Could last seasons hurricane model reign supreme this year

Upgrades set to increase accuracy

Comparison of models and the National Hurricane Center forecast track errors for landfalling Hurricanes. Lower bars mean better forecasts.
Comparison of models and the National Hurricane Center forecast track errors for landfalling Hurricanes. Lower bars mean better forecasts.

JACKSONVILLE, Fla. – Last seasons best preforming hurricane model may not be this year's top forecasting pick once an array of upgrades are added.

After the three unprecedented Category 4 hurricanes made landfall in the country it should be no surprise the 2017 hurricane season was the United States costliest year on record with a $250 billion price tag.

The attack of storms certainly tested forecast model skill yet overall they were quite good and continued to improve in all forecast periods compared with past seasons.

It was the best year on record for track forecasting by the National Hurricane Center thanks to improvements in forecasting tools.

Which model is the best? It typically comes down to the U.S. National Weather Service’s Global Forecast System (GFS) and the European Centre for Medium-Range Weather Forecasts (ECMWF).

Most of the time the correct model was Europe's. It reigned supreme with day 3 and day 5 forecasts compared with the previous year.

Over the past 5 years typical track errors are 42 miles a day away and increase about 40 miles each day in the forecast. So by the fifth day the storm could be around 210 miles from the center track line.   

No model is perfect, and their performance varies each season and from storm to storm.  

Last year the GFS performed better than previous years but still had problems forecasting the birth of storms. It did poorly forecasting Irma but did better forecasting hurricane Maria compared to the ECMWF.

The GFS has consistently lagged behind other modeling centers in forecasting precipitation across the U.S according to Vijay Tallapragada, chief forecaster at NCEP's Global Climate and Weather Modeling Branch. 

He says it struggles to capture mid-latitude systems which often are the features that make storms turn north like Irma. 


The ECMWF was the best-performing model most of the time in 2017, but that does not mean it will always be the best for every storm.

The FV3GFS is an experimental model in testing set to eventually replace how the GFS processes data by 2020. 

Results are promising with this new model coming close to the GFS with Hurricane Harvey during the first five days but it predicted Harvey's rapid intensification just before landfall while the GFS took longer to catch up to Harvey's impressive strengthening in the Gulf.

With Irma, the FV3GFS forecasted the center of Irma west of the GFS and it matched what actually happened better. For the Atlantic, the FV3GFS was generally more skillful than GFS, but worse than the operational HWRF.

Upgrades to the worlds best model may keep the ECMWF at the top.

Scientists want to correct how it overestimates storm intensity beyond 3 days and they have found some improvements when ocean conditions are incorporated into the model.

This configuration should be in operations this summer in addition to adding lightning flashes and high resolution wind data collected from scatterometers. 

Results from these models are stretching the predictions. Beginning in 2018 the NHC's will extend the 36 hour wind forecasts to 48 hours.


About the Author: