Which Hurricane Forecast Model Should You Trust?

The National Hurricane Center (NHC) set a new record in 2012 for accuracy of their 1, 2, 3, and 4-day Atlantic tropical cyclone track forecasts, but had almost no skill making intensity forecasts, according to the 2012 National Hurricane Center Forecast Verification Report, issued in March 2013. The new records for track accuracy were set despite the fact that the season’s storms were harder than average to forecast. The average error in a 1-day forecast was 46 miles, and was 79 miles for 2 days, 116 miles for 3 days, 164 miles for 4 days, and 224 miles for 5 days. The official track forecast had a westward bias of 10 - 17 miles for 1 - 3 day forecasts (i.e., the official forecast tended to fall to the west of the verifying position), and was 38 and 75 miles too far to the northeast for the 4- and 5-day forecasts, respectively.


Figure 1. Verification of official NHC hurricane track forecasts for the Atlantic, 1990 - 2012. Over the past 15 - 20 years, 1 - 3 day track forecast errors have been reduced by about 60%. Track forecast error reductions of about 50% have occurred over the past ten years for 4- and 5-day forecasts. Image credit: 2012 National Hurricane Center Forecast Verification Report.

NHC Intensity Forecasts: Little Improvement Since 1990
Official NHC intensity forecasts did better than usual in 2012, and had errors lower than the 5-year average error for 1, 2, 3, 4, and 5-day forecasts. However, 2012's storms were easier to predict than usual, due to due to a lack of rapidly intensifying hurricanes. These rapid intensifiers are typically the source of the largest forecast errors. The skill of official NHC 24-hour intensity forecasts made in 2012 for the Atlantic basin were only about 15% better than a "no-skill" forecast; 2, 3, 4, and 5-day intensity forecasts had no skill.


Figure 2. Verification of official NHC hurricane intensity forecasts for the Atlantic, 1990 - 2012. Intensity forecasts have shown little to no improvement since 1990. Image credit: 2012 National Hurricane Center Forecast Verification Report.

Which Track Model Should You Trust?
As usual, in 2012 the official NHC forecast for Atlantic storms was almost as good as or better than any individual computer models--though NOAA's GFS model did slightly better than the NHC official forecast at 12, 24, and 48-hour periods, and the European model forecast was slightly better at 12-hour forecasts. Despite all the attention given to how the European Center (ECMWF) model outperformed the GFS model for Hurricane Sandy's track at long ranges, the GFS model actually outperformed the European model in 2012 when summing up all track forecasts made for all Atlantic named storms. This occurred, in part, because the European model made a few disastrously bad forecasts for Tropical Storm Debby when it was in the Gulf of Mexico and steering currents were weak. For several runs, the model predicted a Texas landfall, but Debby ended up moving east-northeast to make a Northwest Florida landfall, like the GFS model had predicted. However, the best-performing model averaged over the past three years has been the European Center model, with the GFS model a close second. Wunderground provides a web page with computer model forecasts for many of the best-performing track models used to predict hurricane tracks. The European Center does not permit public display of tropical storm positions from their hurricane tracking module of their model, so we are unable to put ECMWF forecasts on this page. Here are some of the better models NHC regularly looks at:

ECMWF: The European Center's global forecast model
GFS: NOAA's global forecast model
NOGAPS: The Navy's global forecast model (now defunct, replaced by the NAVGEM model in 2013)
UKMET: The United Kingdom Met Office's global forecast model
GFDL: The Geophysical Fluid Dynamics Laboratory's hurricane model, initialized using GFS data
HWRF: The intended successor for the the GFDL hurricane model, also initialized using GFS data
CMC: The Canadian GEM model
BAMM: The very old Beta and Advection Model (Medium layer), which is still useful at longer ranges

If one averages together the track forecasts from the first six of these models, the NHC official forecast will rarely depart much from it. As seen in Figure 3, the HWRF and UKMET were well behind the ECMWF and GFS in forecast accuracy in 2012, but were still respectable. The simple BAMM model did well at 3, 4, and 5-day forecasts. The GFDL and CMC models did quite poorly compared to the ECMWF, GFS, UKMET, and HWRF. The Navy's NOGAPS model also did poorly in 2012, and has been retired. Its replacement for 2013 is called the NAVGEM model.


Figure 3. Skill of computer model forecasts of Atlantic named storms in 2012, compared to a "no skill" model called "CLIPER5" that uses just climatology and persistence to make a hurricane track forecast (persistence means that a storm will tend to keep going in the direction it's current going.) OFCL=Official NHC forecast; GFS=Global Forecast System model; GFDL=Geophysical Fluid Dynamic Laboratory model; HWRF=Hurricane Weather Research Forecasting model; UKMET=United Kingdom Met Office model; ECMWF=European Center for Medium Range Weather Forecasting model; TVCA=one of the consensus models that lends together several of the above models; CMC=Canadian Meteorological Center (GEM) model; BAMM=Beta Advection Model (Medium depth.) Image credit: National Hurricane Center 2012 verification report.

Which Intensity Model Should You Trust?
Don't trust any of them. NHC has two main statistical intensity models, LGEM and DSHP (the SHIPS model with inland decay of a storm factored in.) In addition, four dynamical models that are also use to track hurricanes--the GFS, ECMWF, HWRF, and GFDL models--all offer intensity forecasts. With the exception of the GFS model, which had a skill just 5% better than a "no-skill" intensity forecast for predictions going out 36 hours, all of these models had no skill in their intensity forecasts during 2012. The ECMWF and HWRF models were the worst models for intensity forecasts of 3, 4, and 5 days, with a skill of 20% - 60% lower than a "no-skill" forecast. The LGEM model, which was a decent intensity model in 2011, tanked badly in 2012 and had near-zero skill. The only model that was any good in 2012 was the IVCN "consensus" model, which averages together the intensity forecasts of two or more of the intensity models such as LGEM, GFDL, HWRF, and DSHP.

Some Promising Models From the Hurricane Forecast Improvement Project (HFIP)
Last year was the fourth year of a ten-year project, called the Hurricane Forecast Improvement Project (HFIP), aimed at reducing hurricane track and intensity errors by 50%. The new experimental models from HFIP generally performed poorly in 2012. However, the new FIM9 15-km global model was competitive with the ECMWF and GFS models for track, and the new CIRA Statistical Intensity Consensus (SPC3) model for intensity performed better than many of the traditional intensity models.

For those interested in learning more about the hurricane forecast models, NOAA has a 1-hour training video (updated for 2011.) Additional information about the guidance models used at the NHC can be found at NHC and the NOAA/HRD Hurricane FAQ.

Sources of Model Data
You can view 7-day ECMWF and 16-day GFS forecasts on wunderground's wundermap with the model layer turned on.
Longer ten-day ECMWF forecasts are available from the ECMWF web site.
FSU's experimental hurricane forecast page (CMC, ECMWF, GFDL, GFS, HWRF, and NAVGEM models)
NOAA's HFIP model comparison page (GFS, ECMWF, FIM, FIM9, UKMET, and CMC models.)
Experimental HFIP models

Quiet in the Atlantic
There are no tropical cyclone threat areas in the Atlantic to discuss today, and none of the reliable models for tropical cyclone formation is predicting development during the coming seven days. I plan on having a detailed update on Friday to discuss the latest long-range forecasts for the coming peak part of hurricane season.

Jeff Masters