The International Division is responsible for the development and maintenance of the main GLOBE Website (excluding data visualizations), real-time GLOBE data acquisition tools, the central GLOBE database, and the mirrored GLOBE Web and database systems.

In looking at ways to improve site performance and keep abreast of recent technology, the GLOBE team explored implementation of J2EE components (Java Servlets and Java Server Pages) into the site. As a case study, all of the content-rich pages of the site were imported into the database, code was developed that dynamically requested content, and then the Webpage was built through the Tomcat Servlet Container. The Oracle large object (LOB) support was examined to see how more content could be added into the central database. The GLOBE student investigation reports are now stored in CLOB columns (character large objects, ideal for large text strings) in the database.

The GLOBE data acquisition code base is contantly growing so that students, teachers, and researchers can continue to collect new datasets. GLOBE students can send their data via Web forms on the very interactive site or via an email message (typically used by schools wishing to report a lot of data at once). In 2002, tasks involved developing code for acquisition, processing, and storage of data from a digital multiple-day or single-day maximum/minimum thermometer, hummingbird observations (the first protocol to study animals), and phenological gardens.

In view of GLOBE's strong international roots, the GLOBE team is committed to having the Website translated into the six United Nations languages, and with interest shown in non-UN languages, these are being accommodated as well. In addition to Dutch and German translation accomplishments, a Japanese coordinator has begun translating the GLOBE site into Japanese – the first Asian language. It is rewarding to participate in tasks related to displaying the first data entry pages in Japanese and foresee their completion in the coming year.

A separate, nonpublic Website allows GLOBE Headquarters staff in Washington, D.C., GLOBE partner groups, and country coordinators to track GLOBE workshop participation, school contact information, school reporting rates, etc. New interfaces were designed so that GLOBE partners can select and specify trainers for their own workshops based on the prospective trainer's experience and qualifications (for specific protocols), availability, and location. This information on the trainers is maintained at Headquarters along with another set of interfaces. The FSL GLOBE team enhanced and released a Web-enabled database query tool that allows Headquarters to look at any of the data stored in the central database without needing to know any database query languages such as SQL.

When running an operational system, the back-end systems require constant maintenance and upgrades to help ensure that they stay highly available and are kept current with the latest software technologies. Another accomplishment involved moving a significant fraction of our software and database files to a NetApp filer to centralize data storage and to improve I/O performance. Taking advantage of the "snapshotting" capability of the filer, we increased the uptime of the database for backups, and shifted from cold to hot backups so that the database does not need to be shut down as often as was required in the past. The Oracle database was upgraded from version 8i to 9i to enable the addition of more new features to the Website.

The Web servers also needed attention since all of our nonvisualization Web servers now operate on Linux platforms. Furthermore, the GLOBE training server that accommodates teachers at workshops uses a database that also runs on Linux. The German GLOBE Program was fully supported by helping them move their legacy GLOBE mirror server to a new platform. This operational mirror server can now run the entire Website, including the current visualization system. Though schools in Germany and the contiguous countries predominantly use the mirror, they may decide to convert this mirror to a failover for the mirrors located in the United States (FSL and NASA/Goddard Space Flight Center).

When the GLOBE science and education releases a GLOBE 2003 Teacher's Guide in the spring, many new protocols will be added. These include contrail observations, fire fuel ecology, ground surface temperature, freshwater macroinvertebrates, bird and seaweed phenology, and various modifications to the hydrology and land cover measurements. GLOBE staff will also continue encouraging schools to use automated data logging devices from providers such as HOBO, Davis, and AWS/WeatherNet. With a potential for the daily data ingest to grow by as much as a factor of 100 due to the 15-minute increment reports, steps will be taken to ensure proper storage handling for the additional data. A primary undertaking will be to complete the work to accommodate the new protocols which will allow entry of these new datasets.

The GLOBE Website will continue to be improved and made more user-friendly. Randomized "smart info" (e.g., "A GLOBE school in Sydney, Australia, reported the same temperature and humidity you reported on this day.") will be added to data entry verification pages to help ensure data quality while making the data entry process less repetitive to students. Another plan is to develop an interface whereby schools can upload images (such as data graphs and site photos) directly to the Website, at which time they could be approved before they are posted. Now is an appropriate time to redesign the "administrative" Website from the ground up. Years ago this site was designed for two or three users, but today its use is much more widespread — about 300 users in the large GLOBE partner community.

The mission-critical GLOBE database will continue to be modified and maintained as more data (quantity and type) are ingested from additional participating schools. With the upgrade to Oracle 9i software, spatial queries will be developed that help to improve visualizations by making them more GIS-capable. Also, with more complete XML support in Oracle 9i, GLOBE developers will need to familiarize themselves with the new version. Building on successful tests of migrating Website content to the database, the content will be placed into the operational database to centralize content management and make authoring and editing easier.

The strong forecasting infrastructure that has been built at CWB includes greater data collection, improved observation systems, high-performance computing, and efficient management capabilities. CWB is positioned to generate new and more useful forecast products and take advantage of the more powerful techniques under development at FSL and other NOAA laboratories. The effectiveness of the CWB-FSL cooperation is based in large part on CWB’s willingness and ability to develop and use customized products with associated technical support. FSL’s mandate to provide useful technologies fits with CWB’s real-world forecasting needs.

• Local Analysis and Prediction System (LAPS)
• Forecast Assistant System (FAS)
• Continuing interaction on earlier cooperative projects.

Forecast Assistant System – Through this project, support continued toward upgrading CWB's WINS II system. FSL provided an upgraded AWIPS Build 5.1.2 with the upgraded GFESuite (Graphical Forecast Editor) software. FSL also provided D3D software to CWB for further evaluation and customization. Last October, FSL's Technical Lead for GFESuite visited CWB to provide extensive training to four forecasters on the use of GFESuite, including the spatial and temporal editors, grid manager, and the necessary steps required to create and execute smart tools. They worked with two developers from CWB, answered questions about GFESuite software, including the latest verion (RPP18), nd provided training on the use of Python for developing smart tools. A less technical seminar was presented to the CWB general audience on the functions and use of GFESuite.

• Upgrade of the FAS nowcasting system
• Implementation of the Local Analysis and Prediction System (LAPS)
• Implementation of the Mesoscale Analysis and Prediction System (MAPS) Surface Analysis System (MSAS) quality control and monitoring system
• Provision for forecast training and risk reduction.

Upgrade of the KMA Nowcasting System – The FAS became operational at KMA and was also deployed at six Regional Offices in July 2002. A KMA visiting scientist worked with FSL staff to upgrade the FAS to the AWIPS 5.2.2 Build, and incorporated Korean menu changes (Figure 76) and necessary unit conversion changes to MKS units. FSL provided KMA with the AWIPS 5.2.2 Build, which is the latest and the last development version. This version has many new features and improvements, such as the ability to customize the AWIPS MSAS analysis to different localizations.

Implementation of the Local Analysis and Prediction System – Over the past five years, KMA scientists have adopted three-dimensional analysis software developed from LAPS. The goal of this task was to implement the latest FSL/LAPS II analysis as an integral part of the KMA nowcasting system (FAS) so that forecasters could access LAPS II products through a single display system provided by FAS. The diabatic initialization capability (Hot Start) has been incorporated into LAPS to perform dynamic balance adjustment.

The available FX-Net products are categorized into four groups: satellite data, model graphics and observations, radar imagery, and model imagery. Wavelet transform is used to compress model and satellite imagery. The application of this relatively new compression technique is critical to the success of delivering very large-size imagery via the Internet in a reasonable amount of time. The small loss of fidelity in the imagery is acceptable in exchange for very high compression ratios. Processing time can be further minimized by pregenerating and compressing all satellite data on the FX-Net server side. In contrast to the satellite imagery, the radar imagery is encoded in a standard lossless image compression format (GIF) and the small-sized model graphics are represented in a standard vector graphics format.

The new FX-Net/AQ datasets include six parameters (O₃, CO, NO, NO_y, SO₂, and condensation particles) that are continuously measured at three UNH sites (Mount Washington, Castle in the Clouds, and Thompson Farm) located in the state of New Hampshire (Figure 77). The FX-Net/AQ user also has access to the data from 13 wind profilers recently installed across the New England states. In addition, hundreds of new meteorological surface observation data, fixed buoy records, and ship measurements are available on the latest FX-Net menu. On the continental U.S. scale (CONUS), FX-Net displays the hourly average of the national EPA low-level ozone data. The FX-Net team is still working on the ingest and display of the MM5 air quality model, and once this task is completed FX-Net can truly be called a "real-time air quality workstation."

The National Interagency Fire Center (NIFC) requested (in 2001) that FX-Net be modified to permit its use as the primary real-time meteorological workstation by fire weather forecasters at NIFC and at the Geographic Area Coordination Centers (GACC). The plan called for the FX-Net workstation to be used during the 2002 fire season on an experimental basis, with the FX-Net server located at FSL in Boulder. If the workstation was accepted by the fire weather forecast community at NIFC and GACC offices, the agreement called for the introduction of an operational solution for the 2003 fire season. Figure 78 shows an FX-Net display in a GACC room set up.

The FX-Net team added a variety of new functions to the FX-Net client with the goals of making additional products available to the fire weather community and adding new user-friendly tools to the client. One of the outstanding new datasets is a complete text browser that allows for the display of a large number of National Weather Service (NWS) forecast and discussion products. Additional new tools allow for the export of products displayed in the primary window, preference client settings rather than former changes to the configuration file, and the change of contour intervals for displayed model products. Another addition to FX-Net involves two special display scales, the Northern Rocky Mountains and Southern Rocky Mountains, used for viewing high-resolution satellite imagery in areas with high potential for wild fires.

As is now well documented, the 2002 fire season was one of the worst ever recorded for many states. The operational demand by NIFC and GACC forecasters for complete sets of meteorological products was paramount to support the issuance of the best possible daily forecasts. In response to the high operational tempo and requirements, the FX-Net team pursued every possibility to improve the reliability of all components of the system involved in the FX-Net data stream. Hardware was exchanged with newer systems, data streams rerouted, and multiple backup systems were installed. By the end of July, a significant improvement in reliability of the FX-Net-related systems was achieved, with reliability exceeding 98%.

The new list of fire weather products, the new functions added to the client, and the significant increase in reliability made FX-Net a truly functional fire weather forecaster workstation. It also resulted in very positive feedback from the NIFC management as well as from the GACC forecaster community.

Real-time Air Quality Forecast Workstation – FX-Net will increasingly focus on adding products to support real-time air quality forecasting.

Fire Weather Forecasting – Since FX-Net has become the primary meteorological workstation to support the fire weather forecasters in all national GACC offices and the NIFC headquarters in Boise, Idaho, many special products and functions will be added to the system as part of ID's continued support.

NWS: Western, Southern, Alaska, and Pacific Region – Early in 2003, four complete FX-Net systems will be installed at four NWS Regional Headquarters to support Incident Meteorologists (IMETs) in the field as well as to provide remote data collection offices with AWIPS-like products.

The average error reflects the overall quality of the reconstructed data. In this example, the average error was nearly an order of magnitude smaller than the predefined maximum error. This large difference between the maximum and the average error means that most of the grid points in the reconstructed field show an error much smaller than the defined maximum error.

The compression test was done on an 850-MHz Pentium III desktop computer. The average compression time to encode each field (2.5 MB) is about 2 – 3 seconds.

Compared to typical lossless codecs with the same precision requirements, this codec achieves 2– 6 times higher compression ratios. It implies that for a typical model output sized at 1 GB, if transmitted over a 1 Mbps communication channel, the transmission time can be reduced from about 2.8 hours to about half an hour.

The presented data compression scheme is asymmetric by nature, in that it takes more time to encode the data than to decode them. This is beneficial in the practical implementation, since there is usually more computing power in the encoding machine than in the decoding machine.

The approach to control the maximum round off error is computationally simple, or somewhat ad hoc. It is feasible with our current operational environment; however, an ideal algorithm should be able to find the best bits allocation that minimizes the maximum error. An efficient algorithm that can carry this out would be very useful both in theory and in practice. The vertically and timely adjacent frames are highly correlated. Current results only reflect the compression performance of this scheme on the two-dimensional field (in the horizontal plane). Three- or four-dimensional separable wavelet transforms can be applied to the volume data. To meet the robust (error propagation control) requirement, each partitioned group of coefficients can possibly be encoded individually into an independent bitstream to build a more error-resilient codec.

The above technology will feed directly into the project's workstation developments. Compression ratios of the above magnitude now allow for sending high-resolution forecast models (with typical model outputs sized at 1 GB and larger) via low bandwidth to an FX-Net or WorldWide Weather Workstation (W₄) client in very reasonable amounts of time. This will make satellite broadcasting mechanisms (with a bandwidth of 128 kbs or less) for meteorological datasets feasible.