Dynamic wheels-up time predictions

The study described in this paper addressed cases when FAA´s Enhanced Traffic Management System (ETMS) had experienced substantial errors in its departure time predictions. At some airports on some days, the ldquoerror wavesrdquo had scores of minutes in magnitude for the majority of flights and lasted for hours. One of the key ETMS functions is to predict the future demand for elements of the National Airspace System (NAS) - airports, sectors, and fixes. The main source of ETMS errors in demand prediction is directly attributed to departure time errors. Departure times enter the system with flight plans filed by NAS users. This study evaluated archived historical data of ETMS operations. Analysis of one month of data yielded an algorithm that mitigated departure time errors in cases of significant airport delays. The core of the algorithm consisted of: middot heuristic filtering criteria that told when a correction had to be applied to departure time estimates; middot the technique of improving the estimates for the cases that passed the criteria. The algorithm was validated using 5+ years of archived data for 34 pacing airports and showed substantial improvement in departure time estimates. For every month of data, the filtering criteria had selected between 2%-8% of wheels-up time estimate cases. For these cases, monthly average estimates improved by 5-10 minutes (about 20%). The improvements were especially good for cases when lead time to departure was smaller than 15 minutes: of these 35% were impacted with the average improvement of about 11 minutes. The good and stable performance of the algorithm provides a reasonable expectation for improved departure time predictions if the algorithm is deployed within the context of operational ETMS.