Use Data Assimilation to Reduce Uncertainties in Ensemble Forecasting for Strategic Traffic Flow Management

Ensemble Forecasting (EF) is indispensable in dealing with unavoidable uncertainties in Numerical Weather Prediction (NWP). Nevertheless, in Strategic Air Traffic Flow Management (TFM), the dispersion among the ensemble members presents an obvious challenge in decision making, especially when uncertainty is significant and the planning horizon is relatively far.  With the help of real-time observation data, Data Assimilation (DA) algorithms not only can improve the prediction accuracy of individual ensemble members, but are also able to help narrow down the dispersion range of available EF results. Our implementation, based on the sliding-window Bayesian Model Averaging (BMA) approach, has demonstrated the capability to consistently identify one to two most matching predictions (validated by the actual observations in hindsight analysis) out of the 21 ensemble members of the Short Range Ensemble Forecasting (SREF) within a 2 to 6-hour prediction window. Moreover, the auto bias-correction function has successfully corrected “group” errors with the help of real-time observation data. The resulting algorithm has been proven to be effective and robust, and is being evaluation as part of a proposed strategic TFM decision support system to be deployed in the Federal Aviation Administration’s Traffic Flow Management System.