Assessment of imperfect weather forecasts on airline and passenger planning

Abstract

This paper studies the impacts on aviation and related costs of a snow storm in the northeastern United States that occurred on October 29, 2011, focusing on the question of what would have happened had an accurate forecast been available to airlines and the traveling public at least twenty-four hours before the actual event. Snow storms are remarkably difficult to forecast. In the case of the October 29th 2011 event, snow fell over a larger area than initially forecast, the event moved through the Northeast faster than predicted, and the storm occurred in the middle of the day, maximizing its impact on air travel. A mitigating factor is that the event occurred on a Saturday, a day less traveled than other days of the week. The unexpected nature of the storm caused twenty-nine flights to be diverted to Bradley International Airport (BDL) near Hartford, Connecticut. One diverted flight remained at BDL for over seven hours before continuing to its intended destination. Additionally, all major Northeast airports had reduced capacity, as some runways closed for snow removal operations while others operated under Instrument Meteorological Conditions (IMC) for an extended time. The impact was larger than the aviation system was prepared to handle in part due to the fact that the weather forecasts were accurate only within two hours of the storm's onset in the Northeast. The question we seek to answer is the extent to which the impact can be mitigated had an accurate forecast been available twenty-four hours in advance. To study the question, we configured NASA's Airspace Concepts Evaluation System (ACES) to replicate the day with a two-hour forecast, and a second time with a hypothetical twenty-four hour perfect forecast. The forecasts as well as actual conditions not only in the Northeast, but throughout the National Airspace System (NAS), were provided by the Air Force Environmental Data Cube Support System group [1]. The ACES agents involved in airline decision making, particularly the Airline Operations Center (AOC) agent, contain logic that mimics an airline's decision making logic (cancellation, flight delay, or diversion) given a forecast of future conditions. By providing the AOC agent a two-hour perfect forecast and a twenty-four hour perfect forecast, the flight impact can be estimated. We monetize the flight impact in two ways, one from the airlines' perspective and another from the passengers' perspective. The airline monetization is achieved using standard cost of airline operations as derived from other studies [2]. Passenger monetization is computed using a study of the impact of flight delays and cancellations on passenger costs [3]. The costs are compared, and a classic tradeoff between airline and passenger costs associated with imperfect weather forecasts is observed.