Sample Data Interactive Publication:

Comparisons of the forecasts by the NCEP ETA model and
a local Workstation ETA for the Colorado area:
surface air temperature predictions

Notes on the underlying technology

The technology used in this publication enables you to interact directly with datasets, catalogs of datasets, as well as documents stored on remote servers. To utilize the technology effectively, it's necessary to have a well-configured desktop workstation (at least 512 MB of memory) with the Java Webstart application installed.

The images and animations that show up on your screen via the Unidata Interactive Data Viewer (IDV) are not just pictures; they are interactive visualizations generated from the datasets, portions of which are accessed from remot servers or from the disks on your own desktop workstation.

In this sample case, the datasets you start with are the output of forecast models:

The NCEP model output is approximately is approximately 52 MB in size. The Workstation ETA covers less area and time but at a much higher resolution and results in an output file that's over 800 MB -- nearly a Gigabyte of data. These datasets are stored on a computer at the National Center for Atmospheric Research; the IDV interacts with them via established client/server protocols. A key element of the client/server technology used in these documents and visualizations is that only the needed parts of the huge files are transferred as they are requested. In spite of that, it takes a while to transfer even that subset and render the images on your workstation.

Since these are not pre-generated pictures, you have access to the complete IDV visualization system and to all the data on the server. Thus you can interact directly with the data to create your own visualizations. In fact, you can access other datasets and include them in the analysis.

Just Do It

If you don't already have Webstart installed, the instructions for doing so can be found at The Unidata WebStart Instruction Page.

Once you have webstart running on your workstation, the IDV configured with the appropriate datasets can be started by clicking on the following line:

Temperature forecasts from NCEP national forecast and high-res. local Colorado Workstation ETA.

What you see (initially): temperature forecast comparisons

These forecast models were run at nearly the same time. In fact the NCEP ETA model output was used to initialize the higher resolution Workstation model which covers only the Colorado area.

 

The forecasts were for July 16, 2004, a day in which there was considerable weather activity through Colorado. The visualization shown in this bundle compares the predicted temperatures at 2 meters above the surface. For those of you who aren't able to get the IDV running on your workstation, the screen dump below gives you a sense of the kind of visualization you'd be looking at.

The ETA model output is shown as contour (plan view) white lines whereas the higher resolution Workstation model output is shown as color filled contours. One can readily see the detail in the Workstation model that's not there in the NCEP model which covers most of North America. Using the controls in the upper right corner of the display window, the visualization can be animated through the entire forecast time. This animation reveals the higher temporal resolution of the Workstation model since the color fill image changes for some frames where the white line contours of the national model stay constant because there is no national forecast for those intermediate times.. The frames toward the end only show the NCEP model because the Workstation forecast only goes out 24 hours in time.

For more general background information , a concise description of the contents of various forecast models is at:

<http://dss.ucar.edu/datasets/ds335.0/docs/>

There is also a more all-encompassing description of datasets used in the atmospheric and ocean sciences in an NCAR tech note at:

<http://www.cgd.ucar.edu/cas/tn404/>

Integrated text, tools, and data

As you will see, HTML files (such as this one) are also viewable within the IDV which accesses it from a Web server at the UPC -- a different server from the one that provides the data. The catalogs that contain the URL pointers to the data files can be stored on still another network computer or on one's own workstation for that matter. The catalogs are in the XML language. For those who understand XML or are curious about it, an example catalog can be viewed at:

<http://www.unidata.ucar.edu/projects/THREDDS/BenStuff/Datadocs/Case2004Jul16.xml>

Topology from geopotential height at surface

There are two additional displays available but not active in the IDV as it comes up. These are images of a parameter called geopotential height at the surface which serves as a "proxy" for the topography underlying the model. By clicking on the control lines a the bottom of the image window, you can make either or both of them visible. One of these is for the NCEP national model; the other is for the local workstation model. Comparing these images by turning them on and off gives you some sense of the difference in resolution between the NCEP ETA and the Workstation ETA.

Exercises

One of the main goals of this mini case study is to provide an entry point for learning more about the capabilities of the client/server technology that makes it possible to interact directly with the datasets and to learn a bit about the science behind the models in the process. The following activities should help in that process:

  1. Work with the animation controls in the upper right of the image window to get a sense of how the different forecasts evolve in time.
  2. Think about what sort of additional data would be useful to overlay in order to verify the accuracy of the predicted conditions.
  3. Related to the objective of learning about the science, it is left as an exercise for the reader to figure out why and how "geopotential height at the surface" serves as a proxy for the topography used in the forecast models.
  4. To gain a better sense of the topologies, click on the #1 or #2 lines at the bottom of the image to turn on the topography displays, then hold down the right mouse button in the image window and move the mouse around to gain a 3D perspective on the image. If you rotate the image enough, you can actually view the temperature data from underneath the topography images. The next release of the IDV will allow you to "drape" the data over the topography.
  5. To examine other runs of the forecast models and to visualize other parameters, a special THREDDS catalog of the datasets has been prepared at: http://www.unidata.ucar.edu/projects/THREDDS/BenStuff/Datadocs/Case2004Jul16.xml
  6. You can view the XML source for that catalog by pasting the URL into your browser window, but more importantly, you can access the catalog from within the IDV by going through the Data -> New Data Source menues and selecting the "Cataloged Data" tab. Then paste the URL above as the "Catalogs:" entry. Then clicking on Update allow you to access all the national and local Colorado model run datasets saved in this case study. It also has pointers to the latest run of both the national and local ETA model runs so you can compare the current conditions to those in the case study collection.
  7. You can click on the icons at the beginning of each line in the catalog to open and close the lists within. Then you can highlight the model run of interest and click on "Add source" to make it available in the data selector where you can choose the Fields and Display you want to include in the visualization. Try a few different Fields and displays to gain a sense of how the IDV brings in subsets of data from remote servers and integrates it into the local interactive display.
  8. A serious issue of concern is to determine how much of the additional detail shown in the local Colorado model reflects reality so that it makes it possible to predict weather conditions more accurately for each locality in the state. Of course answering that question is also left as an excercise for the reader.