Boulder meeting kicks off initiative for fine-scale mesoscale analysis
by Bob Henson
In days of yore, meteorologists judged their colleagues on how deftly and quickly they could produce a surface analysis. Their challenge was to accurately draw highs, lows, fronts, and other key features, interpolating from scattered surface reports.
Computers now take care of this task, by and large. However, the need for a careful depiction of current and previous conditions (such as temperature, relative humidity, winds, clouds, and precipitation) hasn't vanished.
More than 70 forecasters and researchers met at NOAA's Boulder laboratories on 29-30 June to launch a campaign for enhanced surface analyses, this time at a level of detail only dreamed of by yesterday's pencil-and-paper set.
Brad Colman and John Horel. (Photo by Bob Henson.)
Chaired by Brad Colman (NOAA National Weather Service, or NWS) and John Horel (University of Utah), the meeting was arranged by the U. S. Weather Research Program (USWRP) to gain consensus and build community momentum for what's being called an "analysis of record" (AOR) program. Such an effort is sorely needed, say its proponents, in order to meet the increasing demand for environmental information at fine spatial resolution.
"The skillful generation of mesoscale objective analyses is one of the key atmospheric challenges for the 21st century," says USWRP lead scientist Bob Gall (NCAR)."We wanted this meeting to serve as a starting point to bring forth the focus and energy that this problem merits."
How to check on a million forecasts
Funding for the meeting came from the NWS, where the mesoscale forecast revolution has arrived in a big way. The experimental National Digital Forecast Database (NDFD; see "On the Web") provides a forecast of weather variables at resolutions as fine as once per hour and up to a week in advance, all on a 5-by-5-kilometer (3.1-by-3.1-mile) grid. The database is produced by a blend of lower-resolution model output and human-forecaster input. Output is provided routinely on the Web sites of NWS local forecast offices.
The NDFD could provide an idea, for instance, of how much the temperature might drop from 10 p.m. to 11 p.m. in the Times Square area on New Year's Eve—and how much cooler it might be in the neighborhoods of Harlem or Wall Street.
Such outlooks promise to bring a new level of service to private-sector meteorologists, emergency managers, and the public. However, millions of weather stations would be needed in order to directly assess the quality of the NDFD's precision outlooks.
The AOR would provide the next best thing, according to its proponents. It would likely start with a recent model-generated forecast on the 5-km grid and adjust it based on the available observations. The final products might include a provisional analysis, available to forecasters within 30 minutes, and an archival version, produced a day or so later. Forecasters would likely use the fast analysis for short-term outlooks, since its resolution and detail would be an improvement over that in the existing operational models.
The AOR's keen sight could extend backward in time. A set of global reanalyses from 1948 onward, produced by NOAA's National Centers for Environmental Prediction and NCAR, has become a mainstay of climate researchers. In the same vein, the AOR team envisions a set of retrospective U. S. analyses with much finer resolution, ranging as far back as the 1970s.
"Once this data set gets out, there will be hundreds of people who want to use it," says Fred Carr (University of Oklahoma)."Gridded data are great if they can represent the mesoscale well." Along with everyday weather outlooks, the AOR could help gauge the accuracy of other mesoscale research and forecasting efforts in areas ranging from dispersion of airborne hazards to fire management. According to Colman, "Mesoscale models are increasingly being used to predict not only the weather but where pollutants and smoke from wildfires are likely to go. The AOR may serve as the yardstick by which the success of those models can be measured."
How to check on a million forecasts
Follow-up activities from the meeting are in full swing (see "On the Web"). The first meeting of a new NOAA/USWRP committee on mesoscale analysis, recommended by attendees of the June workshop, will likely take place in September. Participants are working on a summary paper for publication this winter. The organizers will also work to build financial support for the AOR, a challenge in the midst of tight budgets.
One of the issues that the committee will face at the outset is the limit of current analysis methods. "Any analysis is imperfect, because observations have errors," says Horel. "We want to take advantage of current technology to make the best possible mesoscale analyses now, as well as spur research and development that will lead to improved analyses in the future."
Some other potential flies in the ointment include thunderstorms, low-level inversions, and cold air trapped in valleys-in short, any small-scale weather features that hide within the confines of our current observing and modeling systems. To help spot more of these features, says Horel, "We'll coordinate closely with other NOAA- and USWRP-sponsored programs aimed at deploying additional sensors, such as the modernization of the NWS Cooperative Observer Program."
Researchers will need to figure out how best to prioritize observations based on a sensor's likely error rate, and they'll have to account for phenomena that can't be observed. Carr believes a reliable portrait of individual, garden-variety showers and thunderstorms, for example, is perhaps five years away. "You're always going to have substantial error at some scale," he adds.
Several participants stressed the need for care in using model-derived data to fill gaps in the observed record. "The frequent assumption by many modelers that model analyses are better than observations has always concerned me," says participant Barb Brown (NCAR).
Even with these caveats, many researchers and forecasters at the June meeting came away enthusiastic about the promise of the AOR. "It has huge scientific implications," says NCAR's Joshua Hacker,"if we can make it work."
For more details, contact Brad Colman or John Horel.
A Community Meeting on Real-Time and Retrospective Mesoscale Objective Analysis (includes presentations)
NWS National Digital Forecast Database