•         advanced

End-to-End Early LEAD Prototype in Operation

Superficial draft document and glitz by Ben Domenico
Actual work done by others (especially Chiz) who shall remain nameless.
February 28, 2007

In a conversation inspired by Kelvin's AMS presentation, a number of us in the Unidata booth discussed whether there was a way to connect a set of existing tools so that a high resolution local model could be guided by some metric of where the "interesting" weather would occur during the forecast time. Steve Chiswell had been thinking about GEMPAK analysis tools that would automatically create high resolution radar and satellite datasets based on NCEP national Eta forecast precipitation rates. He decided to use that algorithm to determine where to focus the next 24-hour Workstation Eta run. And the rest is history.

Early LEAD Prototype

What we ended up with can be thought of as an end-to-end illustration of how the basic components of LEAD will work together to guide high resolution local forecast models -- based on the weather itself. In this prototype, the components are hardwired together in the following fashion:

1. IDD feeds NCEP national Eta data into decoders which create files in GEMPAK and netCDF form on motherlode.
2. National Eta GEMPAK input files go into GEMPAK "engine" that determines interesting weather based on forecast precipitation.
3. GEMPAK creates location on which to center next 24-hour Workstation Eta run. This GEMPAK algorthm is not the sophisticated ADAM data mining algorithm that will eventually be used in LEAD, but it serves the same function in this Early LEAD prototype and gives valuable experience for the kind of input, output and configuration that will be needed in the final component.
4. Zero-hour forecast of NCEP national Eta is used to initialize the Workstation Eta. Of course, the NCEP Eta zero-hour forecast is not the same as assimilating local observational data into the local model. But the zero-hour forecast is the result of NCEP's national data assimilation system so it's a reasonable placeholder until the full ADAS system is available as a web service.
5. Workstation Eta output is stored on motherlode server in netCDF where it can be analyzed with the IDV. In the final version of LEAD, WRF will be used rather than the Workstation Eta. But the Workstation Eta provides a chance to see how a local forecast model can be guided by where the interesting weather is happening.

Schematic

The diagram below is an adaptation of a LEAD schematic developed in the initial discussions of how LEAD tools would be connected to one another. The diagram has been updated to show the components running in the Early LEAD prototype with the final components and dataflows identified with dashed lines and outlines.

We need to begin creating a storage area for the input and output files -- perhaps along with the high resolution radar and satellite "floater" images over the same area as the local forecast. These datasets and catalogs should provide the raw materials for case studies that the meteorological and education teams can begin building on. For a quick demo, a special storage area was created manually (in this case on motherlode since our LEAD testbed has hardware problems) for storing case study collections and catalogs. But, of course, in the long term the catalogs will be created automatically

Early LEAD in Operation

Last week, Atlanta's bad luck was our good fortune in that the Early LEAD system focused in on the ice storm that ravaged the Southeast. CNN.com described it thus.

The gray background shows the extent of the two different (but overlapping) local forecasts at this particular time. The forecast precip at the surface is shown in the blue color shaded areas. The national NCEP Eta forecast precip is shown in the contour lines. You can see the differences in spatial resolution and, if you get the IDV running, you can animate the images and see the difference in time resolution as well. It's fascinating to watch the local forecast domain anticipate the storm location based on the GEMPAK analysis of the national Eta forecast.

One thing you might want to try is to examine the forecast vertical temperature profile to see why the result was freezing rain rather than snow or sleet. Something like this:

Then you can move the probe around and see how the profile changes with position.

Note also that a THREDDS catalog for the complete set of datasets in this ice storm case study is available at:

http://www.unidata.ucar.edu/projects/THREDDS/DataPublications/EarlyLEAD/EarlyLEAD.xml

By plugging that into the IDV "Catalogs" tab in the Data Source Chooser, you can get access to the latest Workstation and NCEP Eta datasets as well as those in the case study collection. If you do access the latest workstation Eta dataset, keep in mind that the location is now being driven dynamically and it will likely be centered on a completely different area where the interesting weather is occurring at this time. And it will be appended to the end of the animation since the time is later than that of the case study.

An automated gif of tomorrow's high resolution Workstation Eta forecast for the automatically calculated region of interest is shown below:

See if you can figure out where it is and what's in store for the area.

Background Information

The algorithm Chiz uses in GEMPAK to automatically set the moving location of the Workstation Eta forecast model is described in:

http://www.unidata.ucar.edu/software/gempak/wseta/index.html

Hopefully this Early LEAD Prototype will give us a sense of how all the pieces have to work together in the end, what the key data inputs are for each component, what the components will produce, and the sorts of configuration options will be needed in the final system. Clearly there is a huge task before the LEAD team to put the final components (ADAM, ADAS, WRF, automated THREDDS catalog generation) into place and to transform them into web services that can be orchestrated by the end user. But at least this early prototype provides a working example of one of the main objectives (automated steering of local models based on incoming data).

Contact Us   •   Site Map   •   Search   •   Terms and Conditions   •   Privacy Policy   •   Participation Policy

Unidata is a member of the UCAR Office of Programs, is managed by the University Corporation for Atmospheric Research, and is sponsored by the National Science Foundation. P.O. Box 3000   •   Boulder, CO 80307-3000 USA   •   Tel: 303-497-8643   •   Fax: 303-497-8690