10.9
Using probabilistic scenarios-based weather event forecasts for decision-based TFM

- Indicates paper has been withdrawn from meeting
- Indicates an Award Winner
Thursday, 2 February 2006: 4:00 PM
Using probabilistic scenarios-based weather event forecasts for decision-based TFM
A301 (Georgia World Congress Center)
Goli Davidson, Metron Aviation, Inc., Herndon, VA; and D. J. Krozel, C. Mueller, and W. Chan --Pending

The uncertain nature of strategic 2-6 hr weather predictions makes it difficult for traffic managers (TMs) and Airline Operation Centers (AOCs) to adequately plan traffic flow management (TFM) strategies for weather related disruptions. We cannot wait for dramatic improvements in weather prediction capabilities to provide the precision and accuracy needed to plan around weather events with sufficient lead time to avoid these disruptions. Stochastic weather forecasts and stochastic TFM planning need to be developed to complement each other.

In and of themselves, probabilistic weather forecasts may be of little value to traffic flow planners. Experience with today's consensus forecasts shows a strong tendency of planners to treat uncertain information as deterministic, even when confidence levels have been provided. No scientific methods have been adopted, as of yet, for formally incorporating uncertainty into TFM planning; TFM relies heavily on experience and intuition of traffic managers.

Collaborative research between meteorologists and TFM experts is needed to drive out the salient features that a probabilistic weather forecast should have. Meteorologists are well acquainted with some of the standard techniques for measuring this; however, this is not the same as given probabilities of outcomes, especially when costs are to be taken into account. We have determined that potential future weather scenarios affecting the National Airspace system (NAS) should be expressed in terms of a small, finite number traffic flow possibilities, each with an associated probability.

We use an Air Traffic Management (ATM)-Weather Probabilistic Decision Tree to model the problem of TFM decision making under uncertain, strategic weather conditions. We start by building an underlying event tree to represent a set of discrete potential (convective) weather outcomes, while associating a probability of occurrence with each one, then, assign potential decisions based on these events. This framework of the TFM decision-making process creates a map of current formal and informal weather avoidance practices to temporal and spatial TFM decision points.

Preliminary research shows that probabilistic planning provides benefits over strategies that plan for the most likely scenario, or those that wait-and-see what weather will materialize before executing TFM control actions. We conclude that multiple scenarios should be considered when planning TFM initiatives, where the full spectrum of potential future outcomes is distilled into a more manageable number of outcomes. This not only reduces the size of the underlying search space for algorithmic approaches, but also allows decision-makers to discuss the pros and cons of various strategies. A consensus forecast could be formed and disseminated much in the fashion of today's collaborative forecast, except with alternative future weather scenarios. This will admit decision support tools to aid in the decision-making process. Flow routing and ground holding strategies can be suggested, using techniques already established in the literature.