Are air traffic models chaotic?

Network queueing models, including some models of air traffic, exhibit sensitivity to very small changes in the times of events that are nominally simultaneous, such as flight departures (pushbacks). Such models might exhibit chaotic behavior when the ordering of otherwise simultaneous (or closely spaced) departures are changed. This paper explores the question as to whether such models do, indeed, exhibit chaotic behavior, or whether the effect is bounded. This question is explored using an empirical study of delays in National Airspace System (NAS) simulations. The study consisted of 32 replications of a NAS scenario, in which scheduled departures were given a small pseudorandom pushback delay, independent across replications and across airports. The simulated scenario was a hypothetical one configured for a generally good weather day. The standard deviation of total (viz., 59-airport) at-gate delays is reported; it exceeds the simulated reduction in delay that would be produced by increasing arrival capacity by ten percent at any airport, except for the five most-congested airports. We further analyze the results to determine whether chaotic effects are manifest in the mathematical model