An Experiment to Measure the Value of Statistical Probability Forecasts for Aerodromes

Over the past couple of decades, considerable research has been performed to investigate alternatives to traditional numerical weather model prediction of surface weather conditions. In particular, observations-based statistical techniques have shown great promise for improving short-term forecasts of surface weather conditions. Vislocky and Fritsch (1997) demonstrated that such a statistical forecasting system has superior skill to numerical models for the short-term prediction of ceiling and visibility. This system considered a network of surface observations surrounding an observing site (often an airport) to produce probabilistic forecasts of ceiling and visibility for that airport. The results indicated that such a system has greater skill than model output statistics (MOS) derived forecasts as well as persistence climatology out to a lead time of 6h. Leyton and Fritsch (2003, 2004) extended this work by demonstrating a further increase in skill when utilizing higher density and higher frequency surface observations on these short-term forecasts.

While observations-based statistical systems have shown improved forecast skill, little has been done to investigate whether the economic value of the forecast has been improved as well. Keith (2003) demonstrated how human subjective forecasts of the probability of the weather being below alternate minimum should increase the value to airlines compared to the traditional categorical aerodrome forecasts. The alternate minimum is that level of visibility and cloud base which requires an aircraft to carry extra fuel, which, in turn, incurs a cost.

Considering the promising results found by these recent studies, it was of interest to combine these two concepts and investigate whether observations-based forecasts can further increase the value of the forecast to aviation interests. The concepts behind these two studies are synergistic, as the lead times utilized in the observations-based forecasts coincide with the lead times utilized for the planning and flight period of the majority of domestic flights around the world. It is important to note that human forecasters produce aerodrome forecasts that generally cannot match even raw persistence out to between 3 to 6 hours ahead (depending on how this forecast skill is measured).

This study focused on probabilistic forecasts of low ceiling and/or reduced visibility and the corresponding impact on forecast value for flights arriving at three major domestic hubs in the United States. As a basis of comparison, National Weather Service Terminal Aerodrome Forecasts (TAFs) that correspond to the developed statistical forecasts have been obtained for each location. In addition, costs for several flights arriving at each airport have been provided by a major domestic carrier in the United States. These have been utilized to determine the optimal forecast probability above which extra fuel should be carried. The combination of improved short-term forecasts and identification of optimal forecast probabilities will lead to greater forecast value, potentially saving the aviation industry millions of dollars per year.

Our results indicate that by simply transitioning from a traditional, categorical forecasting system to one that produces probabilistic forecasts could save a major carrier $50 million in operating costs annually.