Live/Forecast engine/October 9, 2026

Forecast engine

Storm Station’s own forecast, set every day beside NOAA’s National Blend of Models, the National Weather Service forecast, NOAA’s GFS, HRRR and NAM, and ECMWF’s IFS and AIFS, and scored against what the airfields reported, hour by hour. Who came closest, by how much, where, why, and how the engine learns from it. A reading is first checked against the stations around it, so a faulty instrument never counts against a forecast.

Where Storm Station came closer than the NBM

October 9, 2026, the latest day scored, every airfield against what it reported, hour by hour. For temperature, Storm Station came closer than the NBM at 1,310 airfields, was level at 457 and further at 731. Blue is closer, red further, gray level; the brighter the dot, the larger the gap. Scored each morning for the day before.

TemperatureDew pointWindSky cover

1,310 of 2,498 airfields closer than the NBM64 in 100 of those not level

closer by 0.5 F or more629
closer by 0.1 to 0.5 F681
level, within 0.1 F457
further by 0.1 to 0.5 F506
further by 0.5 F or more225

At the 70 panel airports, ringed, Storm Station’s own forecast: closer at 2 of 70, level at 63 and further at 5. Elsewhere, its correction of the NBM, scored on the same hours. Point to an airfield for its figures; choose it for its page.

964 of 2,497 airfields closer than the NBM50 in 100 of those not level

closer by 0.5 F or more418
closer by 0.1 to 0.5 F546
level, within 0.1 F577
further by 0.1 to 0.5 F666
further by 0.5 F or more290

At the 70 panel airports, ringed, Storm Station’s own forecast: closer at 1 of 70, level at 67 and further at 2. Elsewhere, its correction of the NBM, scored on the same hours. Point to an airfield for its figures; choose it for its page.

899 of 2,508 airfields closer than the NBM47 in 100 of those not level

closer by 0.5 mph or more333
closer by 0.1 to 0.5 mph566
level, within 0.1 mph601
further by 0.1 to 0.5 mph687
further by 0.5 mph or more321

At the 70 panel airports, ringed, Storm Station’s own forecast: closer at 12 of 70, level at 54 and further at 4. Elsewhere, its correction of the NBM, scored on the same hours. Point to an airfield for its figures; choose it for its page.

820 of 1,658 airfields closer than the NBM63 in 100 of those not level

closer by 5 points or more601
closer by 1 to 5 points219
level, within 1 point351
further by 1 to 5 points209
further by 5 points or more278

At the 70 panel airports, ringed, Storm Station’s own forecast: closer at 1 of 65, level at 54 and further at 10. Elsewhere, its correction of the NBM, scored on the same hours. Point to an airfield for its figures; choose it for its page.

How close was Storm Station on October 9, 2026?

The temperature Storm Station forecast was off by 2.2 F on average, and within 5 F in 93 of every 100 hours. It ran 0.2 F cool.

Hours of reports: 24 of 24
Temperature forecasts scored: 237,889
Readings checked: 118,260
Readings that failed: 380
Misses over 10 F: 746

Who came closest on October 9, 2026

Every forecast at the same 70 airports, every hour of the day, against what the airports reported. The bar is how much closer than the NBM each came, to the right, or further, to the left, over temperature, dew point, wind and sky cover together.

1Storm Station correctionThe NBM with each station’s habit taken out, learned the day after+9.3%Temperature2.0 FDew point2.3 FWind2.5 mphSky cover15.5 points
2NOAA National Blend of Models (NBM)NOAA’s statistical blend of many models, the national referenceReferenceTemperature2.1 FDew point2.2 FWind2.5 mphSky cover22.4 points
3Storm StationThe engine’s own forecast, as every page prints it-0.9%Temperature2.2 FDew point2.3 FWind2.5 mphSky cover22.6 points
4National Weather Service forecastEach forecast office’s gridded forecast, the National Digital Forecast Database-3.5%Temperature2.1 FDew point2.4 FWind2.6 mphSky cover23.7 points
5ECMWF AIFSArtificial Intelligence Forecasting System, the European Centre’s AI model-7.8%Temperature2.5 FDew point2.2 FWind3.1 mphSky cover20.1 points
6ECMWF IFSIntegrated Forecasting System, the European Centre’s global model-16.5%Temperature2.5 FDew point2.7 FWind3.1 mphSky cover23.2 points
7NOAA HRRRHigh-Resolution Rapid Refresh, NOAA’s 2 mile grid model, run every hour-33.4%Temperature2.8 FDew point3.8 FWind2.8 mphSky cover27.3 points
8NOAA NAMNorth American Mesoscale model, NOAA’s 2 mile grid model over the lower 48-34.4%Temperature3.2 FDew point3.7 FWind2.8 mphSky cover26.4 points
9NOAA GFSGlobal Forecast System, NOAA’s global model-34.9%Temperature2.9 FDew point3.8 FWind3.0 mphSky cover25.3 points

Every forecast, every day

The mean miss of each forecast on each day scored, all at the same 70 airports, over every hour ahead. Lower is closer. Choose a figure; choose a forecast below to show or hide it. Point to a day for its figures.

TemperatureDew pointWindSky cover

What changed

October 14, 2026 Oct 13 is the first day all eight of its forecasts carry the NBM on Storm Station’s own hours, and from it the targets for sky cover in its category and for the first six hours compare Storm Station with that NBM.
October 12, 2026 From the forecasts of 07 UTC Oct 11, the NBM is also scored on Storm Station’s own hours: the cycle Storm Station was made from, counted from the same hour. Until then the NBM was scored only from its own cycle, an hour earlier.
October 12, 2026 Oct 11 is the first day the first six hours are also scored against the last reading carried forward, the measure the first hours must beat (T8).
October 12, 2026 Oct 11 is the first day sky cover is also scored as the airport reports it: each forecast’s cover put in the category the airport’s sensor would report, from clear to overcast, against the category it reported.
October 9, 2026 From the forecasts of 13 UTC on, Storm Station is made from the NBM’s cycle of the hour before each forecast, the newest posted; until then it was made from the cycle of two hours before, while the NBM is scored from its own cycle of the hour.
October 7, 2026 Oct 6 is the first day each of the five models is scored from all eight of the day’s forecasts, so each is ranked beside Storm Station.
October 6, 2026 NOAA GFS with HRRR, the two models laid together as one forecast, retired. Its days are kept as scored.
October 5, 2026 From the forecasts of Oct 5 on, the five models are read from NOAA’s and ECMWF’s own published runs.
October 4, 2026 NOAA’s GFS, HRRR and NAM and ECMWF’s IFS and AIFS joined, each compared on its own, from the forecasts of 19 UTC. A day that holds a model from fewer forecasts than Storm Station sets it aside.
October 3, 2026 Storm Station became the engine’s own forecast at the forecasts of 13 UTC; before them it was the middle of the National Weather Service forecast and NOAA GFS with HRRR. Oct 3 and Oct 4 are scored from both, so its line changes meaning there.

Scorecard

Each forecast set against the NOAA National Blend of Models (NBM), the national reference, at the same places, figure by figure and by how many hours ahead it was made. A blue dot is closer than the NBM, a red one further, a gray one level; the larger the dot, the larger the gap. Point to one for both misses.

Each forecast against the NBM, October 9, 2026: how much closer (+) or further (-) its mean miss was, by figure and hours ahead
Forecast	Temperature					Dew point					Wind					Sky cover
Forecast	1 to 6	7 to 12	13 to 18	19 to 25	All	1 to 6	7 to 12	13 to 18	19 to 25	All	1 to 6	7 to 12	13 to 18	19 to 25	All	1 to 6	7 to 12	13 to 18	19 to 25	All
Storm Stationat the 70 airports	-1%	-2%	-2.3%	-3.9%	-2.4%	-1.9%	-1.3%	+0.4%	-0.8%	-0.8%	+0.8%	0%	+0.8%	+0.8%	+0.6%	-1.6%	-1.7%	-1.4%	+0.2%	-1%
National Weather Service forecastat the 70 airports	-4%	-2.4%	+2.8%	+3.4%	+0.2%	-11.5%	-9.4%	-4%	+0.4%	-5.6%	-5.3%	-2.8%	-3.2%	-0.4%	-2.8%	-4.4%	-5.7%	-5.4%	-8%	-6%
NOAA GFSat the 70 airports	-41.3%	-41%	-36.6%	-26.7%	-35.7%	-76.9%	-70.4%	-72.6%	-66.7%	-71.2%	-18.7%	-19.4%	-20.5%	-19.6%	-19.5%	-18.3%	-20.2%	-13.3%	-3.2%	-13.1%
NOAA HRRRat the 70 airports	-32.8%	-32.7%	-26.9%	-25.4%	-29.8%	-53.4%	-65%	-76.1%	-83.1%	-68.3%	-10.2%	-12.5%	-16.9%	-16.5%	-13.8%	-29.5%	-24.2%	-19%	-17.4%	-21.8%
NOAA NAMat the 70 airports	-57.7%	-52.2%	-42.6%	-38.8%	-47.8%	-91.8%	-65.9%	-57.1%	-47.7%	-62.1%	-8.9%	-8.9%	-11.6%	-11%	-10%	-23.4%	-18.3%	-15.6%	-15.9%	-17.8%
ECMWF IFSat the 70 airports	-28.4%	-24.9%	-14.8%	-10.3%	-18.8%	-28.4%	-14.8%	-17.3%	-22.8%	-20.3%	-22%	-24.2%	-24.9%	-23.1%	-23.2%	-5.9%	-7%	-3.7%	+1%	-3.6%
ECMWF AIFSat the 70 airports	-23.4%	-22.4%	-17.6%	-8.2%	-17.1%	-8.7%	+0.4%	+2.2%	+4.6%	+0.3%	-22.8%	-25%	-26.1%	-24.7%	-24.5%	+8.9%	+8.6%	+9%	+13.3%	+10.2%
Storm Station correctionat every airfield	+8.7%	+9%	+8.2%	+8.1%	+8.3%	+4.6%	+3.7%	+2%	+2.7%	+3.2%	+0.4%	+0.4%	+0.4%	+1.2%	+0.8%	+11.5%	+11.1%	+11.7%	+12.7%	+11.8%

10% or more closer
5 to 10% closer
1 to 5% closer
level
1 to 5% further
5 to 10% further
10% or more further

How the engine works

Everything the engine reads, and what happens to it. Readings are checked against the stations around them before they count. The forecast is served on every page, scored the next day against what happened, and every large miss is explained. What it learns becomes a challenger, and replaces the lead only when it wins over 14 days by a margin the test says is not luck. The counts are October 9, 2026’s.

Measured
- Airfield reports every hour, every airport
- Weather stations every open network heard
- Radar every few minutes
- Rain and hail analysis radars and gauges, every hour
Forecast
- NOAA National Blend of Models NBM, four cycles a day
- National Weather Service forecast compared every day
- NOAA and ECMWF models GFS, HRRR, NAM, IFS, AIFS: compared
- Storm Station engine every input, each place, each hour
- Served every page, card and answer
Checked
- Checked 118,260 readings, 380 set aside
- What happened 24 of 24 hours, hour by hour
Scored
- Scored the next day 2,522,425 forecast hours
- Each large miss explained 746 over 10 F, who was wrong
Learned
- Learned 33,929 figures, 17 days kept
- Challenger scored beside the lead
- Promoted by test 14 days, under 5 in 100 by chance

The targets

Storm Station sets out to be closer than the NBM, the strongest forecast the country publishes, on every figure. A target counts as reached only when it is ahead over 14 days or more by a margin the test says is not luck. Today’s standing, at the same airports:

T1Temperature closer than the NBMStorm Station 2.19 F, the NBM 2.14 FBehind today by 0.05 F
T2Dew point closer than the NBMStorm Station 2.26 F, the NBM 2.24 FBehind today by 0.02 F
T3Wind closer than the NBMStorm Station 2.48 mph, the NBM 2.50 mphAhead today by 0.02 mph
T4Sky cover closer than the NBMStorm Station 22.61 points, the NBM 22.38 pointsBehind today by 0.23 points
T4Sky cover as the airport reports it: in its category more often than the NBMStorm Station -, the NBM -; off by - categories, the NBM -Not yet scored
T5An honest band: 66 to 70 in 100 readings inside it71 in 100 for temperature; the NBM 50Too wide today
T6Rain chance: more skill than the NBMskill 0.24 against 0.25, 1 to 6 hours aheadBehind today by 0.01
T7Rain amount: a higher threat score than the NBM0.18 against 0.17 at 0.01 inchAhead today by 0.01
T8The first six hours: closer than the NBM and than the last reading carried forward-Not yet scored

Every day scored

The NBM and the Storm Station correction at every airfield, scored on the same hours, which is the evidence the correction is tested on; how often the temperature forecast came within 5 F; and how many misses of more than 10 F each day had. The table holds every day; choose one for its own page.