Forecasters at National Weather Service (NWS) Weather Forecast Offices (WFO) prepare and disseminate Terminal Aerodrome Forecasts (TAFs) operationally. The Aviation Forecast Preparation System (AvnFPS) assists forecasters in these tasks. The Localized Aviation MOS Program (LAMP) is a statistical system which provides forecast guidance for sensible weather elements by using the most recent surface observations, output from simple models, and MOS guidance based on the Global Forecast System.

One of the features of AvnFPS is the automated generation of TAFs based on LAMP guidance. The technique that is used currently to generate TAFs from LAMP has met with little acceptance among NWS forecasters. This is because the automated techniques fall short in making judgments about TAF content and summarizing techniques. Also, a number of TAF phrases which come naturally to human authors are difficult to generate based on current guidance techniques, e.g., thunderstorms in the vicinity (VCTS).

This paper will describe a new technique for automatically generating a TAF in AvnFPS. The TAF Update with Guidance (TUG) technique compares data from the official TAF with probabilistic guidance from LAMP to generate an updated version of the official TAF. In those cases where LAMP's best category forecast agrees with the official TAF, that portion of the official TAF is preserved. Conversely where LAMP's best category forecast disagrees with the official TAF, the associated LAMP probability and threshold used to decide the best category are compared to measure the amount of disagreement. If the values are close (say, within 10%), that portion of the official TAF is carried into the new TAF. If the values are not close, that portion of the new TAF is generated from the LAMP guidance.

The TUG technique generates automated TAFs that bear considerable resemblance to those written by humans but incorporates LAMP guidance in a judicious way. TAFs generated by this method will be more useful to both forecasters and the aviation community.