Commercial aircraft encounters with thunderstorms in the Memphis terminal airspace
Thunderstorms are dynamic obstacles to the flow of air traffic. Aircraft routing in the presence of thunderstorms is as dynamic as the position and intensity of the storms. The question of where pilots will and will not fly is relevant to the decisions made by human air traffic managers as well as to the development of automated decision aid tools. In order to accurately anticipate which routes will be useable one needs to be able to 1) forecast the relevant weather variables, and 2) convert those weather variables into a quantitative probability that pilots will request deviations from the nominal route. The Convective Weather Integrated Product Team at the FAA is improving the accuracy and lead time of forecasts of thunderstorm products. This paper provides an update on our examination of the issue of probability of deviation. In our recent examination of 63 hours of weather and flight track data from the DFW airspace (Rhoda and Pawlak, 1999a,b) we combined several weather variables (measurements, not forecasts) to correctly predict pilot deviation and penetration behavior for 70-85% of the encounters between thunderstorms and aircraft arriving at DW and Dallas Love (DAL) airports. We also found that pilots were more likely to penetrate strong precipitation when they: 1) were near the arrival airport, 2) were following another aircraft, 3) were flying after dark, 4) had been delayed in the air by 15+ minutes upstream of the DFW airspace. We did not find any statistically significant difference between the percentages of thunderstorm penetrations by various airlines. We also found that persistent penetration of storms near the airport is sometimes abruptly interrupted presumably by wind shear alerts from air traffic controllers or cautionary pilot reports from the penetrating aircraft. When the arrivals cease, aircraft on the final approach course may turn suddenly to the left or right to avoid the weather that caused the interruption. Aircraft that abort the approach sometimes fly through very intense precipitation-sometimes through downdrafts that are causing microburst outflows at the surface. The work described in this paper applies the methodology from the DFW study to data collected in the Memphis Terminal Radar Approach Control (TRACON). The methodology is described briefly here and in more detail in (Rhoda and Pawlak, 1999b). We developed several probability of deviation classifiers using a portion of the Memphis data and tested them on the remaining data to determine if it is possible to predict whether pilots will penetrate or deviate around the storms. We also tested the classifiers that were developed in the DNV study on the MEM data and vice versa. We repeated the DFW hypothesis tests for various dichotomies of encounters: near/far, leading/following, light/dark, delayed/undelayed.