Site Presented By
Tuesday, Nov 10, 2009
Who's to blame when a 737 goes skidding off a snowy runway?
Dec 23, 2005 | At Chicago's Midway Airport on the evening of Dec. 8, conditions were slick and the runway was short. All of Midway's runways, arranged in tick-tack-toe crosshatch, are in fact short. The longest, runway 31C/13C, is a hair over 6,500 feet (shorter, even, than what's found at La Guardia or Reagan National). Bound for this runway came Southwest Airlines flight 1248, a year-and-a-half-old Boeing 737-700 on arrival from Baltimore in the midst of a heavy snowstorm. The 737 touched down and promptly failed to stop, tobogganing off the end of the runway, through a blast fence and perimeter wall, and into the middle of a street, where it collided with two passing automobiles, pancaking one of them. A 6-year-old boy, a rider in the crushed car, was killed. None of the 103 passengers or crew on the aircraft was seriously hurt.
Investigators found the 737's autobrakes switch set to the "maximum" position. To many that would seem the logical choice, except that Southwest Airlines had been training its crews not to use the autobrakes system under any circumstances. Why this restriction existed is somewhat unclear, though, as it happens, Southwest was on the verge of reversing policy at the time of the Midway overrun. It's possible the pilots' decision to employ automatic brakes was influenced by the pending authorization.
The airline has pointed out that the crew had not been "trained" in their use. While that's technically correct, there's little training required beyond adding a couple of pages to one's manual and reviewing some fine print. (On the flight deck of the 737, the autobrakes controls are extremely simple. In this photo, look below the glare shield for a gray rotary switch to the left of the green, triangularly arranged landing-gear indicators.) Still, Southwest can rightfully claim that its pilots relied on a piece of machinery that was neither sanctioned nor, officially, functioning.
That's the legalese, and although the pilots might ultimately be penalized, it's looking less and less like autobrakes had anything to do with the matter. The preliminary report from the National Transportation Safety Board shows that, according to the jet's digital flight recorder, "autobrakes were active and provided high brake pressure upon touchdown."
So, ironically, the crew of flight 1248 seems to have broken company rules and done the wrong thing, which in many people's eyes was the right thing, or at worst an irrelevant thing -- and still the plane couldn't stop in time. What, then, was to blame?
Modern aircraft-landing gear is augmented by sophisticated braking and antiskid technology, and autobrakes are part of that bundle. They work exactly as the name suggests: Electronics take care of the stopping while the crew uses the rudder pedals for directional control only. Pilots are able to select particular levels of autobraking power. Autobrakes oversee the rate of deceleration. If the combination of spoilers and reverse thrust, for example, is adequately slowing the aircraft, wheel braking may not occur at all. Nevertheless, on some aircraft, like the 737, performance can vary from not very effective to overly aggressive. Especially at higher settings, the deceleration can be jarring, and so the equipment isn't always used.
"I rarely use autobrakes," says an Airbus A320 pilot. "Our settings for landing are low and medium. [In the A320's case, maximum is reserved for aborted takeoffs.] For regular operations, low is too low, and medium is too high. On longer runways, I use reverse thrust to about 100 knots, then plain old foot [manual] braking."
When conditions are slippery and/or runways are short, however, autobrakes can be helpful. The pilots can focus on maintaining a straight course while the plane takes care of slowing down. "On a short or slick runway, the same pilots explains, "autobrakes do an awesome job."
Let's look at several other factors, beginning with the weather. At the time of the plane's arrival, seven inches of snow had fallen; braking action was recorded as "fair" to "poor" along portions of the stubby runway; and the plane made its landing with a six-knot tailwind.
This may have exceeded flight 1248's performance envelope. Investigators say that under the known combination of aircraft weight, weather and surface contamination, the jet needed about 5,300 feet of runway once its tires met the pavement. Touching down about 2,000 feet from the threshold -- that's within designated touchdown zone parameters, if a bit excessive under the circumstances -- only 4,500 feet remained. That difference of 800 feet put the 737 into the street.
But when planning an approach, real-time performance data is easily available to both pilots and airline dispatchers. If conditions are tight for a given runway, crews are obligated to know whether or not their machine has the room to perform safely. In interviews with investigators, both of flight 1248's pilots maintain that all pertinent information had been entered into an onboard performance computer, which affirmed the landing was legal. En route, the flight had been contacted at least twice by company dispatchers, who concurred that all appropriate parameters had been met. Had those numbers not worked out, a different runway (apparently impossible under the known ceiling and visibility conditions), or a different airport, would have been the only options.
Then there's the issue of reverse thrust. In the NTSB account, the captain states that he could not get the reverse-thrust actuating levers out of their stowed position after landing. Normally, reverse thrust is activated only a few moments after touchdown. Here, the levers remained stowed for 18 seconds, until the first officer reached over and successfully deployed them. Much has been made of this in the newspapers, but it's not as significant as it sounds: Reverse thrust accounts only for a small portion (10-15 percent) of a plane's stopping power -- most of the work is done by the brakes -- and, for an added safety buffer, all performance data is calculated without the effects of reverse.
