There are dozens of hurricane forecast models all with pros and cons when it comes to predicting where a storm will track or how strong it might become.
Forecasters and storm enthusiasts often refer to a collection of models as a spaghetti plot. When those maps of storm track lines are clustered tightly, it indicates various models are in similar agreement on future storm behavior. When the spread is large, the forecast confidence is low.
Last year, the GFS was the best performing track model, followed by the European.
This is a switch from a long-standing rivalry between the GFS and the ECMWF model, which outperformed the American model last year.
In 2020, the National Hurricane Center issued almost 600 forecasts in its second-busiest year behind 2005, nearly double the normal amount compared to an average year. That was when the ECMWF came out on top as the best individual model in the short term, but in 2021 the GFS had the most skill at the long lead times.
In addition to the two big global models, there are other regional models: HWRF, HMON, COAMPS-TC, and CMC which all did well for forecasts up to three days. The Canadian CMC and Navy’s COAMPS-TC models performed poorly at longer time periods.
Many factors set models apart from each other
Some blanket the entire globe while others are limited to specific regions right around the storm. Because the global models take time to crunch through the mathematical calculations, there are speedier interpolated versions of dynamical models to help forecasters meet a deadline.
The best computer forecasts come from dynamical models which use physical equations to predict the motions in the atmosphere. Dynamical models may treat the atmosphere either as two-dimensional or as having multiple three-dimensional layer components.
Statistical models, in contrast, do not consider the characteristics of the current atmosphere explicitly but instead are based on the persistence that a current storm will behave like past storms in a similar location and time of year.
The CLIPPER5 is one of the statistical models that serve as a baseline to beat. If a forecast proves more accurate than this model, it has skill.
This is how forecasters at the National Hurricane Center grade themselves. In 2021, they had their best two- and three-day forecasts on record in addition to their forecast being as accurate or better than the five-year average.
Part of their success is not relying on one model but using a variety of consensus models which are combinations of results from other models.