Ultra-computer should detect killer storm from a creampuff

[HurricaneJoe22]

Now meteorologists might be able tell us about an incoming storm's intensity.

By Ken Kaye
South Florida Sun-Sentinel
Posted March 31 2007

It can portray a hurricane in stunning detail. It's powered by a supercomputer that can perform 14 trillion calculations a second. And starting in June, it should help tropical meteorologists project whether a storm will arrive as a killer or a creampuff.

The sophisticated new forecast model could be the Holy Grail that forecasters have long sought to sharply improve their hurricane intensity predictions and give emergency managers and residents alike more time to prepare accordingly.

"This is the first time a hurricane model will have its own analysis of the center of the hurricane's structure," said Naomi Surgi of the National Weather Service, who spearheaded the model's development. "This is really pushing the frontiers of science."

Although forecasters at the National Hurricane Center in Miami-Dade County have achieved record accuracy in projecting the path of a storm, they still struggle to gauge its power. South Floridians received an unpleasant reminder of this meteorological soft spot in August when Tropical Storm Ernesto was predicted to plow ashore as a hurricane. After thousands of residents scrambled to put up shutters, the system arrived as nothing more than a squally nuisance.

"Often times, the intensity forecasts can be poor, and that's not because forecasters aren't doing the best they can," said John Gamache, field program manager of the National Oceanic and Atmospheric Administration's Hurricane Research Division in Miami. "It's just that, to an extent, our understanding of the processes within a storm are kind of limited."

Enter the new model, officially called the Hurricane Weather and Research Forecast system, or "H-werf "for short. It will zero in on the ocean's interaction with a storm as never before and produce an elaborate three-dimensional picture of a hurricane's core, where the most vicious winds lurk.

Hopes are that it will outperform older storm intensity models, said Surgi, hurricane modeling program leader at the weather service's Environmental Modeling Center in Camp Springs, Md.

To function, the model needs to gorge on data: sea surface temperatures, wind conditions and barometric pressures, gleaned by hurricane hunter planes, satellites, buoys and other sensors.

After this banquet of information is ingested, it is sifted, studied and manipulated by NOAA's supercomputer in Gaithersburg, Md., which is capable of absorbing 240 million global weather observations daily and doing 14 trillion data operations in a single second.

To give some perspective, if a human were to undertake 14 trillion calculations at the pace of one per second, the job would be complete in 443,632 years.

The H-werf is sure to be a hot topic at the National Hurricane Conference, which starts Monday in New Orleans and helps meteorologists, emergency managers, government officials and others gear up for the coming storm season.

The model already has proven to be a powerful forecasting tool. While being tested, it accurately predicted that Hurricane Katrina would spin into a Category 5 monster as it marched across the Gulf of Mexico toward New Orleans in August 2005. It did "a really good job" tracking other major hurricanes that summer, including Dennis, Rita and Wilma, Surgi said.

Nevertheless, it will take a few seasons for forecasters to gain confidence in the new model, she said. While it should produce positive results this year, it will require annual adjustments and upgrades before it becomes a truly reliable prophet of a hurricane's development.

"We will see an accelerated rate of improvement over the next five to 10 years," she said.

NOAA, the parent agency of both the weather service and the hurricane center, has made intensity forecasting a top priority because the Atlantic basin is entrenched in a period of heightened activity. That was clearly seen in 2005, the most destructive and active season since records began in 1851.

The fear is that more hyperactive seasons may lie ahead, and that off-target intensity predictions could lead to disaster. Yet, try as they might, forecasters have so far been unable to grasp the complex mechanisms that influence storm strength. As a result, the hurricane center last year erred by an average of 21 mph in predicting the sustained winds of storms three days in advance. That wasn't much better than the average error a decade and a half ago.

The most dreaded scenario for forecasters and emergency managers alike is for a system to rapidly intensify just prior to landfall, as Charley did in August 2004. In five hours, the system's sustained winds surged from 110 to 150 mph, from Category 2 to Category 4, before battering Punta Gorda, Port Charlotte and other Southwest Florida towns.

The reason Charley bulked up so quickly: It crossed over a patch of unusually warm water near the coast and feasted greedily on the thermal energy. Forecasters hadn't expected that. And, in general, they have difficulty foreseeing when a storm will suddenly spin up, said Nick Shay, of the University of Miami Rosenstiel School of Marine and Atmospheric Science.

"Rapid intensification happens in only 10 to 15 percent of the systems," he said. "But those are the ones that give forecasters fits."

In its proposed $3.8 billion budget for next year, NOAA has requested $2 million specifically for hurricane intensity research. It has amassed an arsenal of technological tools to sample the atmosphere around hurricanes, including a high-flying Gulfstream jet equipped with Doppler radar. And it has conducted in-depth post-mortem studies of storms. For instance, during the 2006 storm season, it examined the entire lifecycle of some tropical systems to determine what makes them fluctuate in strength.

But the real optimism for progress lies with the new model, officials said. Although it is "one piece of the puzzle" in the hurricane center's ongoing campaign to improve forecast accuracy, it is a key one, said hurricane specialist James Franklin.

"That is our hope for the future," he said.

[brunota2003]

Link please? Interesting read...but I would like to see a link

[x-y-no]

Does anyone here know if the H-WRF runs will be available online?

[wxman57]

The HWRF and GFDL will be run together in 2007. I'm not sure, but the HWRF model output may be published along with GFDL. That link I posted last week to the Interdepartmental Hurricane Conference has a paper on the HWRF-GFDL comparison. They are fairly close, but the GFDL still leads in some areas. Lots of tweaking to do to the HWRF. As for the potential for better intensity forecasts, that's there. But garbage in/garbage out. We need initialization data, and that's hard to come by out in the ocean.

Look in session 5:
http://www.ofcm.noaa.gov/ihc07/linking_file_ihc07.htm

[Derek Ortt]

HWRF also has Ernesto as a cat 3 in the GOM

[wjs3]

Where does HWRF get track from? Another model I assume, but which model(s)?

Thanks

WJS3

[wxman57]

Where does HWRF get track from? Another model I assume, but which model(s)?

Thanks

WJS3

HWRF uses at least part of the GFS for initialization. You can read all about it at the link I posted above:
http://www.ofcm.noaa.gov/ihc07/linking_file_ihc07.htm

[Derek Ortt]

I think it uses the GFDL, which uses the GFS

As many here know, I would never touch anything within 100 miles of the GFS

[wjs3]

Thanks to both of you.

I may be oversimplifying this, as I am nowhere near the expert you two are, but if an intensity model has a cyclone in the wrong place, well, then it has little chance of getting the intensity right, becasue the environment can be so different. I think we've all seen this with models like SHIPS. Put the storm in the wrong place and the shear values (for instance) assumed are incorrect, and the whole thing is out the window.

From what I gather, no different with the HWRF, right? Intensity and track are hopelessly linked.

I think I need to attend this conference next year!
WJS3

[caneflyer]

I think it uses the GFDL, which uses the GFS

While the HWRF shares much of its physics with the GFDL, it does not use any GFDL output to run. The GFDL could be turned off today (indeed, that was the original plan) and the HWRF would still run.

[wxman57]

I think it uses the GFDL, which uses the GFS

While the HWRF shares much of its physics with the GFDL, it does not use any GFDL output to run. The GFDL could be turned off today (indeed, that was the original plan) and the HWRF would still run.

Correct, the HWRF doesn't use anything from the GFDL, but like the GFDL, it uses GFS data for initialization. Early tests of the HWRF show it at least equal to the GFDL out to 72 hours but the HWRF lags a bit behind the GFDL beyond 72 hours. Further tweaks/modifications of the HWRF should allow it to surpass the GFDL's performance at all time periods.

[Derek Ortt]

I believe, though I may be mistaken, it uses the GFDL initial vortex (only model that has an initial vortex of comparable intensity).

They could go through the GFDL relocation procedure, rendering the GFDL totally obsolete

[wxman57]

Here is a PPT presentation that answers a lot of questions. You'll need MS PowerPoint or the free viewer to open the link:

http://www.ofcm.gov/ihc07/Presentations ... ya_rev.ppt

[jasons2k]

I think it uses the GFDL, which uses the GFS

As many here know, I would never touch anything within 100 miles of the GFS

It's not so bad:

http://www.ofcm.gov/ihc07/Presentations ... 5,17,Slide 17