For those puzzling over the various hurricane computer forecast models to figure out which one to believe, the best answer is: Don’t believe any of them. Put your trust in the National Hurricane Center, or NHC, forecast.
While an individual model may outperform the official NHC forecast in some situations – for example, the European model’s track forecasts for Hurricane Isaias did better than the NHC’s forecast at all forecast time periods – the 2019 National Hurricane Center Forecast Verification Report documented that overall, it is very difficult for any one model to consistently beat the NHC forecasts for track and for intensity.
During the 2019 Atlantic hurricane season, NHC track forecasts had accuracies near the five-year average, with five-day track forecasts setting a new record for accuracy. Over the past 30 years, one- to three-day track forecast errors have decreased by 70 – 75%; over the past 15 years, four-day and five-day track forecast errors have decreased by 60%. That amounts to an extraordinary accomplishment, one undoubtedly leading to huge savings in lives, damage, and emotional angst.
The 2019 NHC Atlantic track forecasts tended to have a south to southwest bias of 31 – 95 miles for three- to five-day forecasts (i.e., the official forecast tended to fall to the south or southwest of the verifying position). The NHC track forecasts for Humberto and Lorenzo were particularly excellent, but the overall track forecast errors would have been much better had it not been for one oddball storm that caused major frustrations for forecasters – Tropical Storm Sebastien. The average NHC three-day track forecast error for Sebastien was 390 miles – much worse than the typical NHC three-day track error of 118 miles used in the cone of uncertainty.
Best track model in 2019: the European
As usual, in 2019 the official NHC track forecasts for Atlantic storms were tough to beat, though the European Center (ECMWF) model did outperform the official NHC forecast for one-day and two-day forecasts. The UKMET model was the second-best model in 2019, closely followed by the COAMPS-TC model. The performance of the Navy model (NAVGEM) and NOAA’s GFS, HWRF, and HMON models lagged behind. Here is a list of some of the main hurricane forecast models used by NHC:
Euro: 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
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: COAMPS-TC regional model, initialized using GFS data
Navy: Navy Global Environmental Model (NAVGEM).
If one averages together the track forecasts from three or more of these models, the NHC official forecast will rarely depart much from it. These seven models are used as input to various “consensus” models, such as “TVCN”, often referenced in NHC discussions for a storm. Improved consensus modeling techniques are one major reason NHC track forecasts have seen such a large improvement in the past 20 years.
NHC intensity forecasts in 2019: not much improvement
Official NHC intensity forecast errors in the Atlantic in 2019 were similar to the five-year average for 12-hour, one-day, two-day, and three-day forecasts, but above the mean for four-day forecasts, and well above the mean for five-day forecasts. Mean intensity forecast errors in 2019 were 5 mph at 24 hours, and increased to 30 mph for five-day forecasts. The official forecasts were biased too low at all forecast times, with the most underprediction coming for four-day and five-day forecasts. Part of the reason for the lack of better long-range intensity forecasts in 2019 was that NHC did a poor job forecasting the rapid intensification of Hurricane Dorian, which topped out as a category 5 storm with 185 mph winds over the Bahamas.
Best intensity model in 2019: the COAMPS-TC
In 2019, the official NHC intensity forecast outperformed the five top intensity models at all forecast times, save for five-day forecasts, for which the COAMPS-TC model from the Naval Research Laboratory did best. The five main intensity models are the regional/dynamical HWRF, HMON, and COAMPS-TC models (which subdivide the atmosphere into a 3-D grid around the storm and solve the atmospheric equations of fluid flow at each point on the grid), and the statistics-based LGEM and DSHP models (DSHP is the SHIPS model with inland decay of a storm factored in).
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 made poor intensity forecasts in 2019. In 2019, the HMON model was the best-performing intensity model for one-day to 1.5-day forecasts, with the other four main intensity models close behind. The COAMPS-TC model was the best one at 2-day, 3-day, 4-day, and 5-day forecasts. The HMON model performed relatively poorly at the longer-range forecast times. All of the models had a low bias in predicting intensity.
Sources of track model data
– Tropical Tidbits, which has become the best source of free model data on the web;
– ECMWF forecasts from the ECMWF website;
– FSU’s model page (CMC, ECMWF, GFS, GFS HWRF, HMON, and NAVGEM models);
– NOAA’s HWRF model page;
– NOAA’s HWRF and HMON model data page;
– The Navy’s COAMPS-TC model data page;
– Experimental HFIP models (note that the HAFS-A model outperformed the GFS, HWRF, and HMON models on 2019 hurricane track forecasts, but fell behind the HWRF and HMON on intensity forecasts);
– UKMET text forecast;
Additional information about the guidance models used at the NHC can be found at NHC (updated in 2019).
Model improvements for 2020
Many of the top models used to forecast hurricanes received significant upgrades for the 2020 season. These upgrades, unfortunately, may be largely offset by the reductions in model accuracy as a result of less atmospheric sounding data being taken by commercial aircraft because of the Covid-19 pandemic. (An open-access paper published in July in Geophysical Research Letters analyzed the data gap and its impact on global model performance this past spring.)
The latest upgraded version of the European model came online on June 30.
The HWRF model, upgraded at the end of July, will benefit from getting more data from the multiple satellites, including the new radio occultation sounding data from the COSMIC-2 satellites launched last year. The model will also be ingesting Doppler radar data from land-based radars for the first time. In an interview with spacenews.com, NOAA’s Jim Yoe said that he expects “improvements on the order of 5 percent for track and perhaps as much as 8 percent for the intensity of hurricanes in the North Atlantic and the Pacific basins.” The HWRF model uses three nested grids zoomed in on hurricanes with resolutions of 13.5 km, 4.5 km, and 1.5 km. Graphics for the legacy HWRF model are available at tropicaltidbits.com, along with the new version, labeled “HWRF-PARA”.
The most recent major upgrade to the GFS model occurred in June 2019. The next significant upgrade, the GFSv16, is scheduled for the winter of 2021.
Output from the new GFSv16 can be reviewed at tropicaltidbits.com, labeled “GFS-Para” (for parallel GFS model).
Ensemble models 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 21. A new version of the GFS ensemble model (called GEFSv12) is in the experimental phase, and is the first update to GEFS in five years. The new GEFS increases the resolution of its members to ~25 km, expands to 31 members, and the 0Z run 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. Plots of the experimental GEFS are available from Tropical Tidbits as GEFS-Para.
Forecasts for Hurricane Isaias
Using data compiled from SUNY Albany’s Brian Tang, I plotted up the model performance for 2020’s most significant Atlantic tropical cyclone thus far, Hurricane Isaias.
For track, the European model continued its stellar performance, outperforming every model and the official NHC forecast at every time period. The official NHC forecast performed well, as did the UKMET and Navy NAVGEM models.
For intensity, the official NHC forecast and various models did not show a lot of separation, though the LGEM model performed noticeably worse than the other top models. Note that the best track model, the European, is not considered a top intensity model, and performed poorly for Isaias.
Tropical cyclone genesis forecasts
NHC issues a Tropical Weather Outlook four times per day, offering two-day and five-day forecasts of tropical cyclone genesis. For the Atlantic in 2019, these forecasts were pretty reliable for five-day genesis forecasts of 40 – 70%: So, when NHC gave a 40% chance a tropical cyclone would form within five days, one did actually form about 40% of the time.
However, NHC’s genesis forecasts were too conservative at the lower and upper ends of the distribution. One hundred percent of the storms that it gave an 80% or 90% chance of development did in fact develop, and 20% of the systems that NHC gave a 0% chance of development for did in fact develop. Interestingly, the poorest NHC forecasts occurred when it gave a 30% chance of development – fully 80% of these systems ended up developing.
A 2016 study by a group of scientists led by Florida State’s Daniel Halperin found 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 fewest false alarms) was the European (ECMWF), followed by the UKMET, the GFS, and Canadian models.
The 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 greatest chance of making a correct genesis forecast, but also had the highest number of false alarms. The take-home message: if the Canadian model is predicting genesis, it is suggestive that something may be afoot, but don’t bet on tropical cyclone genesis 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 found.
Sources of tropical cyclone genesis forecasts
– NHC 5-day Tropical Weather Outlook;
– NOAA/CIRA (two-day forecasts);
– Florida State University Experimental Tropical Cyclone Genesis Guidance page; and
– SUNY Albany 10-day Experimental Genesis Probabilities (Alan Brammer).
Bob Henson assisted with and provided some of the links for this post.
Website visitors can comment on “Eye on the Storm” posts (see below). Please read our Comments Policy prior to posting comments. (See all EOTS posts here. Sign up to receive notices of new postings here.)