Steering models in the right direction
A chat between Steve Chiswell (left) and Ben Domenico at a conference last January resulted in a promising new prototype from UNIDATA. (Photo by Carlye Calvin.)
Sometimes a casual interaction at a conference can shape the course of research for months to come. That’s what happened to software engineer Steve “Chiz” Chiswell (Unidata) at the annual meeting of the American Meteorological Society in San Diego last January.
Unidata is a major partner in LEAD (Linked Environments for Atmospheric Discovery). This NSF-funded project is building an interconnected, distributed system to help universities and other users work more fluidly with observations and forecast models. Seven universities are joining Unidata on the five-year LEAD effort.
After an AMS talk on LEAD by Kelvin Droegemeier (University of Oklahoma), Unidata deputy director Ben Domenico strolled over to the Unidata booth. There, he found online displays centered on Colorado’s uneventful weather, rather than the spectacular floods then unfolding in California. Before heading to another talk, Ben asked Steve if he could configure a quick prototype for LEAD that would steer itself—in essence, detect weather features of interest in the upcoming forecast and then focus on those features at higher resolution with movable grids.
“Chiz said, ‘Sure, I can do that in an afternoon,’” Ben recalls.
It took a little bit longer, but Steve quickly produced a striking preview of what LEAD could produce. On its
own, the new prototype successfully predicted and followed a damaging ice storm across the Southeast in late January (see graphic).
Thinking nationally, modeling locally
Nested observing and modeling systems, which place a sharper grid inside a larger, lower-resolution one, have been around for years. Starting in the mid-1990s, Unidata began offering a “floater” system so that universities with limited Internet access could obtain high-
resolution satellite and radar data for a specific region of weather interest, rather than for the entire country. The areas were chosen each day by students at Millersville University.
“The idea is that our data and forecast systems should be responding to the weather ”
“This was a way of allowing universities to get imagery focused on interesting weather and not have to obtain or pay for all of the data,” says Steve. However, he adds, “Most universities now have the bandwidth to receive those data. The whole paradigm that we developed 12 years ago needs some revamping.”
Once it’s approved by the Unidata Users Committee, the newly automated system created by Steve should relieve Millersville of having to select each day’s target area manually.
While not formally part of LEAD yet, this work is in line with the project’s stated mission: building a new framework that can be used in “accessing, preparing, assimilating, predicting, managing, mining/analyzing, and displaying a broad array of meteorological and related information, independent of format and physical location.”
One of LEAD’s many tasks is to make it easier for modeling systems to assimilate data from local sources. For example, mesoscale networks can include more than 100 surface weather stations across a single state. Many of these networks are proprietary, designed to serve utilities and other firms for a charge. Such data aren’t routinely incorporated into models produced by the National Weather Service. But local modeling groups might be able to arrange to use the data, says Ben, if the right setup were in place to make them useful.
Staffers in LEAD also intend to develop a modular set of connectivity components, including grid technology, that will allow users to link various models and data sources more easily. Unidata’s prototype shows how the pieces could fit together.
Unidata plans to update those pieces as LEAD evolves. Software engineer Jeff Weber is upbeat about the possibilities. In theory, he says, “you’d be able to use local modeling expertise and then stitch these local high-resolution runs into a national mosaic,” much like the national radar mosaics familiar to TV weather watchers. “That’d be a pretty exciting product.”
• by Bob Henson
The graphic below shows two different, but overlapping, forecasts of the precipitation that led to a major ice storm across the Southeast on Jan. 28-29. The outlook from the National Weather Service's Eta model is illustrated by contour lines. Unidata's Early LEAD Prototype identified the area of peak forecast interest (the gray rectangle) and produed a more detailed forecast (shown by shading) that automatically tracked the storm during the same period. (Illustration courtesy Unidata.)
On the Web
Unidata's Early LEAD Prototype
Also in this issue...
Playing it Safe
Spring Fling '05
Taking the LEAD
Just One Look
Staff Notes home page | News Center