Issues in modeling a national network of airports

The modeling of networks consisting of a large number of geographically dispersed airports is considered. The need for models of this type has become clear recently as a result of growing systemwide congestion of air traffic, the propagation of delays from one airport to others, and the desire, at the national policy level, to intelligently allocate scarce federal resources among competing alternatives and local airport projects. Some fundamental difficulties associated with network models of this type are examined. Important issues include problem size and data requirements; the probabilistic and dynamic nature of the airport system's demand and capacity; the combinatorially explosive number of possible network states and the resulting need for careful statistical sampling and analysis; the sensitivity of computational performance to the level of detail in the network model; the difficulty of preparing demand scenarios that predict future connections between pairs of airports; and user requirements for model robustness, portability, and transparency. These issues are discussed and illustrated in some detail, including references to specific existing network models of the ATC (air traffic control) system.<
>