Quantifying the Benefit of Temporal Correlation in Probabilistic Weather Forecasts for Aviation Traffic Management

A major goal of the Next Generation Air Transportation System (NextGen) is to increase the use of automated decision support tools (DSTs) to proactively manage air traffic congestion. In particular, these tools must recommend efficient traffic management solutions that are robust to uncertainty in weather forecasts over a horizon of multiple hours. The correlation of weather outcomes over time is an important factor in determining the optimal traffic flow management (TFM) response to a weather event. Advanced TFM DSTs will require probabilistic forecast products that account not only for uncertainty within discrete time intervals but also for these temporal correlations across intervals. In this paper, we demonstrate the importance of incorporating temporal correlation into TFM DSTs by using forecasts with and without these correlations in the ground delay program (GDP) planning problem. GDPs are used to limit arrivals to a capacity-constrained airport by assigning delays to flights at their departure airports. The resulting ground delays are less expensive to flight operators than airborne holding and reduce or eliminate the likelihood of flight diversions. We simulate a GDP DST with a simple linear programming (LP) formulation first described by Ball, Hoffman, Rifkin, and Odoni. We apply this LP using probabilistic forecasts with and without temporal correlation to scenarios at multiple airports. A comparison of the performance – in terms of ground delay issued and expected airborne holding – of the optimal GDP parameters derived from the two LP solutions provides a quantitative estimate of the value of including temporal correlations in weather forecasts for TFM DSTs. The implications of these results for future aviation weather research are discussed.