The 8th Conference on Aviation, Range, and Aerospace Meteorology


Dale A. Rhoda, MIT Lincoln Lab, Lexington, MA; and M. L. Pawlak

The Thunderstorm Penetration/Deviation Decision in the Terminal Area*+

Dale A. Rhoda and Margita L. Pawlak
Massachusetts Institute of Technology
Lincoln Laboratory
Lexington, Massachusetts 02420-9185

Anticipating air carrier aircraft deviations around storms is important for effective terminal air traffic management. For example, automation systems such as CTAS and URET, that depend on making accurate time-to-fly estimates, will need to know which terminal air routes will effectively be closed by convective weather during the control period. The study described in this paper models the penetration and deviation behavior of air carrier aircraft flying near actual storms in the Dallas/Fort Worth airspace as a function of aircraft and thunderstorm characteristics. The objective is to determine which variables either influence the penetration/deviation decision directly or serve as proxies for variables that influence that decision.

Sixty hours of weather and flight track data were analyzed from nine storm days in the summer of 1997 at Dallas/Fort Worth International Airport (DFW). Weather data were collected with the MIT Lincoln Laboratory prototype of the Integrated Terminal Weather System (ITWS). Weather sensors included the National Lightning Detection Network (NLDN), three fan-beam Airport Surveillance Radars (ASR-9s), two pencil-beam Terminal Doppler Weather Radars (TDWRs), and one pencil-beam WSR-88D (NEXRAD). Flight track data were collected from an ASR-9 located on the airport.

The unit of analysis in the study was each aircraft encounter with a storm cell in the plane's intended flight path. A combination of software and human analysis were used to reduce the weather and flight data to one value of each weather variable per storm cell encounter. The data set was divided into encounters where aircraft penetrated storm cells and encounters where aircraft deviated from their intended flight paths to avoid part or all of the storms.

The data set was split randomly into a training set and a testing set and analyzed using a pattern classification approach. Several statistical classifiers were trained and tested and were able to correctly classify more than 85% of the encounters in the test dataset. The variables that contributed to the most successful classifiers were: range of the encounter from the airport, pencil-beam radar reflectivity, vertically integrated liquid water, percent of the region covered with light precipitation, and percent of the region covered with heavy precipitation. The classifier performance bodes well for plans to develop an algorithm that will predict the probability that pilots will deviate around weather.

In addition to the weather-related classifier results, several hypothesis tests involving flight-related variables yielded statistically significant results. We found that aircraft were more likely to penetrate heavy weather if:

a. They were near the destination airport,
b. They were following another aircraft,
c. Their flying time to the destination TRACON was more than 15 minutes longer than the scheduled flying time to the TRACON (i.e., they were running late), or
d. The time of day was more than one hour after twilight.

Given what we know about wind shear accidents, it is disturbing that a significant number of aircraft penetrated heavy weather near the airport. Aircraft that are preparing to land are in a critical phase of flight with very little airspeed or altitude to spare. The paper will discuss a number of considerations regarding this result. In particular, it will be shown that the cost of deviating around weather, namely longer flight path length inside the TRACON that implies more fuel spent and a later arrival time, is much higher when the deviation occurs near the airport rather than farther away. The pilots' penetration behavior is quite possibly a rational choice based on their estimates of the relative costs of penetrating and deviating.

* This work was sponsored by the National Aeronautics and Space Administration. The views expressed are those of the authors and do not reflect the official policy or position of the U.S. Government.

+ Opinions, interpretations, conclusions and recommendations are those of the authors and are not necessarily endorsed by the United States Air Force

The 8th Conference on Aviation, Range, and Aerospace Meteorology