Improving Efficiency of Terminal Aerodrome Forecast Production

Terminal Aerodrome Forecasts (TAFs) are a widely accepted international form of aviation forecast used for flight planning procedures at all major airports. Typically, the production of TAFs is a time-consuming, manual process carried out by Operational Meteorologists. In recent years, a process has been developed at the Met Office that uses Numerical Weather Prediction (NWP) data to automatically generate TAFs. These are referred to as ‘first guess TAFs’ as they are intended to be used by Operational Meteorologists as a first guess solution in their TAF production process. The main aim of providing first guess TAFs is to improve the efficiency of TAF production, which would free up Operational Meteorologists’ time for other activities. Based on standard TAF verification scores, the first guess TAFs perform very well, often better than manually produced TAFs. Furthermore, a year-long trial conducted at Heathrow Airport demonstrated the potential for first guess TAFs to increase the efficiency of TAF production. Despite these successes, the trial revealed a significant shortcoming: first guess TAFs were prone to ‘going bust,’ meaning the observed weather deviated enough from the forecast that amendments to the TAF were necessary. To address this, machine learning techniques were introduced to predict when first guess TAFs might go bust, enabling adjustments to the NWP data. This approach has successfully reduced busts and improved verification scores. The close collaboration between operational and research teams has been crucial to the ongoing success of first guess TAFs and will continue to be essential as this forecasting method transitions from a research project to a valuable operational tool. This talk will explore the challenges and lessons learned in this transition.