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

Daily verification · methodology v0.3.1

How far off was each weather model?

No group in this view yet has enough scored days to rank (n < 30). Every number below is still published, greyed out, with its sample size.

Scope: raw model output on the native 0.25° grid. The headline variable is the instantaneous 2 m temperature at 00/06/12/18 UTC, scored against the observed 2 m temperature at the same instant (ASOS routine METAR) — the same four instants on both sides, so nothing in the number depends on a model's own diurnal amplitude. Errors are computed in °C and shown in °F (methodology v0.3.1).

Window
the last 30 days
Variable
instantaneous temperature
Init
00Z
Interpolation
bilinear
Data through
2026-10-28
Updated
Next update

Leader, lead day 1

—

no ranked group with an interval yet

n < 30 everywhere

Persistence baseline

4.29°F

[3.69, 5.09]

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

Scored days

29

23 stations · 12 systems

in the last 30 days, lead day 1

Systems ranked

0 of 12 ranked yetn < 30

next update

below n = 30: published, greyed, unranked

Mean absolute error with 95 % confidence intervals

Lead day 1 · the last 30 days · 23 stations pooled · 00Z · bilinear · instantaneous temperature. 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, 30d, 00Z, bilinear0246810mean absolute error, °FECMWF AIFS SingleECMWF AIFS Single: 2.32 °F2.32Aurora (IFS init)Aurora (IFS init): 2.42 °F2.42GraphCast (IFS init)GraphCast (IFS init): 2.56 °F2.56ECMWF IFS HRESECMWF IFS HRES: 2.6 °F2.6NCEP GFSNCEP GFS: 2.63 °F2.63Aurora (GFS init)Aurora (GFS init): 2.7 °F2.7FourCastNet v2 (IFS init)FourCastNet v2 (IFS init): 2.74 °F2.74GraphCast (GFS init)GraphCast (GFS init): 2.78 °F2.78FourCastNet v2 (GFS init)FourCastNet v2 (GFS init): 2.79 °F2.79Pangu-Weather (IFS init)Pangu-Weather (IFS init): 2.8 °F2.8Pangu-Weather (GFS init)Pangu-Weather (GFS init): 2.81 °F2.81Persistence (baseline)Persistence (baseline): 4.29 °F4.29

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 30 days, all 23 stations pooled, 00Z initialization, bilinear interpolation, instantaneous temperature against the observed 2 m temperature at the same instant (ASOS routine METAR). 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. — in the rank column means fewer than 30 scored days, so the group is published but not ranked.

Lead day 1, the last 30 days, 00Z, bilinear, instantaneous temperature.
RankModelMAE °FBias °F±3 °F SkillSkill, debiased (out-of-sample) nvs leader
—ECMWF AIFS Singleaifs_single2.32[2.21, 2.44]−1.05[−1.28, −0.83]71% +0.46[+0.34, +0.56]vs 4.29 (n=29)+0.63n=29 2922.9 stns
—Aurora (IFS init)aurora_ifs2.42—−1.22—69% +0.40—vs 4.06 (n=25)+0.61n=25 2522.9 stns
—GraphCast (IFS init)graphcast_ifs2.56[2.48, 2.67]−1.47[−1.66, −1.31]66% +0.41[+0.30, +0.52]vs 4.31 (n=28)+0.61n=28 2822.9 stns
—ECMWF IFS HRESifs_hres2.60[2.50, 2.73]−0.94[−1.10, −0.79]65% +0.39[+0.27, +0.50]vs 4.29 (n=29)+0.55n=29 2922.9 stns
—NCEP GFSgfs2.63[2.54, 2.73]+0.48[0.19, 0.82]66% +0.39[+0.27, +0.49]vs 4.29 (n=29)+0.51n=29 2922.9 stns
—Aurora (GFS init)aurora_gfs2.70—−1.54—63% +0.37—vs 4.28 (n=17)+0.59n=17 1722.9 stns
—FourCastNet v2 (IFS init)fourcastnet_ifs2.74[2.61, 2.85]−1.07[−1.31, −0.78]64% +0.36[+0.27, +0.46]vs 4.29 (n=29)+0.49n=29 2922.9 stns
—GraphCast (GFS init)graphcast_gfs2.78—−1.65—62% +0.35—vs 4.28 (n=17)+0.59n=17 1722.9 stns
—FourCastNet v2 (GFS init)fourcastnet_gfs2.79—−1.26—64% +0.35—vs 4.28 (n=17)+0.49n=17 1722.9 stns
—Pangu-Weather (IFS init)pangu_ifs2.80—−1.10—62% +0.34—vs 4.24 (n=27)+0.58n=27 2722.9 stns
—Pangu-Weather (GFS init)pangu_gfs2.81—−1.16—63% +0.34—vs 4.28 (n=17)+0.58n=17 1722.9 stns
—Persistence (baseline)persistence · baseline4.29[3.69, 5.09]+0.25[−0.23, 0.76]48% —— 29all 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 30 days, 00Z, bilinear, instantaneous temperature.
RankModelMAE °FBias °FSkillnvs leader
—ECMWF AIFS Singleaifs_single2.54[2.41, 2.66]−1.10[−1.38, −0.85]+0.64[+0.54, +0.72]vs 7.02 (n=29)2922.9 stns
—Aurora (IFS init)aurora_ifs2.71—−1.35—+0.60—vs 6.80 (n=24)2422.9 stns
—GraphCast (IFS init)graphcast_ifs2.84[2.69, 2.99]−1.60[−1.83, −1.42]+0.60[+0.49, +0.69]vs 7.05 (n=28)2822.9 stns
—NCEP GFSgfs2.92[2.82, 3.02]+0.54[0.30, 0.79]+0.58[+0.49, +0.67]vs 7.02 (n=29)2922.9 stns
—FourCastNet v2 (IFS init)fourcastnet_ifs2.96[2.80, 3.10]−1.19[−1.37, −0.97]+0.58[+0.49, +0.66]vs 7.02 (n=29)2922.9 stns
—FourCastNet v2 (GFS init)fourcastnet_gfs3.04—−1.39—+0.56—vs 6.85 (n=16)1623 stns
—GraphCast (GFS init)graphcast_gfs3.04—−1.72—+0.56—vs 6.85 (n=16)1623 stns
—ECMWF IFS HRESifs_hres3.05[2.93, 3.17]−1.17[−1.46, −0.87]+0.57[+0.47, +0.65]vs 7.02 (n=29)2922.9 stns
—Pangu-Weather (IFS init)pangu_ifs3.18—−1.31—+0.54—vs 6.95 (n=26)2622.9 stns
—Aurora (GFS init)aurora_gfs3.22—−1.88—+0.53—vs 6.85 (n=16)1623 stns
—Pangu-Weather (GFS init)pangu_gfs3.37—−1.45—+0.51—vs 6.85 (n=16)1623 stns
—Persistence (baseline)persistence · baseline7.02[5.76, 8.76]+0.88[−0.28, 2.26]—29all 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 30 days, 00Z, bilinear, instantaneous temperature.
RankModelMAE °FBias °FSkillnvs leader
—ECMWF AIFS Singleaifs_single3.19[2.94, 3.44]−1.07[−1.47, −0.67]+0.60[+0.52, +0.67]vs 7.93 (n=29)2922.9 stns
—Aurora (IFS init)aurora_ifs3.21—−1.15—+0.58—vs 7.70 (n=24)2423 stns
—GraphCast (IFS init)graphcast_ifs3.62[3.35, 3.87]−1.51[−2.03, −1.07]+0.55[+0.45, +0.64]vs 7.97 (n=28)2822.9 stns
—ECMWF IFS HRESifs_hres3.75[3.33, 4.23]−1.35[−1.77, −1.03]+0.53[+0.44, +0.59]vs 7.93 (n=29)2922.9 stns
—GraphCast (GFS init)graphcast_gfs3.78—−1.91—+0.52—vs 7.93 (n=18)1822.9 stns
—FourCastNet v2 (IFS init)fourcastnet_ifs3.85[3.37, 4.35]−1.42[−1.78, −1.09]+0.51[+0.42, +0.60]vs 7.93 (n=29)2922.9 stns
—FourCastNet v2 (GFS init)fourcastnet_gfs3.92—−1.94—+0.51—vs 7.93 (n=18)1822.9 stns
—Pangu-Weather (GFS init)pangu_gfs3.96—−1.69—+0.50—vs 7.93 (n=18)1822.9 stns
—NCEP GFSgfs3.96[3.52, 4.48]+0.20[−0.35, 0.77]+0.50[+0.39, +0.58]vs 7.93 (n=29)2922.9 stns
—Aurora (GFS init)aurora_gfs4.06—−2.32—+0.49—vs 7.93 (n=18)1822.9 stns
—Pangu-Weather (IFS init)pangu_ifs4.14—−1.61—+0.47—vs 7.81 (n=26)2623 stns
—Persistence (baseline)persistence · baseline7.93[6.37, 9.89]+1.49[−0.13, 3.58]—29all 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 30 days, 00Z, bilinear, instantaneous temperature.
RankModelMAE °FBias °FSkillnvs leader
—Aurora (IFS init)aurora_ifs4.25—−0.13—+0.47—vs 8.00 (n=24)2422.9 stns
—FourCastNet v2 (GFS init)fourcastnet_gfs4.53—−1.64—+0.43—vs 8.01 (n=18)1822.9 stns
—ECMWF AIFS Singleaifs_single4.55[3.99, 5.23]−1.57[−2.13, −0.98]+0.45[+0.36, +0.53]vs 8.25 (n=29)2922.9 stns
—Pangu-Weather (GFS init)pangu_gfs4.71—−1.33—+0.41—vs 8.01 (n=18)1822.9 stns
—FourCastNet v2 (IFS init)fourcastnet_ifs4.77[4.30, 5.35]−1.76[−2.36, −1.09]+0.42[+0.32, +0.52]vs 8.25 (n=29)2922.9 stns
—ECMWF IFS HRESifs_hres4.84[4.23, 5.62]−1.31[−2.34, −0.02]+0.41[+0.32, +0.49]vs 8.25 (n=29)2922.9 stns
—Aurora (GFS init)aurora_gfs4.92—−2.12—+0.39—vs 8.01 (n=18)1822.9 stns
—GraphCast (IFS init)graphcast_ifs4.94[4.15, 6.03]−1.81[−2.94, −0.67]+0.41[+0.29, +0.52]vs 8.31 (n=28)2822.9 stns
—GraphCast (GFS init)graphcast_gfs5.11—−1.78—+0.36—vs 8.01 (n=18)1822.9 stns
—NCEP GFSgfs5.16[4.42, 5.95]−0.02[−0.93, 0.95]+0.37[+0.25, +0.48]vs 8.25 (n=29)2922.9 stns
—Pangu-Weather (IFS init)pangu_ifs5.21—−2.00—+0.36—vs 8.15 (n=26)2622.9 stns
—Persistence (baseline)persistence · baseline8.25[6.67, 9.86]+2.04[−0.16, 4.63]—29all days

The same four samples as a daily maximum and minimum

Lead day 1, the last 30 days, 00Z, 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
—ECMWF AIFS Singleaifs_single1.87[1.66, 2.08]−0.30[−0.47, −0.10]+0.59vs 4.55 (n=29)29
—NCEP GFSgfs2.29[2.05, 2.56]−0.43[−0.76, −0.12]+0.50vs 4.55 (n=29)29
—GraphCast (IFS init)graphcast_ifs2.33[2.10, 2.56]−1.49[−1.65, −1.32]+0.49vs 4.54 (n=28)28
—Aurora (IFS init)aurora_ifs2.34—−1.62—+0.47vs 4.39 (n=25)25
—ECMWF IFS HRESifs_hres2.52[2.31, 2.77]−1.26[−1.58, −0.95]+0.45vs 4.55 (n=29)29
—Pangu-Weather (IFS init)pangu_ifs2.59—−0.02—+0.43vs 4.54 (n=27)27
—FourCastNet v2 (IFS init)fourcastnet_ifs2.63[2.30, 2.92]−1.38[−1.52, −1.22]+0.42vs 4.55 (n=29)29
—Pangu-Weather (GFS init)pangu_gfs2.76—−0.59—+0.40vs 4.62 (n=17)17
—FourCastNet v2 (GFS init)fourcastnet_gfs2.78—−1.62—+0.40vs 4.62 (n=17)17
—Aurora (GFS init)aurora_gfs3.06—−2.59—+0.34vs 4.62 (n=17)17
—GraphCast (GFS init)graphcast_gfs3.08—−2.47—+0.33vs 4.62 (n=17)17
—Persistence (baseline)persistence4.55[3.82, 5.25]+0.20[−0.27, 0.65]—29all days

Sampled daily minimumtmin_s

RankModelMAE °FBias °FSkillnvs leader
—Aurora (IFS init)aurora_ifs2.05—−0.23—+0.42vs 3.53 (n=25)25
—GraphCast (GFS init)graphcast_gfs2.14—−0.22—+0.41vs 3.61 (n=17)17
—GraphCast (IFS init)graphcast_ifs2.16[2.00, 2.31]−0.64[−0.91, −0.38]+0.43vs 3.78 (n=28)28
—Aurora (GFS init)aurora_gfs2.17—−0.07—+0.40vs 3.61 (n=17)17
—Pangu-Weather (GFS init)pangu_gfs2.26—−0.47—+0.37vs 3.61 (n=17)17
—ECMWF IFS HRESifs_hres2.29[1.95, 2.70]−0.32[−0.59, −0.08]+0.39vs 3.75 (n=29)29
—ECMWF AIFS Singleaifs_single2.30[2.04, 2.60]−1.35[−1.66, −1.02]+0.39vs 3.75 (n=29)29
—Pangu-Weather (IFS init)pangu_ifs2.31—−0.73—+0.38vs 3.73 (n=27)27
—FourCastNet v2 (GFS init)fourcastnet_gfs2.65—−0.94—+0.27vs 3.61 (n=17)17
—FourCastNet v2 (IFS init)fourcastnet_ifs2.69[2.53, 2.86]−0.81[−1.29, −0.22]+0.28vs 3.75 (n=29)29
—NCEP GFSgfs2.95[2.48, 3.44]+1.86[1.21, 2.56]+0.21vs 3.75 (n=29)29
—Persistence (baseline)persistence3.75[3.19, 4.56]+0.28[−0.37, 0.90]—29all days

Every model × every lead day

MAE in °F with the bias and n underneath, the last 30 days, 00Z, bilinear, instantaneous temperature. The sparkline is the same model's MAE across lead days 1–9 on a shared vertical scale.

Modeld0d1d2d3d4d5d6d7d8d9 lead 1–9
ECMWF AIFS Singleaifs_single2.22−1.02 · n=292.32−1.05 · n=292.39−1.05 · n=292.54−1.10 · n=292.82−1.09 · n=293.19−1.07 · n=293.76−1.09 · n=294.55−1.57 · n=295.11−1.38 · n=295.49−0.56 · n=29ECMWF AIFS Single MAE by lead day 1 to 9
Aurora (GFS init)aurora_gfs2.40−1.32 · n=172.70−1.54 · n=172.98−1.83 · n=173.22−1.88 · n=163.62−2.09 · n=174.06−2.32 · n=184.60−2.00 · n=184.92−2.12 · n=185.47−2.04 · n=186.01−2.05 · n=18Aurora (GFS init) MAE by lead day 1 to 9
Aurora (IFS init)aurora_ifs2.30−1.23 · n=252.42−1.22 · n=252.52−1.30 · n=252.71−1.35 · n=242.93−1.29 · n=243.21−1.15 · n=243.81−0.60 · n=244.25−0.13 · n=244.62+0.06 · n=235.35+1.10 · n=23Aurora (IFS init) MAE by lead day 1 to 9
FourCastNet v2 (GFS init)fourcastnet_gfs2.69−1.34 · n=172.79−1.26 · n=172.90−1.31 · n=173.04−1.39 · n=163.44−1.50 · n=173.92−1.94 · n=184.10−1.58 · n=184.53−1.64 · n=184.86−1.43 · n=185.86−1.41 · n=18FourCastNet v2 (GFS init) MAE by lead day 1 to 9
FourCastNet v2 (IFS init)fourcastnet_ifs2.66−1.35 · n=292.74−1.07 · n=292.81−1.10 · n=292.96−1.19 · n=293.31−1.31 · n=293.85−1.42 · n=294.16−1.57 · n=294.77−1.76 · n=295.27−1.80 · n=285.79−1.66 · n=28FourCastNet v2 (IFS init) MAE by lead day 1 to 9
NCEP GFSgfs2.50+0.60 · n=292.63+0.48 · n=292.75+0.44 · n=292.92+0.54 · n=293.36+0.46 · n=293.96+0.20 · n=294.62−0.03 · n=295.16−0.02 · n=296.01−0.21 · n=296.42−1.22 · n=29NCEP GFS MAE by lead day 1 to 9
GraphCast (GFS init)graphcast_gfs2.50−1.36 · n=172.78−1.65 · n=172.91−1.76 · n=173.04−1.72 · n=163.33−1.74 · n=173.78−1.91 · n=184.46−1.75 · n=185.11−1.78 · n=186.02−1.76 · n=186.87−2.06 · n=18GraphCast (GFS init) MAE by lead day 1 to 9
GraphCast (IFS init)graphcast_ifs2.40−1.29 · n=282.56−1.47 · n=282.69−1.56 · n=282.84−1.60 · n=283.11−1.51 · n=283.62−1.51 · n=284.21−1.44 · n=284.94−1.81 · n=285.55−1.99 · n=286.08−2.08 · n=28GraphCast (IFS init) MAE by lead day 1 to 9
ECMWF IFS HRESifs_hres2.38−0.80 · n=292.60−0.94 · n=292.81−1.03 · n=293.05−1.17 · n=293.33−1.31 · n=293.75−1.35 · n=294.32−1.38 · n=294.84−1.31 · n=295.02−0.99 · n=295.66−0.73 · n=29ECMWF IFS HRES MAE by lead day 1 to 9
Pangu-Weather (GFS init)pangu_gfs2.51−0.76 · n=172.81−1.16 · n=173.06−1.42 · n=173.37−1.45 · n=163.57−1.47 · n=173.96−1.69 · n=184.59−1.25 · n=184.71−1.33 · n=186.15−1.04 · n=186.78−0.66 · n=18Pangu-Weather (GFS init) MAE by lead day 1 to 9
Pangu-Weather (IFS init)pangu_ifs2.59−0.77 · n=272.80−1.10 · n=272.99−1.20 · n=273.18−1.31 · n=263.58−1.35 · n=264.14−1.61 · n=264.66−1.69 · n=265.21−2.00 · n=265.91−2.14 · n=256.31−1.21 · n=25Pangu-Weather (IFS init) MAE by lead day 1 to 9
Persistence (baseline)persistence—4.29+0.25 · n=295.92+0.56 · n=297.02+0.88 · n=297.58+1.25 · n=297.93+1.49 · n=297.92+1.77 · n=298.25+2.04 · n=298.40+2.35 · n=298.35+2.81 · n=29Persistence (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, instantaneous temperature, 30d window, 00Z, 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, instantaneous temperature, 30d window, 00Z, bilinear125°W115°W105°W95°W85°W75°W65°W25°N30°N35°N40°N45°N50°NKATL Atlanta Hartsfield: bias −1.29 °F, n = 29KATL−KAUS Austin Bergstrom: bias +0.96 °F, n = 29KAUS+KBOS Boston Logan: bias −1.81 °F, n = 28KBOS−KDCA Washington Reagan: bias −1.53 °F, n = 29KDCA−KDEN Denver Intl: bias +1.10 °F, n = 29KDEN+KDFW Dallas-Fort Worth: bias −0.12 °F, n = 29KDFW−KEWR Newark Liberty: bias −1.67 °F, n = 29KEWR−KIAH Houston Bush: bias +0.53 °F, n = 29KIAH+KLAS Las Vegas Harry Reid: bias −2.98 °F, n = 29KLAS−KLAX Los Angeles Intl: bias −1.43 °F, n = 29KLAX−KMIA Miami Intl: bias −2.67 °F, n = 29KMIA−KMSP Minneapolis-St Paul: bias −1.02 °F, n = 29KMSP−KMSY New Orleans Intl: bias −0.27 °F, n = 29KMSY−KNYC New York Central Park: bias −1.10 °F, n = 28KNYC−KOKC Oklahoma City: bias +1.96 °F, n = 29KOKC+KORD Chicago O'Hare: bias −1.77 °F, n = 29KORD−KPHL Philadelphia Intl: bias −2.55 °F, n = 29KPHL−KPHX Phoenix Sky Harbor: bias −3.10 °F, n = 29KPHX−KSAN San Diego Lindbergh: bias −1.76 °F, n = 29KSAN−KSAT San Antonio Intl: bias +0.33 °F, n = 29KSAT+KSEA Seattle-Tacoma: bias −0.78 °F, n = 29KSEA−KSFO San Francisco Intl: bias −1.13 °F, n = 29KSFO−KTTN Trenton Mercer: bias +0.47 °F, n = 29KTTN+

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 HRES2.33−1.2929
KAUS Austin BergstromECMWF IFS HRES3.17+0.9629
KBOS Boston LoganECMWF IFS HRES2.34−1.8128
KDCA Washington ReaganECMWF IFS HRES2.39−1.5329
KDEN Denver IntlECMWF IFS HRES2.74+1.1029
KDFW Dallas-Fort WorthECMWF IFS HRES2.01−0.1229
KEWR Newark LibertyECMWF IFS HRES2.43−1.6729
KIAH Houston BushECMWF IFS HRES2.50+0.5329
KLAS Las Vegas Harry ReidECMWF IFS HRES3.54−2.9829
KLAX Los Angeles IntlECMWF IFS HRES2.41−1.4329
KMIA Miami IntlECMWF IFS HRES2.96−2.6729
KMSP Minneapolis-St PaulECMWF IFS HRES2.53−1.0229
KMSY New Orleans IntlECMWF IFS HRES2.19−0.2729
KNYC New York Central ParkECMWF IFS HRES2.38−1.1028
KOKC Oklahoma CityECMWF IFS HRES3.38+1.9629
KORD Chicago O'HareECMWF IFS HRES2.44−1.7729
KPHL Philadelphia IntlECMWF IFS HRES3.06−2.5529
KPHX Phoenix Sky HarborECMWF IFS HRES3.60−3.1029
KSAN San Diego LindberghECMWF IFS HRES2.99−1.7629
KSAT San Antonio IntlECMWF IFS HRES2.10+0.3329
KSEA Seattle-TacomaECMWF IFS HRES1.92−0.7829
KSFO San Francisco IntlECMWF IFS HRES2.43−1.1329
KTTN Trenton MercerECMWF IFS HRES2.04+0.4729

All-model mean bias

Mean bias averaged over every scored model at each station (lead day 1, instantaneous temperature, 30d window, 00Z, 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, instantaneous temperature, 30d window, 00Z, bilinear125°W115°W105°W95°W85°W75°W65°W25°N30°N35°N40°N45°N50°NKATL Atlanta Hartsfield: bias −1.29 °F, n = 29KATL−KAUS Austin Bergstrom: bias +0.20 °F, n = 29KAUS+KBOS Boston Logan: bias −1.44 °F, n = 28KBOS−KDCA Washington Reagan: bias −1.05 °F, n = 29KDCA−KDEN Denver Intl: bias −1.05 °F, n = 29KDEN−KDFW Dallas-Fort Worth: bias −1.79 °F, n = 29KDFW−KEWR Newark Liberty: bias −1.19 °F, n = 29KEWR−KIAH Houston Bush: bias −0.49 °F, n = 29KIAH−KLAS Las Vegas Harry Reid: bias −3.07 °F, n = 29KLAS−KLAX Los Angeles Intl: bias −2.11 °F, n = 29KLAX−KMIA Miami Intl: bias −1.01 °F, n = 29KMIA−KMSP Minneapolis-St Paul: bias −1.62 °F, n = 29KMSP−KMSY New Orleans Intl: bias −0.76 °F, n = 29KMSY−KNYC New York Central Park: bias −0.35 °F, n = 28KNYC−KOKC Oklahoma City: bias +0.61 °F, n = 29KOKC+KORD Chicago O'Hare: bias −0.98 °F, n = 29KORD−KPHL Philadelphia Intl: bias −1.63 °F, n = 29KPHL−KPHX Phoenix Sky Harbor: bias −2.69 °F, n = 29KPHX−KSAN San Diego Lindbergh: bias −1.98 °F, n = 29KSAN−KSAT San Antonio Intl: bias −0.66 °F, n = 29KSAT−KSEA Seattle-Tacoma: bias −0.56 °F, n = 29KSEA−KSFO San Francisco Intl: bias −1.38 °F, n = 29KSFO−KTTN Trenton Mercer: bias +1.27 °F, n = 29KTTN+

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 models2.52−1.2929
KAUS Austin Bergstrom11 models3.25+0.2029
KBOS Boston Logan11 models2.28−1.4428
KDCA Washington Reagan11 models2.12−1.0529
KDEN Denver Intl11 models2.89−1.0529
KDFW Dallas-Fort Worth11 models2.64−1.7929
KEWR Newark Liberty11 models2.41−1.1929
KIAH Houston Bush11 models2.48−0.4929
KLAS Las Vegas Harry Reid11 models3.44−3.0729
KLAX Los Angeles Intl11 models3.15−2.1129
KMIA Miami Intl11 models1.95−1.0129
KMSP Minneapolis-St Paul11 models2.73−1.6229
KMSY New Orleans Intl11 models2.22−0.7629
KNYC New York Central Park11 models2.31−0.3528
KOKC Oklahoma City11 models2.87+0.6129
KORD Chicago O'Hare11 models2.15−0.9829
KPHL Philadelphia Intl11 models2.63−1.6329
KPHX Phoenix Sky Harbor11 models3.76−2.6929
KSAN San Diego Lindbergh11 models3.27−1.9829
KSAT San Antonio Intl11 models2.22−0.6629
KSEA Seattle-Tacoma11 models2.29−0.5629
KSFO San Francisco Intl11 models3.02−1.3829
KTTN Trenton Mercer11 models2.33+1.2729

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 → 2026-10-27216
2026-01-012026-10-27
Aurora (IFS init)2026-01-02 → 2026-10-27232
2026-01-012026-10-27
FourCastNet v2 (GFS init)2026-01-02 → 2026-10-27216
2026-01-012026-10-27
FourCastNet v2 (IFS init)2026-01-02 → 2026-10-27237
2026-01-012026-10-27
NCEP GFS2026-08-02 → 2026-10-2787
2026-01-012026-10-27
GraphCast (GFS init)2026-01-02 → 2026-10-27216
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 → 2026-10-27216
2026-01-012026-10-27
Pangu-Weather (IFS init)2026-01-02 → 2026-10-27234
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