Overview of the Center for Analysis and Prediction of Storms

University of Oklahoma

The Center for Analysis and Prediction of Storms (CAPS; http://caps.ou.edu) was established at the University of Oklahoma in 1989 as one of the first 11 National Science Foundation Science and Technology Center. Its mission was, and remains the development of techniques for the computer-based prediction of high-impact local weather, such as individual spring and winter storms, with the NEXRAD (WSR-88D) Doppler radar serving as a key data source.

Along the path toward fulfilling this mission, CAPS developed the Advanced Regional Prediction System (ARPS; http://caps.ou.edu/ARPS). The ARPS is a fully automated numerical prediction system designed for both research and operational application at scales ranging from continents down to cities. It includes a data ingest, quality control, and objective analysis package known as the ARPS Data Analysis System (ADAS; http://caps.ou.edu/adas.html); a single-Doppler radar parameter retrieval and 3DVAR/4DVAR data assimilation system; the prediction model itself; and a web-based data display and decision support system. The value of the ARPS was recognized in 1997, when CAPS received both the Discover Magazine Award for Technology Innovation as well as the Computerworld-Smithsonian Award.

When CAPS was established over 10 years ago, few in the scientific community believed that storm-scale weather, especially thunderstorms, possessed any reasonable degree of predictability. By 1996, CAPS had conclusively demonstrated that prediction on this scale indeed was possible, and had done so through a mix of case studies as well as real-time operational tests in close collaboration with the Norman, Oklahoma National Weather Service Forecast Office, NCEP Storm Prediction Center, US Air Force, FAA, National Center for Atmospheric Research, Pittsburgh Supercomputing Center, and the National Center for Supercomputing Applications.

The culmination of CAPS' research can best be illustrated through an example. Shown below is a sequence of hourly radar images (top panels) from the Fort Worth, Texas tornadic storms of 29 March 2000. Colors indicate precipitation intensity, with higher rates indicated by the warmer colors. The city of Fort Worth is shown by the magenta dot. Shown in the lower three panels is the equivalent radar reflectivity from a 3 kilometer-grid spacing ARPS forecast, initialized at 2300 UTC with WSR-88D (NEXRAD) radar and other data. The degree of agreement between observations and forecast, even out to 4 hours, is exceptionally good, and further improvements are anticipated with more sophisticated data assimilation techniques and forecast cycling. Of course, not all forecasts exhibit this degree of quality, and thus CAPS continues to explore fundamental issues in the predictability of small-scale atmospheric motion, including the use of ensembles as a means for dealing with uncertainty.

The private sector began to see value in CAPS research by the mid-1990s. In 1996, American Airlines funded a 3-year, $1M grant to apply the CAPS prediction system to commercial aviation. Around this same time, CAPS received an NSF Academic Research Infrastructure (now Major Research Infrastructure) grant to purchase a Cray supercomputer, with additional support provided by the NSF EPSCoR Program. That computer played a major role in the American Airlines project, and American soon contributed financially to its enhancement. In 1999, American established an endowed professorship in the OU School of Meteorology, clearly as a result of its positive experiences with CAPS.

The opportunity to further make available the benefits of its technology led CAPS in 1999 to spearhead the creation of a private company, Weather Decision Technologies, Inc (http://wdtinc.com). WDT now has signed several major contracts, with American Airlines one of the founding customers. The transition of CAPS' basic research concepts to operational utilization took less than a decade, a remarkable achievement and a testimonial to the impact of the STC Program.

In recognition of its responsibility as a national research center, CAPS undertook a number of major initiatives that have delivered substantial benefits to society. The first was a multi-institutional effort, launched in 1998 and known as Project CRAFT (Collaborative Radar Acquisition Field Test; http://kkd.ou.edu/craft.htm), which demonstrated the capability of compressing NEXRAD Doppler radar data in real time and transmitting it over the Internet. This project has evolved into a national Abilene/Internet2 initiative and presently is serving as a blueprint for a new data infrastructure within the National Weather Service. Importantly, this effort was seed funded by the State of Oklahoma and the NSF EPSCoR Program, and now has generated more than $1.5M in support from NOAA.

Another national initiative, conducted during May, 1998 as an un-funded proof-of-concept test and known as the Storm and Mesoscale Ensemble Experiment (SAMEX), involved the generation of real time forecasts to determine the value of coarse-resolution (30 km) ensembles relative to single predictions made over nested domains at finer (3-10 km) resolution. The daily 25-member ensemble forecasts comprised 5 runs of the ARPS by CAPS, 10 runs of the MM5 by the National Severe Storms Laboratory, and 5 runs each by NCEP of the operational Eta and Regional Spectral Models. This output was compared to intermediate (10 km) resolution experiments of MM5, ARPS, and the Eta model over the central US, and to high (2-3 km) resolution forecasts from the ARPS and MM5 over Oklahoma and Texas. Other participating groups included the Air Force Weather Agency (AFWA) and National Center for Atmospheric Research (NCAR). Several National Weather Service Offices took part in the real time evaluation, along with the NCEP Storm Prediction Center. The results show that, in most cases, the ensemble of 25 forecasts is more skillful than any of the individual ensemble members or the ensemble of the controls. The outcome of SAMEX was so compelling that the National Weather Service's National Centers for Environmental Prediction now is organizing an operational forecast system based upon the multi-model concept.

CAPS also is a major participant in the $2 million NSF MRI grant that is creating "SuomiNet" -- a national geophysical instrument of GPS receivers that will provide real time, high-resolution water vapor and geodetic measurements across entire US. Finally, CAPS continues to receive funding from the FAA to conduct experiments in storm-scale weather prediction, and now is involved in mining radar data to better understand the impact of thunderstorms on air traffic delays.

CAPS' visionary goal of storm-scale numerical prediction was aptly demonstrated during a 7-week operational test in the heart of the 1999 Great Plains severe storm season. Having been given fully dedicated access to the largest computer then available to the scientific community -- a new 256-node Origin 2000 at the National Center for Supercomputing Applications -- CAPS operated the ARPS on a daily basis in real time, producing forecasts at 3, 9, and 27 km resolution including a 5-member ensemble over the continental US. The crowning achievement of this test was the prediction, to within one half county accuracy 2 hours in advance, of the storm that produced an F5 tornado in the Oklahoma City metropolitan area on the night of 3 May 1999. This numerical forecast was the first ever of a deep convective storm using NEXRAD Doppler radar data and single-Doppler velocity retrieval techniques in an experimental operational framework. The results were shown in a Congressional Briefing and during Congressional Testimony later in the summer.

In an effort to learn as much as possible about every aspect of the May 3 event, CAPS organized the National Symposium on the Great Plains Tornado Outbreak of May 3, 1999 (http://caps.ou.edu/may3.htm), held from 30 April to 3 May 2000 in Oklahoma City. This conference brought together more than 400 meteorologists, social scientists, construction engineers, emergency managers, policy makers, and disaster relief workers from around the world in the first multi-disciplinary examination of a major tornado disaster. Furthermore, more than 1000 private citizens attended a half-day exhibition of safe room and advanced weather technology prior to the symposium. A special issue of the American Meteorological Society Journal Weather and Forecasting is being dedicated to papers on the May 3rd event, including those on non-meteorological topics.

Numerous other initiatives are underway at CAPS. The ARPS has now been adopted as the Korea Meteorological Administration's heavy rainfall prediction system, and further research is underway to explore better utilization of available Doppler radar data. CAPS also is one of four core participants, along with the NOAA Forecast Systems Laboratory, NCAR, and the NCEP, that is developing the Weather Research and Forecast Model (http://wrf-model.org), which is expected to replace current operational mesoscale models and also serve as a community research platform.

Most recently, CAPS signed a 5-year, $8.1M research and development contract with the Williams Energy Marketing and Trading Company, which is a component of the Williams Companies in Tulsa, Oklahoma. This effort is part of a $10.6M alliance between Williams and the University of Oklahoma, elements of which include an endowed Chair in the School of Meteorology, three Fellowships in the Master of Science in Professional Meteorology Degree Program, a new supercomputer, and funding for numerous scientists and graduate students.

For more information about current research and other activities at CAPS, visit the home page at http://caps.ou.edu/.