UC Berkeley

This dissertation investigates predictability in strategic air traffic management with a focus on ground delay programs (GDPs). Through a survey of flight operators, we confirm the proposition that flight operators care about predictability. We then develop models that incorporate predictability into GDP cost optimization after failing in finding an existing model that can serve this purpose. This is accomplished by modifying traditional GDP delay cost functions so that they incorporate predictability, and determining the sensitivities of the optimal planned capacity recovery time and associated cost to the unpredictability premiums included in the cost functions. To do this, we develop two stochastic GDP models: a GDP no-revision, or static, model; and a GDP revision, or dynamic, model considering one GDP revision. GDP scope, which matters only in the revision model, is also considered. The optimization results from the case study show that the cost of unpredictability clearly matters, particularly in the more realistic case where GDP revision is allowed. Of the two unpredictability cost parameters, the one for unplanned delay has a stronger impact than the one for planned un-incurred delay. The insights from this analysis might eventually be used to develop a decision support tool that air traffic managers could use in determining what the planned end time should be for a GDP in a manner that reflects the importance of predictability to flight operators.