13th Conference on Applied Climatology and the 10th Conference on Aviation, Range, and Aerospace Meteorology

Tuesday, 14 May 2002: 9:00 AM
Forecasting Convective Weather Using Multi-scale Detectors and Weather Classification—Enhancements to the MIT Lincoln Laboratory Terminal Convective Weather Forecast
William J. Dupree, MIT Lincoln Lab., Lexington, MA; and M. M. Wolfson, R. J. Johnson Jr., K. E. Theriault, B. E. Forman, R. A. Boldi, and C. A. Wilson
Thunderstorms account for a significant fraction of air traffic delays in the US, resulting in enormous annual economic losses. In response, the FAA has assembled an expert scientific team to produce automated convective weather forecasting tools. For the past 4 years MIT Lincoln Laboratory has developed, tested and evaluated a prototype Terminal Convective Weather Forecast (TCWF) algorithm, which has proven to be extremely valuable to the air traffic community at several test bed sites.

The success of the TCWF envelope tracking algorithm is due partially to the realization that tracking and extrapolating at scales on the order of a 13 x 69 km [size of the elliptical match filter], produces a good estimate of the location of weather on the 0-1 hour time scale. However the envelope style filtering fails to provide adequate forecasts for smaller airmass storms in some critical operational situations.

The success of the envelope tracking is due in part to eliminating individual cell motions within the larger line storms. Air mass storms can migrate and grow in different directions, so use of a directional average for quality control of line storm scale aberrant vectors is not appropriate when used on air mass scale features.

Our approach to this problem is to classify weather into line storms, isolated cells and stratiform regions by processing precipitation images using matched filter image processing techniques developed at Lincoln Laboratory. Each region is then assigned tracking vectors that are processed and trimmed based on the region statistics, producing superior short-term and equally good long-term forecasts. Techniques for producing convective weather forecasts and evaluation of the new algorithm performance will be discussed.

* This work was sponsored by the Federal Aviation Administration under Air Force Contract No. F19628-00-C-0002. The views expressed are those of the authors and do not reflect the official policy or position of the U.S. Government. Opinions, interpretations, conclusions, and recommendations are those of the authors and are not necessarily endorsed by the US Government.

Supplementary URL: