear · sampled daily minimum. Whiskers are moving-block bootstrap intervals; models whose whiskers overlap are not distinguishable at this sample size. Baselines are drawn in grey and never ranked.

MAE by model with 95 % confidence intervals, lead day 1, 365d, 12Z, bilinear012345mean absolute error, °FECMWF IFS HRESECMWF IFS HRES: 2.12 °F2.12ECMWF AIFS SingleECMWF AIFS Single: 2.26 °F2.26Aurora (IFS init)Aurora (IFS init): 2.31 °F2.31Aurora (GFS init)Aurora (GFS init): 2.38 °F2.38GraphCast (IFS init)GraphCast (IFS init): 2.4 °F2.4GraphCast (GFS init)GraphCast (GFS init): 2.4 °F2.4Pangu-Weather (GFS init)Pangu-Weather (GFS init): 2.44 °F2.44Pangu-Weather (IFS init)Pangu-Weather (IFS init): 2.49 °F2.49NCEP GFSNCEP GFS: 2.55 °F2.55FourCastNet v2 (IFS init)FourCastNet v2 (IFS init): 2.76 °F2.76FourCastNet v2 (GFS init)FourCastNet v2 (GFS init): 2.84 °F2.84Persistence (baseline)Persistence (baseline): 4.02 °F4.02

Source: CastCheck 0.3.3, methodology v0.3.1 · ECMWF open data, NOAA/NCEP GFS, NOAA/CIRA AIWP, NWS ASOS · data through 2026-10-28 · CC BY 4.0. Hover a bar for its value.

Lead day 1target = init + 1 d

permalink

The last 365 days, all 23 stations pooled, 12Z initialization, bilinear interpolation, sampled daily minimum against the minimum of the four *observed* samples on the same day. Lower MAE is better; bias is positive when the model is too warm. Every number links to its permanent page.

Skill is 1 − MAE ÷ MAE(persistence) computed on the days both have a value — the small print in that column is that denominator and the size of the intersection, so the number can be checked against the persistence row rather than contradicting it. The persistence row's own n is its whole record, marked all days. Skill, debiased (out-of-sample) repeats the score after removing a bias estimated on the 30 scored days before each day and applied forward, never on the day itself. An interval reads — when the bootstrap could not run: fewer than 28 scored days or fewer than 4 blocks. Hit-rate intervals are Wilson score intervals, not bootstraps.

Lead day 1, the last 365 days, 12Z, bilinear, sampled daily minimum.
RankModelMAE °FBias °F±3 °F SkillSkill, debiased (out-of-sample) nvs leader
1ECMWF IFS HRESifs_hres2.12[1.98, 2.30]−0.59[−0.72, −0.47]76% +0.27[+0.16, +0.37]vs 2.92 (n=87)+0.47n=72 8722.9 stns★ lowest MAE among ranked models (n >= 30) in this view
2ECMWF AIFS Singleaifs_single2.26[2.12, 2.41]−1.37[−1.53, −1.23]73% +0.23[+0.11, +0.34]vs 2.92 (n=87)+0.48n=72 8722.9 stns= not distinguishable from the leader after the Holm correction within this table
3Aurora (IFS init)aurora_ifs2.31[2.23, 2.39]−1.16[−1.30, −1.01]72% +0.46[+0.38, +0.51]vs 4.25 (n=241)+0.59n=226 24122.9 stns= not distinguishable from the leader after the Holm correction within this table
4Aurora (GFS init)aurora_gfs2.38[2.29, 2.46]−0.93[−1.16, −0.70]72% +0.43[+0.36, +0.49]vs 4.18 (n=250)+0.57n=235 25022.8 stns= not distinguishable from the leader after the Holm correction within this table
5GraphCast (IFS init)graphcast_ifs2.40[2.31, 2.48]−1.29[−1.38, −1.20]69% +0.42[+0.34, +0.48]vs 4.10 (n=256)+0.57n=241 25622.9 stns= not distinguishable from the leader after the Holm correction within this table
6GraphCast (GFS init)graphcast_gfs2.40[2.29, 2.51]−0.82[−1.00, −0.64]71% +0.43[+0.36, +0.48]vs 4.18 (n=250)+0.54n=235 25022.8 stns= not distinguishable from the leader after the Holm correction within this table
7Pangu-Weather (GFS init)pangu_gfs2.44[2.32, 2.57]−0.72[−0.90, −0.53]70% +0.42[+0.36, +0.47]vs 4.18 (n=250)+0.50n=235 25022.8 stns= not distinguishable from the leader after the Holm correction within this table
8Pangu-Weather (IFS init)pangu_ifs2.49[2.40, 2.58]−1.27[−1.37, −1.17]69% +0.41[+0.34, +0.47]vs 4.25 (n=241)+0.54n=226 24122.9 stns▼ worse than the leader; distinguishable after the Holm correction within this table
9NCEP GFSgfs2.55[2.34, 2.82]+0.78[0.36, 1.24]66% +0.12[−0.00, +0.24]vs 2.92 (n=87)+0.34n=72 8722.9 stns▼ worse than the leader; distinguishable after the Holm correction within this table
10FourCastNet v2 (IFS init)fourcastnet_ifs2.76[2.65, 2.88]−1.13[−1.29, −0.99]63% +0.35[+0.28, +0.41]vs 4.25 (n=241)+0.47n=226 24122.9 stns▼ worse than the leader; distinguishable after the Holm correction within this table
11FourCastNet v2 (GFS init)fourcastnet_gfs2.84[2.71, 2.95]−1.20[−1.37, −1.03]62% +0.32[+0.25, +0.38]vs 4.18 (n=250)+0.45n=235 25022.8 stns▼ worse than the leader; distinguishable after the Holm correction within this table
—Persistence (baseline)persistence · baseline4.02[3.60, 4.47]−0.10[−0.34, 0.12]53% —— 300all days

+ model too warm − model too cold bias interval includes zero ★ lowest MAE among ranked models (n ≥ 30) = not distinguishable from the leader ▼ worse than the leader ▲ better than the leader — all three Holm-corrected within this table, over the family of comparisons against the leader; the uncorrected verdict for every pair is published in the pairwise table of each permanent link and in pairwise_latest.csv.gzn < 30 greyed and unranked CI is a 95 % moving-block bootstrap interval on the group's own days, and reads — when it could not be computed skill reads — when fewer than 10 days are common to the model and the baseline: the ratio of two means over a handful of shared days is not a number worth printing

Lead day 3target = init + 3 d

permalink

Same sampling as lead day 1. Skill is 1 − MAE ÷ MAE(persistence) on the days both have a value.

Lead day 3, the last 365 days, 12Z, bilinear, sampled daily minimum.
RankModelMAE °FBias °FSkillnvs leader
1ECMWF AIFS Singleaifs_single2.35[2.23, 2.49]−1.28[−1.44, −1.12]+0.46[+0.34, +0.56]vs 4.37 (n=85)8522.9 stns★ lowest MAE among ranked models (n >= 30) in this view
2ECMWF IFS HRESifs_hres2.41[2.28, 2.58]−0.50[−0.66, −0.34]+0.45[+0.34, +0.54]vs 4.37 (n=85)8522.9 stns= not distinguishable from the leader after the Holm correction within this table
3Aurora (IFS init)aurora_ifs2.56[2.45, 2.66]−1.17[−1.38, −0.94]+0.60[+0.54, +0.64]vs 6.34 (n=239)23922.9 stns= not distinguishable from the leader after the Holm correction within this table
4GraphCast (IFS init)graphcast_ifs2.57[2.47, 2.68]−1.27[−1.39, −1.13]+0.58[+0.53, +0.62]vs 6.17 (n=254)25422.9 stns= not distinguishable from the leader after the Holm correction within this table
5Pangu-Weather (IFS init)pangu_ifs2.74[2.63, 2.86]−0.96[−1.11, −0.80]+0.57[+0.51, +0.61]vs 6.34 (n=239)23922.9 stns= not distinguishable from the leader after the Holm correction within this table
6GraphCast (GFS init)graphcast_gfs2.81[2.68, 2.93]−1.35[−1.54, −1.13]+0.56[+0.50, +0.60]vs 6.38 (n=249)24922.8 stns= not distinguishable from the leader after the Holm correction within this table
7NCEP GFSgfs2.82[2.61, 3.07]+0.82[0.35, 1.36]+0.36[+0.22, +0.47]vs 4.37 (n=85)8522.9 stns▼ worse than the leader; distinguishable after the Holm correction within this table
8Aurora (GFS init)aurora_gfs2.84[2.73, 2.95]−1.45[−1.72, −1.18]+0.56[+0.49, +0.60]vs 6.38 (n=249)24922.8 stns= not distinguishable from the leader after the Holm correction within this table
9Pangu-Weather (GFS init)pangu_gfs2.84[2.67, 3.00]−0.94[−1.18, −0.71]+0.56[+0.51, +0.59]vs 6.38 (n=249)24922.8 stns= not distinguishable from the leader after the Holm correction within this table
10FourCastNet v2 (IFS init)fourcastnet_ifs3.08[2.92, 3.26]−1.24[−1.46, −1.03]+0.51[+0.46, +0.56]vs 6.34 (n=239)23922.9 stns▼ worse than the leader; distinguishable after the Holm correction within this table
11FourCastNet v2 (GFS init)fourcastnet_gfs3.14[2.95, 3.32]−1.43[−1.65, −1.20]+0.51[+0.45, +0.55]vs 6.38 (n=249)24922.8 stns▼ worse than the leader; distinguishable after the Holm correction within this table
—Persistence (baseline)persistence · baseline6.14[5.43, 6.87]−0.20[−0.80, 0.35]—300all days

Lead day 5target = init + 5 d

permalink

Same sampling as lead day 1. Skill is 1 − MAE ÷ MAE(persistence) on the days both have a value.

Lead day 5, the last 365 days, 12Z, bilinear, sampled daily minimum.
RankModelMAE °FBias °FSkillnvs leader
1ECMWF AIFS Singleaifs_single2.64[2.49, 2.80]−1.21[−1.45, −0.98]+0.47[+0.35, +0.57]vs 5.02 (n=83)8322.9 stns★ lowest MAE among ranked models (n >= 30) in this view
2ECMWF IFS HRESifs_hres2.78[2.60, 2.99]−0.36[−0.57, −0.14]+0.45[+0.33, +0.54]vs 5.02 (n=83)8322.9 stns= not distinguishable from the leader after the Holm correction within this table
3GraphCast (IFS init)graphcast_ifs2.92[2.75, 3.10]−1.12[−1.32, −0.93]+0.57[+0.52, +0.61]vs 6.77 (n=252)25222.9 stns= not distinguishable from the leader after the Holm correction within this table
4Aurora (IFS init)aurora_ifs3.02[2.85, 3.18]−0.97[−1.27, −0.65]+0.57[+0.51, +0.61]vs 6.95 (n=237)23722.9 stns= not distinguishable from the leader after the Holm correction within this table
5Pangu-Weather (IFS init)pangu_ifs3.17[2.98, 3.40]−0.60[−0.89, −0.34]+0.54[+0.49, +0.58]vs 6.95 (n=237)23722.9 stns= not distinguishable from the leader after the Holm correction within this table
6NCEP GFSgfs3.22[2.95, 3.54]+0.79[0.29, 1.37]+0.36[+0.24, +0.45]vs 5.02 (n=83)8322.9 stns▼ worse than the leader; distinguishable after the Holm correction within this table
7GraphCast (GFS init)graphcast_gfs3.36[3.17, 3.58]−1.49[−1.83, −1.18]+0.52[+0.47, +0.56]vs 7.01 (n=247)24722.8 stns= not distinguishable from the leader after the Holm correction within this table
8Pangu-Weather (GFS init)pangu_gfs3.39[3.17, 3.66]−1.03[−1.44, −0.70]+0.52[+0.47, +0.56]vs 7.01 (n=247)24722.8 stns= not distinguishable from the leader after the Holm correction within this table
9Aurora (GFS init)aurora_gfs3.52[3.34, 3.73]−1.92[−2.31, −1.58]+0.50[+0.43, +0.55]vs 7.01 (n=247)24722.8 stns▼ worse than the leader; distinguishable after the Holm correction within this table
10FourCastNet v2 (IFS init)fourcastnet_ifs3.62[3.34, 3.91]−1.45[−1.76, −1.16]+0.48[+0.43, +0.53]vs 6.95 (n=237)23722.9 stns= not distinguishable from the leader after the Holm correction within this table
11FourCastNet v2 (GFS init)fourcastnet_gfs3.70[3.44, 3.99]−1.70[−2.02, −1.43]+0.47[+0.42, +0.51]vs 7.01 (n=247)24722.8 stns▼ worse than the leader; distinguishable after the Holm correction within this table
—Persistence (baseline)persistence · baseline6.76[6.00, 7.56]−0.25[−1.14, 0.60]—300all days

Lead day 7target = init + 7 d

permalink

Same sampling as lead day 1. Skill is 1 − MAE ÷ MAE(persistence) on the days both have a value.

Lead day 7, the last 365 days, 12Z, bilinear, sampled daily minimum.
RankModelMAE °FBias °FSkillnvs leader
1ECMWF AIFS Singleaifs_single3.20[2.88, 3.53]−1.13[−1.40, −0.82]+0.40[+0.30, +0.48]vs 5.31 (n=81)8122.9 stns★ lowest MAE among ranked models (n >= 30) in this view
2ECMWF IFS HRESifs_hres3.51[3.07, 4.10]+0.09[−0.30, 0.46]+0.34[+0.24, +0.43]vs 5.31 (n=81)8122.9 stns= not distinguishable from the leader after the Holm correction within this table
3GraphCast (IFS init)graphcast_ifs3.76[3.42, 4.12]−1.07[−1.40, −0.74]+0.48[+0.44, +0.52]vs 7.29 (n=250)25022.8 stns= not distinguishable from the leader after the Holm correction within this table
4Aurora (IFS init)aurora_ifs3.86[3.54, 4.19]−0.35[−0.85, 0.19]+0.49[+0.44, +0.53]vs 7.53 (n=235)23522.8 stns= not distinguishable from the leader after the Holm correction within this table
5NCEP GFSgfs3.90[3.49, 4.37]+0.76[0.18, 1.43]+0.27[+0.14, +0.36]vs 5.31 (n=81)8122.9 stns▼ worse than the leader; distinguishable after the Holm correction within this table
6Pangu-Weather (IFS init)pangu_ifs4.07[3.70, 4.41]−0.36[−0.85, 0.07]+0.46[+0.41, +0.50]vs 7.53 (n=235)23522.8 stns▼ worse than the leader; distinguishable after the Holm correction within this table
7Pangu-Weather (GFS init)pangu_gfs4.16[3.82, 4.49]−0.77[−1.21, −0.32]+0.44[+0.39, +0.49]vs 7.49 (n=245)24522.8 stns▼ worse than the leader; distinguishable after the Holm correction within this table
8GraphCast (GFS init)graphcast_gfs4.19[3.84, 4.59]−1.33[−1.73, −0.84]+0.44[+0.39, +0.48]vs 7.49 (n=245)24522.8 stns▼ worse than the leader; distinguishable after the Holm correction within this table
9Aurora (GFS init)aurora_gfs4.34[4.06, 4.65]−2.01[−2.43, −1.53]+0.42[+0.36, +0.47]vs 7.49 (n=245)24522.8 stns▼ worse than the leader; distinguishable after the Holm correction within this table
10FourCastNet v2 (IFS init)fourcastnet_ifs4.36[3.92, 4.79]−1.49[−1.98, −1.07]+0.42[+0.37, +0.46]vs 7.53 (n=235)23522.8 stns▼ worse than the leader; distinguishable after the Holm correction within this table
11FourCastNet v2 (GFS init)fourcastnet_gfs4.39[4.01, 4.78]−1.69[−2.08, −1.31]+0.41[+0.36, +0.46]vs 7.49 (n=245)24522.8 stns▼ worse than the leader; distinguishable after the Holm correction within this table
—Persistence (baseline)persistence · baseline7.18[6.38, 8.00]−0.29[−1.41, 0.76]—300all days

The same four samples as a daily maximum and minimum

Lead day 1, the last 365 days, 12Z, bilinear. Here the forecast's max/min of the four samples is scored against the observation's max/min of the same four instants. Like for like: whatever the four-sample definition misses, it misses on both sides. The comparison against the true NWS daily extremes is a different question and lives on the station and permanent-link pages.

Sampled daily maximumtmax_s

RankModelMAE °FBias °FSkillnvs leader
1ECMWF AIFS Singleaifs_single1.94[1.84, 2.02]−0.45[−0.54, −0.35]+0.46vs 3.60 (n=87)87★
2ECMWF IFS HRESifs_hres2.33[2.20, 2.46]−0.95[−1.21, −0.70]+0.35vs 3.60 (n=87)87▼
3Pangu-Weather (IFS init)pangu_ifs2.48[2.36, 2.60]−0.75[−1.03, −0.50]+0.50vs 4.91 (n=241)241▼
4NCEP GFSgfs2.69[2.44, 2.95]+0.65[0.11, 1.18]+0.25vs 3.60 (n=87)87▼
5Pangu-Weather (GFS init)pangu_gfs2.80[2.67, 2.94]−1.35[−1.62, −1.05]+0.43vs 4.89 (n=250)250▼
6GraphCast (IFS init)graphcast_ifs2.81[2.73, 2.90]−2.18[−2.30, −2.06]+0.41vs 4.80 (n=256)256▼
7Aurora (IFS init)aurora_ifs2.87[2.76, 2.99]−2.31[−2.46, −2.16]+0.42vs 4.91 (n=241)241▼
8FourCastNet v2 (IFS init)fourcastnet_ifs3.00[2.84, 3.17]−1.81[−2.06, −1.60]+0.39vs 4.91 (n=241)241▼
9FourCastNet v2 (GFS init)fourcastnet_gfs3.09[2.94, 3.25]−1.95[−2.18, −1.70]+0.37vs 4.89 (n=250)250▼
10GraphCast (GFS init)graphcast_gfs3.33[3.17, 3.49]−2.79[−2.99, −2.59]+0.32vs 4.89 (n=250)250▼
11Aurora (GFS init)aurora_gfs3.47[3.26, 3.67]−3.06[−3.31, −2.79]+0.29vs 4.89 (n=250)250▼
—Persistence (baseline)persistence4.74[4.29, 5.22]−0.08[−0.29, 0.13]—300all days

Sampled daily minimumtmin_s

RankModelMAE °FBias °FSkillnvs leader
1ECMWF IFS HRESifs_hres2.12[1.98, 2.30]−0.59[−0.72, −0.47]+0.27vs 2.92 (n=87)87★
2ECMWF AIFS Singleaifs_single2.26[2.12, 2.41]−1.37[−1.53, −1.23]+0.23vs 2.92 (n=87)87=
3Aurora (IFS init)aurora_ifs2.31[2.23, 2.39]−1.16[−1.30, −1.01]+0.46vs 4.25 (n=241)241=
4Aurora (GFS init)aurora_gfs2.38[2.29, 2.46]−0.93[−1.16, −0.70]+0.43vs 4.18 (n=250)250=
5GraphCast (IFS init)graphcast_ifs2.40[2.31, 2.48]−1.29[−1.38, −1.20]+0.42vs 4.10 (n=256)256=
6GraphCast (GFS init)graphcast_gfs2.40[2.29, 2.51]−0.82[−1.00, −0.64]+0.43vs 4.18 (n=250)250=
7Pangu-Weather (GFS init)pangu_gfs2.44[2.32, 2.57]−0.72[−0.90, −0.53]+0.42vs 4.18 (n=250)250=
8Pangu-Weather (IFS init)pangu_ifs2.49[2.40, 2.58]−1.27[−1.37, −1.17]+0.41vs 4.25 (n=241)241▼
9NCEP GFSgfs2.55[2.34, 2.82]+0.78[0.36, 1.24]+0.12vs 2.92 (n=87)87▼
10FourCastNet v2 (IFS init)fourcastnet_ifs2.76[2.65, 2.88]−1.13[−1.29, −0.99]+0.35vs 4.25 (n=241)241▼
11FourCastNet v2 (GFS init)fourcastnet_gfs2.84[2.71, 2.95]−1.20[−1.37, −1.03]+0.32vs 4.18 (n=250)250▼
—Persistence (baseline)persistence4.02[3.60, 4.47]−0.10[−0.34, 0.12]—300all days

Every model × every lead day

MAE in °F with the bias and n underneath, the last 365 days, 12Z, bilinear, sampled daily minimum. The sparkline is the same model's MAE across lead days 1–9 on a shared vertical scale.

Modeld1d2d3d4d5d6d7d8d9 lead 1–9
ECMWF AIFS Singleaifs_single2.26−1.37 · n=872.28−1.33 · n=862.35−1.28 · n=852.47−1.26 · n=842.64−1.21 · n=832.89−1.16 · n=823.20−1.13 · n=813.42−0.99 · n=803.97−1.04 · n=79ECMWF AIFS Single MAE by lead day 1 to 9
Aurora (GFS init)aurora_gfs2.38−0.93 · n=2502.61−1.21 · n=2502.84−1.45 · n=2493.13−1.67 · n=2483.52−1.92 · n=2473.97−2.02 · n=2464.34−2.01 · n=2454.84−1.99 · n=2445.24−1.92 · n=243Aurora (GFS init) MAE by lead day 1 to 9
Aurora (IFS init)aurora_ifs2.31−1.16 · n=2412.41−1.17 · n=2402.56−1.17 · n=2392.76−1.13 · n=2383.02−0.97 · n=2373.38−0.72 · n=2363.86−0.35 · n=2354.31+0.01 · n=2344.78+0.28 · n=233Aurora (IFS init) MAE by lead day 1 to 9
FourCastNet v2 (GFS init)fourcastnet_gfs2.84−1.20 · n=2502.99−1.31 · n=2503.14−1.43 · n=2493.40−1.60 · n=2483.70−1.70 · n=2474.11−1.66 · n=2464.39−1.69 · n=2454.95−1.72 · n=2445.40−1.90 · n=243FourCastNet v2 (GFS init) MAE by lead day 1 to 9
FourCastNet v2 (IFS init)fourcastnet_ifs2.76−1.13 · n=2412.95−1.18 · n=2403.08−1.24 · n=2393.31−1.38 · n=2383.62−1.45 · n=2373.97−1.42 · n=2364.36−1.49 · n=2354.86−1.44 · n=2345.35−1.47 · n=233FourCastNet v2 (IFS init) MAE by lead day 1 to 9
NCEP GFSgfs2.55+0.78 · n=872.69+0.77 · n=862.82+0.82 · n=852.93+0.86 · n=843.22+0.79 · n=833.57+0.77 · n=823.90+0.76 · n=814.42+0.65 · n=804.76+0.63 · n=79NCEP GFS MAE by lead day 1 to 9
GraphCast (GFS init)graphcast_gfs2.40−0.82 · n=2502.62−1.17 · n=2502.81−1.35 · n=2493.06−1.46 · n=2483.36−1.49 · n=2473.81−1.48 · n=2464.19−1.33 · n=2454.76−1.37 · n=2445.10−1.36 · n=243GraphCast (GFS init) MAE by lead day 1 to 9
GraphCast (IFS init)graphcast_ifs2.40−1.29 · n=2562.47−1.31 · n=2552.57−1.27 · n=2542.71−1.20 · n=2532.92−1.12 · n=2523.33−1.12 · n=2513.76−1.07 · n=2504.24−1.02 · n=2494.86−1.06 · n=248GraphCast (IFS init) MAE by lead day 1 to 9
ECMWF IFS HRESifs_hres2.12−0.59 · n=872.24−0.52 · n=862.41−0.50 · n=852.58−0.43 · n=842.78−0.36 · n=833.15−0.21 · n=823.51+0.09 · n=813.63+0.11 · n=804.00+0.01 · n=79ECMWF IFS HRES MAE by lead day 1 to 9
Pangu-Weather (GFS init)pangu_gfs2.44−0.72 · n=2502.64−0.90 · n=2502.84−0.94 · n=2493.05−1.00 · n=2483.39−1.03 · n=2473.78−0.85 · n=2464.16−0.77 · n=2454.75−0.61 · n=2445.30−0.57 · n=243Pangu-Weather (GFS init) MAE by lead day 1 to 9
Pangu-Weather (IFS init)pangu_ifs2.49−1.27 · n=2412.64−1.17 · n=2402.74−0.96 · n=2392.88−0.78 · n=2383.17−0.60 · n=2373.59−0.41 · n=2364.07−0.36 · n=2354.58−0.19 · n=2345.13−0.02 · n=233Pangu-Weather (IFS init) MAE by lead day 1 to 9
Persistence (baseline)persistence4.02−0.10 · n=3005.57−0.15 · n=3006.14−0.20 · n=3006.47−0.22 · n=3006.76−0.25 · n=3007.01−0.27 · n=3007.18−0.29 · n=3007.26−0.30 · n=3007.41−0.32 · n=300Persistence (baseline) MAE by lead day 1 to 9

Where the errors are

Both maps show a fixed quantity per station. Neither shows which model wins where: on samples this short the per-station winner is mostly noise, and a map of winners would invite a comparison the data cannot carry.

ECMWF IFS HRES bias

Mean bias of the fixed reference model ECMWF IFS HRES (lead day 1, sampled daily minimum, 365d window, 12Z, bilinear). One model everywhere, so the colours compare stations, not models. Dot area grows with the number of scored days; fill is the mean bias (warm = model too warm, cool = too cold, grey = interval includes zero). The frame is a latitude/longitude graticule on an Albers projection.

Mean bias in °F of ECMWF IFS HRES, lead day 1, sampled daily minimum, 365d window, 12Z, bilinear125°W115°W105°W95°W85°W75°W65°W25°N30°N35°N40°N45°N50°NKATL Atlanta Hartsfield: bias −1.58 °F, n = 87KATL−KAUS Austin Bergstrom: bias +2.15 °F, n = 87KAUS+KBOS Boston Logan: bias −1.49 °F, n = 83KBOS−KDCA Washington Reagan: bias −1.59 °F, n = 87KDCA−KDEN Denver Intl: bias +0.45 °F, n = 87KDEN+KDFW Dallas-Fort Worth: bias +0.85 °F, n = 87KDFW+KEWR Newark Liberty: bias −0.87 °F, n = 87KEWR−KIAH Houston Bush: bias +0.65 °F, n = 87KIAH+KLAS Las Vegas Harry Reid: bias −4.29 °F, n = 87KLAS−KLAX Los Angeles Intl: bias −0.67 °F, n = 87KLAX−KMIA Miami Intl: bias −2.31 °F, n = 87KMIA−KMSP Minneapolis-St Paul: bias −0.95 °F, n = 87KMSP−KMSY New Orleans Intl: bias +1.25 °F, n = 87KMSY+KNYC New York Central Park: bias −0.83 °F, n = 86KNYC−KOKC Oklahoma City: bias +2.95 °F, n = 87KOKC+KORD Chicago O'Hare: bias −0.93 °F, n = 87KORD−KPHL Philadelphia Intl: bias −2.51 °F, n = 87KPHL−KPHX Phoenix Sky Harbor: bias −2.40 °F, n = 87KPHX−KSAN San Diego Lindbergh: bias −0.54 °F, n = 87KSAN−KSAT San Antonio Intl: bias −0.22 °F, n = 87KSAT−KSEA Seattle-Tacoma: bias −0.58 °F, n = 87KSEA−KSFO San Francisco Intl: bias −0.95 °F, n = 87KSFO−KTTN Trenton Mercer: bias +0.79 °F, n = 86KTTN+

Source: CastCheck · station coordinates NWS/NOAA · Albers conic projection · stations.csv. Hover a dot for its numbers.

model too cold interval includes zero model too warm dot area ∝ scored days

ECMWF IFS HRES bias as a table
StationModelMAE °FBias °Fn
KATL Atlanta HartsfieldECMWF IFS HRES2.02−1.5887
KAUS Austin BergstromECMWF IFS HRES2.68+2.1587
KBOS Boston LoganECMWF IFS HRES2.09−1.4983
KDCA Washington ReaganECMWF IFS HRES2.04−1.5987
KDEN Denver IntlECMWF IFS HRES2.61+0.4587
KDFW Dallas-Fort WorthECMWF IFS HRES1.59+0.8587
KEWR Newark LibertyECMWF IFS HRES1.77−0.8787
KIAH Houston BushECMWF IFS HRES1.60+0.6587
KLAS Las Vegas Harry ReidECMWF IFS HRES4.36−4.2987
KLAX Los Angeles IntlECMWF IFS HRES1.28−0.6787
KMIA Miami IntlECMWF IFS HRES2.53−2.3187
KMSP Minneapolis-St PaulECMWF IFS HRES1.57−0.9587
KMSY New Orleans IntlECMWF IFS HRES1.95+1.2587
KNYC New York Central ParkECMWF IFS HRES1.84−0.8386
KOKC Oklahoma CityECMWF IFS HRES3.68+2.9587
KORD Chicago O'HareECMWF IFS HRES1.55−0.9387
KPHL Philadelphia IntlECMWF IFS HRES2.62−2.5187
KPHX Phoenix Sky HarborECMWF IFS HRES3.08−2.4087
KSAN San Diego LindberghECMWF IFS HRES1.72−0.5487
KSAT San Antonio IntlECMWF IFS HRES1.48−0.2287
KSEA Seattle-TacomaECMWF IFS HRES1.58−0.5887
KSFO San Francisco IntlECMWF IFS HRES1.51−0.9587
KTTN Trenton MercerECMWF IFS HRES1.62+0.7986

All-model mean bias

Mean bias averaged over every scored model at each station (lead day 1, sampled daily minimum, 365d window, 12Z, bilinear). A station that is cold for all of them is a station property — elevation, coastline, a grid cell that is partly sea — not a model ranking. Dot area grows with the number of scored days; fill is the mean bias (warm = model too warm, cool = too cold). The frame is a latitude/longitude graticule on an Albers projection.

All-model mean bias in °F, lead day 1, sampled daily minimum, 365d window, 12Z, bilinear125°W115°W105°W95°W85°W75°W65°W25°N30°N35°N40°N45°N50°NKATL Atlanta Hartsfield: bias −1.86 °F, n = 255KATL−KAUS Austin Bergstrom: bias +1.43 °F, n = 255KAUS+KBOS Boston Logan: bias −1.73 °F, n = 242KBOS−KDCA Washington Reagan: bias −1.69 °F, n = 255KDCA−KDEN Denver Intl: bias −0.74 °F, n = 255KDEN−KDFW Dallas-Fort Worth: bias −0.53 °F, n = 252KDFW−KEWR Newark Liberty: bias −1.26 °F, n = 254KEWR−KIAH Houston Bush: bias +0.14 °F, n = 256KIAH+KLAS Las Vegas Harry Reid: bias −4.07 °F, n = 255KLAS−KLAX Los Angeles Intl: bias −0.60 °F, n = 256KLAX−KMIA Miami Intl: bias +0.09 °F, n = 255KMIA+KMSP Minneapolis-St Paul: bias −2.08 °F, n = 256KMSP−KMSY New Orleans Intl: bias +0.41 °F, n = 256KMSY+KNYC New York Central Park: bias −0.99 °F, n = 254KNYC−KOKC Oklahoma City: bias +1.27 °F, n = 255KOKC+KORD Chicago O'Hare: bias −0.64 °F, n = 256KORD−KPHL Philadelphia Intl: bias −2.04 °F, n = 256KPHL−KPHX Phoenix Sky Harbor: bias −3.33 °F, n = 256KPHX−KSAN San Diego Lindbergh: bias −0.02 °F, n = 255KSAN−KSAT San Antonio Intl: bias −0.90 °F, n = 256KSAT−KSEA Seattle-Tacoma: bias −1.16 °F, n = 256KSEA−KSFO San Francisco Intl: bias −0.92 °F, n = 256KSFO−KTTN Trenton Mercer: bias +1.01 °F, n = 251KTTN+

Source: CastCheck · station coordinates NWS/NOAA · Albers conic projection · stations.csv. Hover a dot for its numbers.

model too cold interval includes zero model too warm dot area ∝ scored days

All-model mean bias as a table
StationModelsMAE °FBias °Fn
KATL Atlanta Hartsfield11 models2.44−1.86255
KAUS Austin Bergstrom11 models2.78+1.43255
KBOS Boston Logan11 models2.49−1.73242
KDCA Washington Reagan11 models2.39−1.69255
KDEN Denver Intl11 models3.05−0.74255
KDFW Dallas-Fort Worth11 models2.07−0.53252
KEWR Newark Liberty11 models2.18−1.26254
KIAH Houston Bush11 models1.97+0.14256
KLAS Las Vegas Harry Reid11 models4.48−4.07255
KLAX Los Angeles Intl11 models1.88−0.60256
KMIA Miami Intl11 models1.96+0.09255
KMSP Minneapolis-St Paul11 models2.59−2.08256
KMSY New Orleans Intl11 models1.87+0.41256
KNYC New York Central Park11 models2.33−0.99254
KOKC Oklahoma City11 models2.79+1.27255
KORD Chicago O'Hare11 models1.96−0.64256
KPHL Philadelphia Intl11 models2.55−2.04256
KPHX Phoenix Sky Harbor11 models4.12−3.33256
KSAN San Diego Lindbergh11 models2.52−0.02255
KSAT San Antonio Intl11 models2.09−0.90256
KSEA Seattle-Tacoma11 models1.96−1.16256
KSFO San Francisco Intl11 models1.87−0.92256
KTTN Trenton Mercer11 models1.94+1.01251

Data availability

Each model is scored only over its own available period (2026-01-01 → 2026-10-27); the windows above are therefore not identical across models. Pairwise comparisons on the permanent-link pages use common days only.

ModelPeriodScored daysCoverage
ECMWF AIFS Single2026-08-02 → 2026-10-2787
2026-01-012026-10-27
Aurora (GFS init)2026-01-02 → 2026-10-26250
2026-01-012026-10-27
Aurora (IFS init)2026-01-02 → 2026-10-27241
2026-01-012026-10-27
FourCastNet v2 (GFS init)2026-01-02 → 2026-10-26250
2026-01-012026-10-27
FourCastNet v2 (IFS init)2026-01-02 → 2026-10-27241
2026-01-012026-10-27
NCEP GFS2026-08-02 → 2026-10-2787
2026-01-012026-10-27
GraphCast (GFS init)2026-01-02 → 2026-10-26250
2026-01-012026-10-27
GraphCast (IFS init)2026-01-02 → 2026-10-27256
2026-01-012026-10-27
ECMWF IFS HRES2026-08-02 → 2026-10-2787
2026-01-012026-10-27
Pangu-Weather (GFS init)2026-01-02 → 2026-10-26250
2026-01-012026-10-27
Pangu-Weather (IFS init)2026-01-02 → 2026-10-27241
2026-01-012026-10-27
Persistence (baseline)2026-01-01 → 2026-10-27300
2026-01-012026-10-27

Stations23

full station table
  • KNYC — New York Central Park
  • KEWR — Newark Liberty
  • KPHL — Philadelphia Intl
  • KTTN — Trenton Mercer
  • KBOS — Boston Logan
  • KDCA — Washington Reagan
  • KATL — Atlanta Hartsfield
  • KMIA — Miami Intl
  • KORD — Chicago O'Hare
  • KMSP — Minneapolis-St Paul
  • KDFW — Dallas-Fort Worth
  • KIAH — Houston Bush
  • KAUS — Austin Bergstrom
  • KSAT — San Antonio Intl
  • KMSY — New Orleans Intl
  • KOKC — Oklahoma City
  • KDEN — Denver Intl
  • KPHX — Phoenix Sky Harbor
  • KLAS — Las Vegas Harry Reid
  • KLAX — Los Angeles Intl
  • KSAN — San Diego Lindbergh
  • KSFO — San Francisco Intl
  • KSEA — Seattle-Tacoma