Safety Pilot: Overrun!

Landmark Accidents: Southwest slides off the runway

December 1, 2008

Bruce Landsberg is the executive director of the AOPA Air Safety Foundation.

It was three years ago this month that a Southwest Airlines (SWA) Boeing 737-700 slid off the end of Runway 31C after landing at Chicago Midway International Airport (MDW) on a snowy evening. Tragically, a child’s life was lost as his parents’ car was in the wrong place at the wrong time. The Boeing rolled through blast and airport perimeter fences before stopping on the roadway with the car underneath. There was one serious injury and several minor injuries in the car.

How could a diligent professional crew flying for one of the world’s most successful airlines in one of the most reliable aircraft find themselves sliding off the end of a runway that they had landed on many times before, at an airport that has seen hundreds of thousands of landings without mishap? Sometimes, situations that seem simple aren’t. Arcane information involving braking action reports, confusing flight operation policies, inconsistent computer programming, and airport environment all played a part in this mishap. There’s plenty to ponder, regardless of what size aircraft you fly.


Southwest Flight 1248 departed Baltimore/Washington International Thurgood Marshall Airport (BWI) on December 8, 2005, about two hours late because of deteriorating weather in Chicago. The six to nine inches of snow that was predicted for Chicago started at Midway just before 2 p.m. An amendment to the forecast surface winds revising “calm to 090 at 11 knots” and runway braking action from “wet-good” to “wet-fair” was sent to the flight en route. The winds were accurate, but the braking action report turned out to be wrong.

At 6:53 p.m., about 20 minutes before the accident, the weather was reported as wind 100 degrees at 11 knots, visibility one-half statute mile in moderate snow, and freezing fog. The ceiling was broken at 400 feet and overcast at 1,400 feet.


The runway was plowed one-half hour before the accident with an average runway friction reading of 0.67 (out of a possible 1.00, which indicates perfect conditions). After the accident a second test revealed the friction at 0.40. Aeronautical Information Manual guidance states that, with values below 0.40, “aircraft braking action begins to deteriorate and directional control begins to be less responsive.” (For more information see the AOPA Air Safety Foundation Safety Brief on braking action reports.)

Braking action advisories were in effect, which require pilots and controllers to provide each other with the latest updates on aircraft stopping performance.

Why did Midway continue to operate on Runway 31 with a significant tailwind? Changing runways might have had a negative impact on operations at Chicago O’Hare International, 13 nm to the northwest, so the air traffic considerations must be weighed against the safety implications.

The crew

The captain, age 59, a retired Air Force pilot, had 15,000 total flight hours with 4,500 hours in the Boeing 737. He had flown for SWA since August 1995 and had an exemplary flight record. He told investigators he had slept well and was not fatigued. He also noted in post-accident interviews that the weather was “the worst” he had experienced, but he had “encountered similar conditions about a dozen times” in his tenure at SWA and “expected to be able to land safely.”

The 34-year-old first officer had flown as a regional airline pilot before joining SWA in 2003. He had about 8,500 hours total time with 2,000 in the 737. He claimed extensive winter weather flying as a regional airline captain.

The flight

It was the captain’s leg to fly, and during the flight the crew discussed braking action reports, required landing distances, use of autobrakes, and possible diversion to an alternate airport. A confounding factor was the mixed braking-action reports. Braking was generally reported “good to fair” on the first half of the runway and “poor” on the second half. The tailwind component was eight knots, and the crew had some discussion on diversions to alternate airports in the Midwest.

Everything was routine, despite some holding that delayed all flights going into MDW as the runway was plowed. The most recent report, about 30 minutes before the accident, showed “Runway 31C—trace to 1/16th of an inch of wet snow over 90 percent of the surface, 10 percent clear and wet.”

The landing was unremarkable as the flight data recorder (FDR) verified main gear touchdown at 1,250 feet from the runway threshold (well within the recommended touchdown zone) at 124 knots airspeed, at target airspeed with a ground speed of 131 knots. Ground spoilers and autobrakes deployed immediately, as designed.

The captain had difficulty applying reverse thrust, which was essential for this difficult landing. As soon as the first officer noticed that the reversers were not engaged, he deployed them. According to the NTSB report, “The first indication of thrust reverser deployment was recorded at 15 seconds after touchdown with full deployment at 18 seconds.” The four airliners that landed in the 20 minutes preceding the accident deployed reverse thrust within four to six seconds of touchdown according to their FDRs. Considering that none of them probably should have landed under these conditions, there is something of a herd mentality that is admittedly very difficult to go against. We’ve all done it. How do you explain to the passengers that you diverted when all the other flights were getting in? It doesn’t mean it was right, it just means that they were lucky.

The autobrakes on SWA Flight 1248 deactivated 12 seconds after touchdown as the pilot manually applied the brakes. Even with maximum braking and maximum reverse thrust, the nosewheel departed the runway overrun at 53 knots. The aircraft stopped 500 feet beyond the runway end on the public road, after sliding through two fences. The passengers and crew evacuated the aircraft with a few minor injuries.

The computer

SWA equips all of its aircraft with an on-board performance laptop computer (OPC) to assist the crew in a variety of calculations, including landing performance and stopping margins. Pilots enter pertinent data such as landing runway, prevailing wind conditions, aircraft weight, ambient temperature, and reported braking conditions. Up to 10 knots of tailwind is allowed with good or fair braking action, but only five knots is permissible under poor conditions.

The OPC computed margins of 560 feet and 40 feet based on fair and poor braking action, respectively. A 40-foot margin in this business is no margin at all, but company guidance was that any positive number was acceptable. Additionally, the computer did not clearly indicate that five knots was the maximum tailwind used in its calculation. It accepted an eight-knot entry but did not compute the distances based on the actual eight-knot tailwind. Had the OPC computed using the actual tailwind entry, the calculation would have shown a 260-foot overrun.

There was another “gotcha” in the computer programming. For the 737-700, stopping distances were based on the use of reverse thrust—but not on the 737-300 and -500, so a crew might extrapolate that with reverse thrust the aircraft would stop several hundred feet shorter than shown on the OPC. This important dissimilarity was not clearly explained in differences training. The accident crew assumed, as did many other SWA crews interviewed after the accident, that the OPC computed the same way for all models. Pilots were required to remember which formula applied to which aircraft, and they might fly all three models in the course of a day. Allowing reverse thrust to be credited with decreased landing distance, while economically and operationally expedient, does lessen the margins.

Procedural confusion

SWA requirements were for crews to use the most adverse braking-action report for planning purposes and not depend on a “blended” report. Thus, any part of the runway that was reported as poor would override a good/fair report on other parts. Post-accident interviews showed that a number of SWA crews did not understand that. With an eight-knot tailwind and a poor braking-action report for any part of the runway, SWA Flight 1248, and the ones preceding it, should have diverted. A five-knot tailwind is the maximum allowed with poor braking.

The Midway tower controller did not provide all the braking action reports to SWA Flight 1248, as required by FAA procedure. A Gulfstream III that landed shortly before the Southwest flight reported “poor” conditions, but the tower did not pass that along. While there might be some validity in SWA guidance to only accept braking action reports from other “commercial” aircraft, this might have alerted the crew that conditions were deteriorating.

SWA did not allow crews to use autobrakes, but was implementing a change to standardize use across the fleet as the system became available on all SWA aircraft. Crews interviewed and tested on the procedure afterward showed some confusion and uncertainty. The change was not to take effect until four days after the accident, but the crew had read about procedural changes before the flight and was unsure on the use of autobrakes. Distraction was clearly a factor.

Midway Airport

At the time of the accident, Midway was not optimally configured for large airline jet operations. Runway 31C was the only approach available because of visibility requirements, according to the NTSB. The runway had a displaced threshold and usable landing distance of 5,826 feet. As early as September 2000 the FAA had determined that the runway safety areas (RSA) were not in compliance with an airport design advisory circular, which considers a 1,000-foot overrun as standard. Runway 31C’s RSA was a mere 81 feet.

In 2003, the FAA again asked for a reassessment, but the city of Chicago stated that there were “no alternatives.” Their contentions: The runway could not be shortened and still allow air carrier operations; there was no space to extend the RSA beyond the existing airport perimeter without a major impact to the surrounding community. The use of an engineered materials arresting system (EMAS—a soft surface where an overrunning aircraft sinks in and stops with minimal damage) was not feasible since it would encroach on either the runway or extend beyond the airport perimeter. After-accident simulations showed that a modified EMAS would have stopped SWA Flight 1248 on the airport property. It was installed on several runways shortly afterward.

Braking-action reports

The NTSB had “long been concerned about runway surface condition assessments.” There is considerable variability between pilot braking-action reports, airport contaminant type and depth observations, and ground surface vehicle friction measurements. There may be little or no correlation between the three, and there is no standard for how a particular aircraft will perform under given conditions. It almost amounts to “Y’all be careful out there.”

Air carriers are required to perform landing distance calculations before departure to determine if the aircraft can land safely. However, the NTSB noted, “The assessments do not attempt to comprehensively account for actual conditions.” There are many variables to consider and much subjectivity. While the FAA requires operators to take the aircraft manufacturers’ maximum performance data and add margins of 62 percent for dry runways and 92 percent for contaminated runways, which may not be sufficient for all conditions.

Probable cause

The NTSB determined the probable cause “was the pilots’ failure to use available reverse thrust in a timely manner to safely slow or stop the airplane after landing. This failure occurred because the pilots’ first experience and lack of familiarity with the airplane’s autobrake system distracted them from thrust reverser usage during the challenging landing.” Numerous contributing factors included SWA’s failure to provide clear guidance on landing distance calculations, programming of the OPC, implementation of the new autobrake procedure without appropriate familiarization, failure to include appropriate margins of safety to account for operational uncertainties, the pilots’ failure to divert to an alternate, and the lack of an EMAS system given the restricted runway environment.

In summary

In all air carrier operations, margins are built into the system to allow for the inevitable human or mechanical malfunction. However, there will be a one-in-a-million series of occurrences that leads to a mishap. The odds of the captain having difficulty deploying thrust reversers are small, but it was a critical—if understandable—lapse. Multiple crews misunderstood company procedure regarding mandatory diversion to alternates. The nonstandard programming of the OPC shows the importance of standardization.

For light aircraft takeoff and landing performance estimates, the AOPA Air Safety Foundation recommends adding a margin of 50 percent to the pilot’s operating handbook for dry, level runways and 100 percent for wet or slippery conditions. Understand that you are operating in largely untested circumstances.

This seemly simple accident demonstrates multiple facets and serves to remind us that the physics of flying never takes a day off. Do you anticipate contingencies? If you’re betting that everything will work exactly as planned on a snowy runway, I have a bridge for sale.