Feasibility of using historical flight track data to nowcast unstable approaches

The approach and landing phase is one of the most complex procedures in airline operations. To mitigate the potential risks, airlines have established “stabilized approach” criteria that require the flight crew to check four conditions at 1000' above ground level (AGL) and 500' AGL. If any of these conditions are not satisfied, the flight crew is required to abort the approach adding operational costs to the flight. Nowcasting unstable approaches prior to the stable approach altitudes (e.g. at 10 nm, 6 nm from the runway threshold) could provide lead time for flight crew to make adjustments to avoid a potential unstable approach. Kinematic models, already used in the Flight Management System (FMS) to predict future aircraft state are not practical as these models cannot account for events that will occur during flight progress (e.g. flap/slat and extension, ATC clearances ...). This paper describes an analysis of massive amounts of surveillance track data for a given approach to assess the feasibility of using of historical flight track data to nowcast the likelihood of a stable approach given the situation at 10 nm, 6 nm, and 3 nm from the runway threshold. A logistic regression model was able to nowcast the 1000' AGL aircraft state with accuracy of 61.7% at 10 nm, 73.6% at 6 nm, and 83.1% at 3 nm. Additional measures of prediction performances are summarized. The implications and limitations of this method are discussed.