A Comparison of Different Approaches to Estimate Unimpeded Flight Times

摘要

Queuing models of the National Airspace System require demand information about when aircraft intend to land. Empirical data typically re°ect when °ights actually landed. Published schedules encapsulate delay expectations and actual landing times encapsulate actual delays, so neither is directly applicable. Instead, one might infer nominal arrival times from actual upstream departure times and unimpeded travel times. This paper presents di?erent ways to estimate unimpeded °ight times. These include summary statistical methods such as observing speciˉc percentiles from ac- tual °ight time distributions, generated from a number of possible data sources. We also use simulation methods, relying on the Future ATM Concepts Evaluation Tool (FACET) to model system-wide airspace operations over the United States. Since this model does not consider interactions between aircraft more than superˉcially, one of its outputs is essentially unimpeded °ight times. Finally we use a statistical estimation approach. The underlying idea is that the observed °ight time is a mixture of two unobserved distributions-the unimpeded °ight time and the delay. The statistical approach estimates these unobserved distributions, taking into account the quarterly periodicity of the data. Finally a graphical and statistical comparison of all approaches is performed and results are discussed.