The problem of generating representative weather-impact scenarios can be viewed as requiring two advances. First, a tool is required for generating an ensemble of stochastic scenarios of weather impact (on both en route and terminal-area capacities) over a strategic time horizon. Second, a technique is needed for selecting representative scenarios from the ensemble that faithfully capture the possible scope of system-wide weather impact and allow the FCM automation (or human operators) to enumerate management contingencies. Recently, several approaches have been posited for the first task, i.e. the generation of stochastic weather-impact scenarios. Specifically, we have introduced a tool for rapidly generating a large ensemble of en-route weather- weather-impact scenarios, whose statistics match available probabilistic forecasts at snapshot times. Alternate approaches that directly translate ensemble weather forecasts to probabilistic weather-impact maps or weather-impact trajectories are also under development. Several approaches for stochastic prediction of airport arrival/departure rate futures have also been introduced, and are under development. In this article, we will review and pursue a preliminary comparison of the available tools for weather-impact scenario generation. We will also highlight remaining needs toward a comprehensive strategic weather-impact scenario-generation tool, including integration of terminal-area and en-route weather and validation of developed tools.
The second task, namely selecting representative weather-impact scenarios from the ensemble, requires further study. The selection process first requires defining metrics for valuation and comparison of weather-impact scenarios that together allow 1) differentiation of scenarios requiring different types of management initiatives and 2) characterization of system-wide impact of the weather scenarios. This problem of selecting metrics is complex, because it requires understanding the relationship of weather-impact futures and system-wide performance and management design, and yet simulation of NAS performance and design of management initiatives for each scenario is infeasible. The challenge, then, is to identify a small number of metrics that allow comparison and valuation of scenarios, based on historical knowledge, sensitivity-analysis of traffic flows, and/or theoretical analyses of weather-impact on traffic. Once the metrics have been selected, a technique is needed for clustering the large ensemble of scenarios according to the metrics, and selecting a representative example in each cluster. Two preliminary techniques for clustering scenarios and selecting representative ones have been introduced in our previous work, but these techniques must be enhanced for use with multiple, realistic selection metrics. In this article, we will work toward a comprehensive solution for the representative-scenario-selection problem, by exploring strategies for metric selection and pursuing enhancement the techniques for scenario clustering and selection.