A Closer Look at the Performance of Automated vs. Traditional Terminal Aerodrome Forecasts

Automation of weather forecasts is a controversial subject that involves many important issues. The purpose of this paper is not to address those issues but rather to point out some interesting verification results regarding an automated aviation weather forecast and the National Weather Service TAF (Terminal Aerodrome Forecast).  

Harris Corporation, provider of WARP and OASIS weather processing systems for the FAA, has a fuzzy logic based forecast system, FuzzyTAF, which produces automated short term aviation forecasts with output similar to that of the TAF. FuzzyTAF forecasts have demonstrated significant skill over a verification period of almost five years. Verification studies of FuzzyTAF have always included comparisons with the TAF, but this verification has previously only compared FuzzyTAFs with TAFs that were issued at the same time and were valid at the same time. Since the FuzzyTAF is automated and is updated whenever observed conditions change, it is never more than one hour old. However, the TAF is issued every six hours and therefore at time of use could be up to six hours old. For the very important next six hours to a decision-maker, this potential difference in forecast age leads to the question: How does an automated forecast that is 0 to 1 hour old compare with a TAF that is 0 to 6 hours old?  

A comparative verification study of FuzzyTAF and TAF forecasts for this “next six hour” period was completed for the year 2004 and part of 2005. Results of this study indicated that the automated FuzzyTAFs consistently outperformed TAFs for “next” hours 1 through 6 for ceiling and visibility, both overall and for only Instrument Flight Rule (IFR) conditions. This was true even for periods when the TAFs outperformed FuzzyTAFs based on traditional verification for forecasts issued at the same time and valid at the same time. This paper presents details of this study.