The "spaghetti plot" has become a very common graphic that we show during the tropical season. It shows a bunch of lines on a map that illustrate a computer forecast.
It looks like spaghetti that has been thrown against a wall.
Each line represents the work of a specific computer model. It wasn't many years ago that meteorologists never showed this graphic on TV, even though we had access to it. The spaghetti plot became popular during and after the 2004 hurricane season.
These models process a tremendous amount of data from all over the atmosphere. With that in mind, it takes very large computers with huge processing power to decipher the data and crunch the calculations. Data for the models are processed either twice or four times a day, depending on the model. So there is no time for rest with these computers. They almost always are crunching numbers so they can crank out the next model output. It's an impressive process that is constantly taking place behind the scenes. This is why the jump in computer technology has helped improve forecasting in the past 20 years.
The more data we can get from the atmosphere and give to the computer, the better output we tend to get. Since this process takes very large, expensive computers, these forecast models are typically processed by the government. The National Centers for Environmental Prediction near Washington, D.C., is where the bulk of the forecast models are processed.
Examples of models that we use, but are not processed in this country, are the United Kingdom Met Office model (UKMET), the European Center for Medium-Range Weather Forecasting (ECMWF) and the Canadian (GEM) model. There are also examples of universities processing a model, but it's rare due to the cost and resources that it takes. The most successful tropical model produced by a university is the Florida State Superensemble (FSSE).
The most accurate models over the past few years have been what we call "consensus" models. This is when you take an average of your best models to get a consensus forecast. The FSSE is an example of a consensus model. The models that are typically used for the consensus are the Global Forecast System (GFS), the NWS/Geophysical Fluid Dynamics Laboratory model (GFDL) and the ECMWF. There are a few other models that are used for the consensus, but these are the main ones that have provided the most accurate results.
Over the past 15 years there has been a slow shift in the way forecasts are delivered on TV. We tend to show more data that support and explain our forecasts.
The spaghetti plot is one of those graphics that became popular very quickly, mainly because it is a simple concept. It allows for a lot of computer data on one simple map so you can see the complex forecast that we're looking at behind the scenes. As with anything, though, it has its drawbacks. We quickly noticed that viewers had a lot of questions about what the spaghetti plot meant. People wanted to know what each line represented and which line was more accurate. So what we've had to do over the years is explain the graphic and also explain what the graphic is showing in each particular situation.
The first and most basic thing to look at is whether the lines are bunched together or whether they are spread far apart. When they are more tightly bunched together it usually means a higher level of confidence in the forecast. When the lines are spread apart that usually means there is a lot of uncertainty about where the storm will go. Some computer models are better than others at predicting paths. We try to convey this message when we show this graphic on TV. Many of the models were developed by different organizations and are given different parameters. Therefore, it's okay for the models to differ. You just have to understand why they differ and then figure out which model has a better understanding of the current atmospheric setup. This is the hard part. It's up to experienced meteorologists to explain how we think the storm will respond to various things in the atmosphere.
You might think there is a dominant model that we lean on, but for tropical storm tracking we've noticed over the years that they all have their good and bad days. Some models will even do better for an entire season and then do poorly the next year. This is why spaghetti models need to be used with caution. There is no such thing as a 100 percent accurate forecast model. However, we have found that there are a few models that are better than the rest, so we tend to lean heavily on their results. The good news is that the models continue to be worked on and tweaked to make them better. Believe it or not, before the 1980s the forecast models didn't have much skill in predicting the path of tropical storms.
Before the 1980s the main model was the CLIPER. It really isn't even a forecast model. It uses climatology to plot a path. For example, it looks at all past storms and plots a likely path based on what those historic storms did. We do not show this on TV because it doesn't serve a purpose for our viewers. Unfortunately, they are shown on websites and some other sources, and they can confuse the public. We like to keep it simple and avoid confusion.
The last thing to remember is that the spaghetti plot does not show intensity predictions. When watching hurricanes, remember that unexpected changes in intensity are always possible.