Tuesday, August 14, 2007

Vicarious Weather

[Post updated with more recent storm track graphic and proper credit to Weather Underground.]

I am a weather geek living in a place that has almost no weather to speak of. Luckily, I have friends in places that actually have weather. A friend sent me a link to the hurricane track below from Weather Underground's Tropical Weather site.

She asked if I thought it would make landfall or go out to sea. Hmm, I need no excuse to go pore over weather data and make proclamations that I am not qualified to make.*

Off I went to look at the jet stream plots from the California Regional Weather Server. Unfortunately, their plots stop short of the tropical region. I could make out at the edge that winds aloft were rather weak and moving toward the east. (See how the arrows are rather puny and point to the left? The longer the arrow, the stronger the winds aloft.)

Then I took a look at the surface winds. In this type of plot, the "wind barbs" have lines showing the directions that the winds came from. That means the winds are coming from the east near the surface.
That is favorable for hurricane formation. If the winds aloft and at the surface moved in different directions, a condition known as wind shear, the storm can be ripped apart. With little or no wind shear, the storm could really grow.

Warm water below also fuel monster storms. So I looked at the sea surface temperature, SST. That looks pretty warm to me. To put it in perspective, it helps to look at the SST anomaly, or the deviation from the climatological mean. Any way you look at it, the water is warm and conducive to formation of a large storm.
We have an office pool about hurricane storm tracks. As the only non-classically trained meteorologist, I use a naive method. I bet the storm will follow the warm water. So right away, you can see that two things can happen. If the storm blows into the hot water north of Puerto Rico, it will blow out to sea instead of hitting the mainland US. If the storm continues to the east, it can linger in the Caribbean over the very warm water and hit the mainland.

What do the real meteorologists do? They take into account momentum and Coriolis forces. Some pore over the surface pressure maps. Others look at the water vapor ahead of the storm. (Water vapor holds a great deal of latent heat energy which also fuels storms.) Better yet, we can feed all the information into a weather model and crank up the supercomputers. The National Hurricane Center does this, but
The National Hurricane Center (NHC) does not generate a graphic of the models it uses to produce its forecasts. We do this because our past experience indicates such plots have confused users and detracted from our final message, which is producing official tropical cyclone forecasts and advisories. Some users have also become too reliant in the individual forecast scenarios presented by the many model forecast lines, some of which have little or no chance of being correct. This is not the message the NHC wants to send. (Please read our Mission and Vision).
It is ok because the Navy is willing to show you their model hurricane track.

Actually, NHC runs an ensemble of many models. Some are more realistic than others so they don't plot them all out, lest they mislead the public (see above). The Environmental Modeling Center is willing to share their ensemble plots. MIT's hurricane group collected hurricane models from multiple weather forecast centers and plotted them all together.
Why am I showing you so many spaghetti plots? Don't take too much stock in a single forecast. In the words of a newspaper headline on my bulletin board, "Weather Tip: Get many forecasts; 1 may be right"

On a more serious note, the ensemble of forecasts help us understand the uncertainty of predicted hurricane tracks. NHC has an (outdated, but still a gem) web page showing model performance against actual storm tracks for several models and for an entire hurricane season. (nm = nautical miles) See how the models vary in accuracy. The geospatial uncertainty for all models increase with forecast time. In other words, the further in the future, the more difficult it is to accurately predict a hurricane track.

Table 3: Average Errors (nm) of the Early Track Models for 1996-97 Atlantic Tropical Cyclones

Forecast Interval (hr)
Model 12 24 36 48 72
No. Cases346310279255207

And what is the point of all of this?
If you live in hurricane country, you need to prepare well in advance, when the uncertainty is still high. Remember the horrible traffic as people tried to evacuate ahead of Katrina?

If, like the couple in the Dream Home Diaries Blog, your home is on an island, then you need to evacuate even further in advance because of the possibility that bridges will be closed due to high winds or gridlock. Thus, many people whose homes have only a moderate to low probability of being directly hit by the storm should evacuate anyway. If they wait too long, they will have missed their opportunity. Now, where are we going to put all those evacuees?

If you want to think about something even scarier, look at the vast area that could be under water in the event of a storm surge. Suppose they had an 8 meter storm surge. How far from their dream home would they need to evacuate? The flood map generator is just way too much fun for a professional worrier like me. ;-)

* Disclaimer * I am not academically trained as a meteorologist. But the American Meteorological Society let me join because I
hold a baccalaureate or higher degree from an accredited institution of higher learning in some other science or a related field and be currently engaged in a professional activity in which his or her knowledge is applied to the advancement or application of the atmospheric or related sciences


  1. Anonymous08:59

    Don't forget about Weatherunderground.They have good shots of the models and Dr McMasters is pretty low key in his forecasts.The forums on the other hand can get crazy with uninformed opinions.

  2. You are so right. I totally forgot to give them credit for the storm track graphic at the top. I will fix that.