Decision Support Tool for Predicting Aircraft Arrival Rates, Ground Delay Programs, and Airport Delays from Weather Forecasts

The principle “bottlenecks” of the air traffic control system are the major commercial airports [1]. Atlanta, Detroit, St. Louis, Minneapolis, Newark, Philadelphia, and LaGuardia all expect to be at least 98% capacity by 2012 [2]. Due to their cost and the environmental and noise issues associated with construction, it is unlikely that any new airports will be built in the near future. Therefore to make the National Airspace System run more efficiently, techniques to more effectively use the limited airport capacity must be developed. Air Traffic Management has always been a tactical exercise, with decisions being made to counter near term problems [3]. Since decisions are made quickly, limited time is available to plan out alternate options that may better alleviate arrival flow problems at airports. Extra time means nothing when there is no way to anticipate future operations, therefore predictive tools are required to provide advance notice of future air traffic delays. This research describes how to use Support Vector Machines (SVM) to predict future airport capacity. The Terminal Aerodrome Forecast (TAF) is used as an independent variable within the SVM to predict Aircraft Arrival Rates (AAR) which depict airport capacity. Within a decision support tool, the AAR can be derived to determine Ground Delay Program (GDP) program rate and duration and passenger delay. The research compares the SVM to other classification methods and confirms that it is an effective way to predict airport capacity. Real world examples are included to highlight the usefulness of this research to airlines, air traffic managers, and the flying consumer. New strategies to minimize the effect of weather on arrival flow are developed and current techniques are discussed and integrated into the process. The introduction of this decision support tool will expand the amount of time available to make decisions and move resources to implement plans. I. PROBLEM STATEMENT Air traffic congestion has become a widespread phenomenon in the United States. The principle bottlenecks of the air traffic control system are the major commercial airports, of which at least a dozen currently operate near or above their point of saturation under even moderately adverse weather conditions [1]. The Macroscopic Capacity Model (MCM) analyzed 16 airports within a 1000 nmi. triangle from Boston, Massachusettes, to Minneapolis, Minnesota, to Tallahassee, Florida. Based on this analysis, the MCM showed that in 1997 these airports were operating at 74% of maximum capacity. The model further went on to predict that the these airports will be at 89% capacity by 2012 [2]. The congestion problem is made worse because most airline schedules are optimized without any consideration for unexpected irregularities. When irregularities occur, the primary goal of the airlines is to get back to the original schedule as soon as possible, while minimizing flight cancellations and delays [4]. When trying to get back on schedule, sometimes it is the complexity of the situation, coupled with time pressure, which results in results in quick decisions that may be less than optimal [5]. Therefore, it would be advantageous to develop techniques to lessen the complexity of the situation and increase the time available. One way to increase the time available is to create a tool that can predict the impact of weather on future inbound flight operations. Weather reports such as the TAF, Aviation Routine Weather Report (METAR), and the Collaborative Convective Forecast Product (CCFP) all provide raw weather forecast information. None of these forecasts though inform National Airspace System (NAS) stakeholders what the effect of that weather will be on flight operations. This research intends to fill this void by developing a process from which a forecast can be entered to produce estimate of the delay and capacity of the airport within the forecast area. Capacity estimates, in the form of AARs are produced for four time periods of the operational day. Ground Delay Program estimates of duration and program AARs along with expected delays can be derived from the predicted AARs. Now the forecast will not only provide the winds and ceiling, but also the AARs, GDPs, and expected