Skip to content
HurriTracker
LIVE · 3 ACTIVE STORMS

Spaghetti models

Hurricane spaghetti models for Rachel, Nolo and Nineteen-E, from the 06 UTC Oct 1 run in NHC’s public guidance feed. Spaghetti plots are raw model guidance, not forecasts: the NHC official track and its cone are the forecast.

C2

Spaghetti models for Hurricane Rachel

NHC adv 17 ·

11 guidance models, 11 of them out to 72 hours. At 72 hours the middle of the pack is about 340 miles (550 km) west of Rachel’s position at the start of the run, and the two tracks furthest apart (HAFS-B and COAMPS-TC) end 220 miles (350 km) apart. The official NHC point at 72 hours is 35 miles (55 km) from the middle of the pack.

Hurricane Rachel tracker
TS

Spaghetti models for Tropical Storm Nolo

NHC adv 43 ·

11 guidance models, 11 of them out to 72 hours. At 72 hours the middle of the pack is about 680 miles (1,100 km) west of Nolo’s position at the start of the run, and the two tracks furthest apart (HWRF and HMON) end 140 miles (230 km) apart. The official NHC point at 72 hours is 25 miles (40 km) from the middle of the pack.

Tropical Storm Nolo tracker
TD

Spaghetti models for Tropical Depression Nineteen-E

NHC adv 8 ·

11 guidance models, 7 of them out to 72 hours. At 72 hours the middle of the pack is about 1,150 miles (1,860 km) east of Nineteen-E’s position at the start of the run, and the two tracks furthest apart (HAFS-A and HMON) end 250 miles (400 km) apart. NHC’s official forecast stops at 24 hours, so treat model tracks beyond that with extra caution.

Tropical Depression Nineteen-E tracker

What are spaghetti models?

A spaghetti plot puts the forecast tracks from many computer models on one map. Each line is where one model expects the center of the storm to go over the next few days, and the tangle of colored lines is what gives the plot its name. NHC calls the same thing model guidance: input that its forecasters weigh, not a forecast.

When the lines bundle tightly, the models agree and confidence in the track is higher. When they fan out, the future track is uncertain. The lines show the path of the center only; they say nothing about how large the storm is or how far its wind, rain and storm surge will reach.

Hurricane model tracks: the computer models on the plot

These are the models you are most likely to see, with the ATCF ID that labels each line. IDs ending in I are the “interpolated” early versions explained under update times below.

IDModelTypeWhat NHC uses it forOn our plots
OFCLNHC official forecastOfficial forecastThe forecast behind the cone, watches and warningsYes, bold line
AVNIGlobal Forecast System (GFS), NOAAGlobal dynamicalTrack and intensity; every 6 hYes
EMXIECMWF global model (the “Euro”)Global dynamicalTrack and intensity; every 12 hNo: data restricted
UKXIUK Met Office global modelGlobal dynamicalTrack and intensity; every 12 hYes, when published
CMCICanadian global modelGlobal dynamicalTrack and intensity; every 12 hYes, when published
NVGINavy Global Environmental Model (NAVGEM)Global dynamicalTrack and intensity; every 6 hYes
HFAI, HFBIHurricane Analysis and Forecast System (HAFS-A, HAFS-B)Regional hurricane modelTrack and intensity; run on request from NHCYes
HWFIHurricane Weather Research and Forecast system (HWRF)Regional hurricane modelTrack and intensity; run on requestYes
HMNIHMON hurricane modelRegional hurricane modelTrack and intensity; run on requestYes
CTCICOAMPS-TC, US NavyRegional hurricane modelTrack and intensity; from the first advisoryYes
AEMIGFS ensemble mean (31 members)Ensemble meanTrack and intensityYes
TVCNVariable consensusSimple consensusTrack: average of at least 2 of GDMI, AEMI, EMXI, UKMI, HFAI, CTCIYes
HCCAHFIP Corrected Consensus ApproachCorrected consensusTrack and intensity: members weighted by past performanceYes
GDMIGoogle DeepMind ensemble meanAI ensemble meanTrack and intensityNo
CLP5CLIPER5 climatology and persistenceStatistical baselineA yardstick for the skill of other track forecastsNo

Sources: NHC model summary (updated June 1, 2026) and NHC’s ATCF technique list. “On our plots” means drawn when the model has run for that storm and NHC’s public feed carries it.

How to read them without fooling yourself

  • Look at the cluster, not the outlier. One line far from the rest is rarely the answer.
  • Check the run time on the map. A run from yesterday can be badly out of date for a fast-changing storm.
  • Give simple aids less weight. Some sites also draw CLP5, a climatology-and-persistence model that NHC keeps as a skill yardstick, not as a forecast.
  • Tracks become far less reliable after three to five days, even when the lines agree: NHC’s own 2026 Atlantic cone circle grows from 39 nautical miles at 24 hours to 200 at 120 hours.

Spaghetti models vs the forecast cone

The official NHC track is built by forecasters who weigh all of these models, and NHC says that on average its official forecasts have smaller errors than any individual model. They also jump around less: over 2023–2025, NHC’s Atlantic 4- and 5-day forecasts changed less from one advisory to the next than the GFS, ECMWF, UKMET and HCCA guidance did.

The forecast cone around that track is not drawn from today’s model spread. It comes from NHC’s historical errors, sized so the storm’s center stays inside about two-thirds of the time. So a wide spread on the spaghetti plot is a sign that this storm may be harder to forecast than usual, even though its cone looks the same size as any other. If the models and the official forecast disagree, plan around the official forecast and your local warnings.

How often spaghetti models update

Model runs are tied to four synoptic times a day: 00, 06, 12 and 18 UTC. The GFS, NAVGEM, the hurricane models and the consensus aids run every six hours; the ECMWF, UKMET and Canadian models run twice a day, at 00 and 12 UTC.

The big models finish late. NHC releases its 12 UTC forecast at 15 UTC, but the 12 UTC GFS is not ready until about 16 UTC. So forecasters use “early” versions: the previous run of each model, shifted so it starts at the storm’s current position and strength. These are the interpolated models, and their IDs usually end in I (AVNI is the early GFS). NHC builds them from a run up to 12 hours old.

This page checks NHC’s public ATCF feed (the “a-deck”) every 15 minutes while a storm is active. It draws the newest run that already includes NHC’s official track, in bold, beside every global, hurricane and consensus model that ran, so new lines usually appear a few hours after each synoptic time, once the advisory is out.

Spaghetti models FAQ

Which spaghetti model is the most accurate?

No single model wins every storm. In NHC’s 2025 verification report, the consensus aids HCCA, FSSE and TVCA were the best Atlantic track guidance over 2023–2025, with the official forecast about as skillful, and ECMWF was the best single model. In 2025, Google DeepMind’s ensemble mean slightly beat NHC’s own Atlantic track forecasts from 12 to 72 hours.

Are spaghetti models official forecasts?

No. They are guidance that forecasters consult. Watches, warnings and the forecast cone come only from the official forecast issued by the National Hurricane Center or your national weather service. Emergency managers plan around that official forecast, so it is the one to act on, even when a single model line looks alarming.

Why is the European model missing from this plot?

Because its data is restricted. NHC’s model summary marks the ECMWF runs, and the UK Met Office’s own track, as restricted by agreements with the data provider, so they are not in the public ATCF feed this page draws from. The UKMET line on our plot is usually UKXI, a UKMET track found with the GFS storm tracker, shown when the feed carries it.

Do spaghetti models show how strong a storm will get?

Not on the plot, which shows only where each model takes the center. Most models also forecast intensity, and NHC’s intensity guidance includes the HAFS hurricane models, the statistical models DSHP and LGEM, and consensus aids such as IVCN and HCCA. The official wind forecast for each storm is on its tracker page.