Which hurricane models should you trust in 2025? » Yale Climate Connections

With an ever-growing plethora of hurricane computer forecast models out there – including two new ones that became operational just last year – it can be ever more puzzling to figure out which model to believe. But the best approach remains: Don’t take any of them as gospel. Put your trust in the National Hurricane Center, or NHC, forecast.

During the highly active 2o24 Atlantic hurricane season, the NHC made record-accurate track forecasts at every time interval (12-, 24-, 36-, 48-, 60-, 72-, 96-, and 120-hour forecasts), and the official forecast outperformed all of the individual models in almost all cases, according to the 2024 NHC Forecast Verification Report.

Best track models in 2024: the European and GFS

In 2024, the official NHC track forecast outperformed all models at four and five days out. However, NHC’s top three consensus models — which use a blend of four to six of the individual models to create a “consensus” model forecast — were nearly as good as the official NHC forecast (Fig. 2). Notably, none of the individual models outperformed the official forecast or the three consensus models at any time frame.

The European and GFS models were the top performers in 2024, with the GFS making better four- and five-day forecasts. The HMON model also performed well at all lead times, but the UKMET and HWRF generally performed less well. The new HAFS-A and HAFS-B models did well for forecasts of 36 hours or less but fell off in accuracy significantly for four- and five-day forecasts. The best five-day model forecasts were made by the HWRF model.

To interpret Fig. 2, note that the CLIPER5 model (which combines the words “climatology” and “persistence” to show the nature of the forecasts it makes) is tough to outperform at short-term forecasts, since a hurricane will tend to keep moving in the same direction and at the same speed as at its initial point (this is called persistence). For that reason, the skill curve in Fig. 2 shows relatively low skill for NHC forecasts as well as individual model runs for short-term forecasts out to one day. The skill compared to CLIPER5 increases for forecasts between one and three days, when persistence tends not to be a good forecast. (Hurricanes generally don’t move in a straight line at a constant speed for days on end.) Beyond three-day forecasts, NHC forecast skill starts to drop off, as the CLIPER5 model starts weighting its forecasts using climatology, which becomes tougher to beat at long ranges.

Here is a list of some of the top hurricane forecast models used by NHC:

European: The European Center for Medium-range Weather Forecasting (ECMWF) global forecast model
GFS: The National Oceanic and Atmospheric Administration (NOAA) Global Forecast System model
UKMET: The United Kingdom Met Office’s global forecast model
HAFS-A and HAFS-B: Two versions of the Hurricane Analysis and Forecast System (newly added in 2024)
HMON: Hurricanes in a Multi-scale Ocean-coupled Non-hydrostatic regional model, initialized using GFS data
HWRF: Hurricane Weather and Research Forecasting regional model, initialized using GFS data
COAMPS: The Navy’s COAMPS-TC regional model, initialized using GFS data

NHC intensity forecasts in 2024: larger errors, but harder storms to predict

Though intensity forecasts have not improved as dramatically as track forecasts over the past 30 years, there has been a notable decrease since around 2010 in intensity errors. Official NHC intensity forecast errors in the Atlantic in 2024 were higher than the previous five-year mean error for forecasts of up to four days in the future, but lower for five-day forecasts.

Because of a large number of rapid intensification events, the season’s storms were notably more challenging than normal to predict. Thus, even though NHC’s errors increased, the skill of their intensity forecasts was near record-high (see Fig. 5 below). Notably, NHC’s low bias for rapid intensity forecasts has decreased sharply in recent years (Fig. 4). It was 26 kt (30 mph) from 2010-2014, and decreased to 16 kt (18 mph) from 2020-2024.

Best intensity forecasts in 2024: The blends and NHC take the prize

In 2024, the official NHC intensity forecast outperformed all individual models, except for four- and five-day forecasts, where the statistical LGEM model slightly outperformed the official forecast. Four consensus models were close to or slightly better than the official forecast at some lead times. Over the past few years, the five top intensity models have typically been:

• The regional/dynamical HWRF, HMON, and COAMPS-TC models (these models subdivide the atmosphere into a 3D grid around the storm and solve the atmospheric equations of fluid flow at each point on the grid)
• The statistics-based LGEM and DSHP models (DSHP is the SHIPS model with inland decay of a storm factored in).

In 2024, the top dynamic models were actually the older HMON and HWRF models. The newer models designed to replace them, the HAFS-A and HAFS-B, did less well. Mean intensity forecast errors in 2024 (expressed in maximum sustained winds) were about 8 mph at 24 hours and increased to about 13 mph for three- to five-day forecasts. The official forecasts had little bias, either high or low.

Two of the top-performing global dynamical models for hurricane track, the European (ECMWF) and GFS models, are typically not considered by NHC forecasters when making intensity forecasts. These models have traditionally made poor intensity forecasts, and this was the case again in 2024 as seen in Fig. 5.

NHC’s cone of uncertainty to shrink in 2025

Over the past 20 years, one- to three-day track forecast errors have been reduced by about 75%, and four-day and five-day track forecast errors have fallen by 60%. Those numbers amount to an extraordinary accomplishment, one undoubtedly leading to huge savings in lives, damage, and stress. The improvement in track forecast accuracy has slowed down in recent years, however, suggesting that forecasts may be nearing their limit in accuracy because of the chaotic nature of the atmosphere.

The lower errors in official NHC track forecasts in 2024 mean that the forecast “cones” in 2025 will be slightly smaller than before (up to 6% in the Atlantic). The cone width at any forecast time is based on average error over the prior five years – so whether a storm is well-behaved or tough to predict, the cone width will be the same for all storms in a given year, as it reflects the average historical error rather than the uncertainty specific to a given hurricane. Also, the widths are calibrated so that on average they will capture about two-thirds of hurricane positions, meaning that the observed track positions can be expected to stray outside the cone about a third of the time. (The two-thirds value is intended to strike a balance between over- and under-warning.) And of course, impacts often extend well away from where the center tracks.

Hurricane Beryl from 2024 is a good example of why the cone should be taken as guidance rather than gospel. In the 48 hours before some of Beryl’s strongest winds smashed into central Houston as the center passed just to the west on the morning of July 8, the city core was consistently near the right-hand edge of the cone and briefly outside the right edge.

New on the scene: the HAFS model

For the 2024 season, NHC brought two variants of the new Hurricane Analysis and Forecast System (HAFS) model into the fold of its model guidance. HAFS, which became fully operational on June 27, 2024, is now the preferred option within the National Weather Service for high-resolution track and intensity forecasts, similar to the guidance long provided by HMON and HWRF. (These two models are still being run this season, but in “legacy” mode, so the underlying code will no longer be updated.) Three years of testing (2020-2022) showed improvements of up to 10% in both track and intensity for HAFS versus HWRF.

HAFS is the hurricane-oriented element of the NWS Unified Modeling System, which uses a common dynamical core that’s designed to help streamline the agency’s key modeling efforts. Also part of this unified system is the current GFS model, which will provide input to the higher-resolution HAFS.

Two versions of HAFS are being run, both out to 126 hours and with maximum resolutions of 2 km in and around tropical cyclones:

• HAFS-A (for all global oceanic basins)
• HAFS-B (only for those basins monitored by NHC, including the North Atlantic and the Eastern and Central Pacific)

Sources of free model data

What are ensemble models?

Ensemble model runs are available for most of the top global models. An ensemble model is created by taking the forecast from the high-resolution version of a model like the GFS or European, then running multiple versions of the model with slightly different initial conditions to generate an ensemble of potential forecasts that suggest uncertainties that may exist. These ensemble members are run at a lower resolution to save computer time. The European model has 51 ensemble members, and the GFS has 31. The 0Z GFS run (called GEFS) goes out to Day 35 (note: there is approximately a 24-hour delay for Days 17-35 to be recorded). Note that Days 17-35 ensemble forecasts should be taken with a large grain of salt for now but may still be useful for tracking long-term or seasonal shifts.

Ensembles are especially useful for setups such as weak steering flow, where the varied starting conditions across a model ensemble may shed light on important features that the observing grid hasn’t yet captured directly. When the spread in a model ensemble decreases as a storm evolves, it’s a good sign that the forecast from that operational model is becoming more reliable. Keep in mind that one model’s ensemble tracks can sometimes be in tight agreement while another model’s ensemble is in tight agreement on a completely different solution. In such a case, it’s often the different physics within each model that is driving the difference, which makes it especially important to watch how the consensus model output evolves (the average forecast from three or more separate models averaged together, like the GFS, European, and UKMET models).

Several of the model-output sites above include the ECMWF’s Artificial Intelligence Forecasting System (AIFS), the first major AI-based global forecast model, which became operational in February 2025.

“The AIFS outperforms state-of-the-art physics-based models for many measures, including tropical cyclone tracks, with gains of up to 20%,” the ECMWF said in a news release on the AIFS debut. A 50-member AIFS ensemble is already in the works, and the center pointed out in a 2024 newsletter article that running such AI-based ensembles would be much less costly than running traditional numerical model ensembles.

Tropical cyclone genesis forecasts

NHC has long issued a Tropical Weather Outlook four times per day, offering two-day and five-day forecasts of tropical cyclone genesis. Since 2021, these outlooks have been issued starting on May 15 rather than on the official season-start date of June 1, and in 2024 the five-day forecasts were expanded to cover seven days. For the Atlantic in 2024, these forecasts were pretty reliable for two-day genesis forecasts of 0-60%. For example, when NHC gave a 60% chance a tropical cyclone would form within two days, one actually did form about 56% of the time. However, NHC’s genesis forecasts were too conservative at the upper end of the distribution. All of the Atlantic storms to which NHC gave a 70% and 90% chance of development did, in fact, develop in 2024; and 89% of the storms that were given an 80% chance developed. For seven-day genesis forecasts, NHC was also too conservative with their forecasts when they gave odds of 40% or higher for a genesis event. For example, when NHC gave a 40% chance a tropical cyclone would form within seven days, one actually did form 62% of the time.

A 2016 study by a group of scientists led by Florida State’s Daniel Halperin, though now nine years old, is worth noting: It found that four models can make decent forecasts out to five days in advance of the genesis of new tropical cyclones in the Atlantic. The model with the highest success ratio (rewarding correct genesis forecasts combined with the fewest false alarms) was the European, followed by the UKMET, GFS, and Canadian models.

The scientists authoring that study found that skill declined markedly for forecasts beyond two days into the future, and skill was lowest for small tropical cyclones. The European model had the lowest probability of correctly making a genesis forecast – near 20% – but had the fewest false alarms. The GFS correctly made genesis forecasts 20-25% of the time but had more false alarms. The Canadian model had the best chance of making a correct genesis forecast but also had the highest number of false alarms. The take-home message: The Canadian model’s predicting genesis suggests something may be afoot, but don’t bet on it until the European model comes on board. In general, when two or more models make the same genesis forecast, the odds of the event actually occurring increase considerably, the study authors found.

Sources of tropical cyclone genesis forecasts