1. Figure 7 - METAR Wind Speed Profile profile is illustrated in Figure 7, where the surface winds increase exponentially up to the altitude of 2000 feet, above which the wind remains constant.
2. The constant dilemma in the choice of modeling parameters is a sensible balance between two criteria, the parameter's ability to capture the behavior of the system realistically and the simplicity of the resulting model. There are over a thousand aircraft type identifiers defined by the FAA that would need to be considered in a model parameterized by aircraft type. By contrast, a trajectory model parameterized by weight class would require less aircraft-specific data, as there are currently only five weight classes specified by the FAA (Small, Large, 757, Heavy, and Super). For this reason, the effect of simplifying aircraft type and weight class modeling on final approach trajectory prediction accuracy for the 737-700 aircraft type was analyzed.
3. Aweight class based, Fourier transform true airspeed model was created using a subset of the aircraft from the modeling data set described earlier. This subset consisted of all Large weight class aircraft (1930 aircraft), not just a single aircraft type as was the case for the 737-700 model. An "all aircraft types" (AAT) model was also created for this analysis that included 3349 aircraft of all weight classes. Once these models were created, final approach trajectories for the 737-700 aircraft from the analysis data set were calculated with each of the three Fourier transform true airspeed models (i.e., 737-700 type, Large weight class, and AAT).
4. This section presents three categories of analysis for final approach trajectories derived from the various speed profile models described earlier. The first analysis category measured the accuracy of the resulting final approach trajectory for each of the speed profile modeling methods. The second analysis category evaluated the effect of key modeling parameters such as wind on final approach trajectory accuracy. The final analysis category applied operational relevant metrics such as inter-arrival separation and landing time prediction error. Aircraft from an independent set of sixteen days (mid-February to March 2010) of KLAX final approach data, different from that used to develop the models, were used for the analysis. A. Final Approach Speed Profile Method Analysis
5. The success of each of the three methods for inferring final approach speed profile on final approach trajectory prediction accuracy was evaluated with respect to the groundspeed based variant of each method. As with the modeling data set, the analysis data set was limited to 737-700 series straight in arrivals resulting in a total of 457 aircraft. For each aircraft, a series of final approach trajectories were calculated using initial conditions at several locations along the final approach route starting at 14 nmi from the runway and ending at 2 nmi from the runway. For each trajectory prediction made for a specified initial condition (e.g., 14 nmi from the runway), groundspeed and path distance error were calculated by holding the predicted values constant and subtracting the actual measured aircraft track values. The mean and standard deviation of the groundspeed error and path distance error for all samples were then calculated. Figure 8 shows the resulting groundspeed and path distance errors for the dead reckoning method. Each star indicates the initial condition for the corresponding trajectory prediction. For the series of dead reckoning predictions initiated for aircraft 14 nmi from the runway, mean groundspeed error increases from zero to approximately 40 knots faster than actual when the aircraft were 6 nmi from the runway, then to approximately 105 knots faster by the time the actual aircraft reach the runway. The corresponding mean path distance errors were negative, indicating the dead reckoning method on average predicts the aircraft to be closer to the runway than was actually the case.