Listen to any TV weather report concerning Atlantic Ocean area tropical storms and hurricanes and you’ll hear the term, “spaghetti plot” mentioned. Today (Sun., Aug. 24, 2014), spaghetti plots (Fig. 01) surfaced as soon as Cristobal was named.
Spaghetti plots are nothing more than a single image that contains storm tracks as forecast by an array of numerical prediction models. Each of these, “models” contains differing mathematical representations of atmospheric variables (e.g., temperature, pressure, wind) and the physics and thermodynamics that govern them. The models handle the lower atmosphere (a.k.a., the troposphere) across different geographical areas employing different ways of describing the atmosphere’s initial condition. Then, the models are integrated over different time steps and over different horizontal and vertical distances.
Meteorologists look at the output of these models in many ways, including spaghetti plots, to try to figure out which model is expected to perform best at predicting a storm’s future movement and intensity.
When most or all model output agrees, meteorologists have a high confidence in their forecast. Fig. 2 and Fig. 3 are examples of “well-behaved” suites of outcomes and situations that offer a high forecast confidence level.
When the storm is weaker and/or steering currents are weak and/or transitioning, projected storm paths (and intensities) can differ significantly. Sandy’s spaghetti plot (Fig. 4), from early in its life cycle, offers an example of diverging storm track forecasts. However, this does not affect the official forecast cone (of uncertainty). This is because the cone is constructed from individual forecast points with error circles placed around the points. The circles are sized to accommodate 67% of the previous five years official forecast errors. Since errors grow with time, the largest uncertainty circle is at the five-day forecast time.
One thing that meteorologists key on when using spaghetti plots is change. Hence, when the models start to show a shift in where the storm is projected to move, this sets the forecaster on a track to update forecasts accordingly. Even though an “average” position within the spaghetti plot may often be the chosen track, meteorologists can adjust a track more to the left or right (or update expected movement speed) based on aircraft reconnaissance, satellite imagery and other data.
Figures 5 and 6 are testimony to how forecasts change based on later data. Fig. 5 shows the forecast for Sandy from Oct. 23, 2012. By Sunday, the storm was expected to be in its declining stages west of Bermuda and moving on a northeast track. Fig. 6 shows the forecast prepared two days later. At that time, Sandy was expected to be in hurricane status by Sunday, at a position several hundred miles to the west of the earlier forecast position. Even more importantly, the hurricane was expected to be moving northward and then northwest, toward the U.S. East Coast. To see how Sandy’s storm track forecasts evolved over time, watch this full animation of Hurricane Sandy’s graphical advisories.
While spaghetti plots may capture one’s attention (after all, they are colorful), non-meteorologists would be better served by following the official National Hurricane Center graphical forecasts and narrative advisories. It’s also a good idea to keep in mind that the longer into the future one looks, the more uncertain the outcome. Many times, a five-day graphical outlook shows a particular landfall, only to be realized by an error of hundreds of miles.
© 2014 H. Michael Mogil
NOTE: Chris Landsea (meteorologist with NOAA) provided further explanation about how the cone of uncertainty is created. That was added into paragraph 5 of the article (with a reference hyperlink).