Hurricane tracking reliability.

August 17, 2007 2:14 PM

Has anyone tracked the historical accuracy of different types of hurricane trajectory prediction models?

Each hurricane season I'm fascinated by storm plots like this. I understand that the different colors and their corresponding stock-tickerish names stand for the different sources of hurrican path predictions, but does anyone track the outcomes of hurricane paths to see if any one of these models is more often right than others?

posted by The Straightener to technology (7 comments total)

The National Weather Service Southern Region Headquarters has an overview of the various storm track guidance models models, including graphs showing the relative skill of the statistical, numerical, and intensity guidance models. That page links to a page at the NWS/NCEP Environmental Modeling Center that contains maps of the spatial variation of the forecast track and intensity errors for Atlantic Storms for the year 1995-1999.

I found these by following the link in your sample storm plot image.
posted by RichardP at 2:54 PM on August 17, 2007

On top of that, note that some of the models have had 'upgrades' for this year .. i.e. the GDFL, which in the last runs predicted a direct hit on NOLA for Dean in 132 hous.
posted by SpecialK at 3:04 PM on August 17, 2007

You might also find interesting Dr. Jeff Masters' recent post about Hurricane Dean over at the Weather Underground. It touches on the subject of the historical accuracy of computer models as it relates to Hurricane Dean (which is currently approaching the Caribbean).
posted by RichardP at 5:02 PM on August 17, 2007

Great Weather Underground post. Would the why behind the better performance of certain models be too complicated for a lay reader to understand? It always seems that there's this broad area of concensus and then a couple outlier predictions that plot a way different course than the others. I'm curious to know why some models produce such variations, and why others produce such concensus, but I doubt I could understand the models to begin with. What kind of real life conditions would have to be satisfied say for Dean to suddenly whipsaw north towards NOLA, like the CLP5 prediction?
posted by The Straightener at 6:48 PM on August 17, 2007

What kind of real life conditions would have to be satisfied say for Dean to suddenly whipsaw north towards NOLA, like the CLP5 prediction?

Deeper prevailing westerlies (i.e. westerly upper level winds that dip further south) would do that. Chaos theory is at the fabric of all these forecasts. With the dynamic models, you are looking at the storm track in the forecast run of a simulated atmosphere, so if various parameters are off (physics, observed weather, etc) then the outcome changes from reality with time.

I work in the field and have written some books, unfortunately I've had a few beers so I may have be very clear right now.
posted by rolypolyman at 7:16 PM on August 17, 2007

"may not be very clear right now"
posted by rolypolyman at 7:16 PM on August 17, 2007

Having lived in Florida for most of my ife I've ridden through and run from many storms. I can say for sure that I have noticed a great leap in accuracy of storm path projections over the last ten years or so. While weather science has a pretty lousy record when it comes to predicting how many storms will occur in a year, once a storm forms, they usually do an accurate job of predicting where it will go.
posted by wsg at 1:49 AM on August 18, 2007