When crash investigators talk about putting the pieces together, they mean it literally. Within hours of an accident, members of one of the National Transportation Safety Board's elite "go teams" are combing through hundreds of pieces of charred and twisted wreckage, putting them back together again like an enormous jigsaw puzzle. The sleuths pick up, catalog and tag all the debris, down to the last bolt and screw. They either lay the fragments on a grid, wingtip to wingtip and nose to tail, or hang them inside the shell of the same kind of plane that went down. Then they check each one, some smaller than an aspirin, for clues to the tragedy. "There's a genius to this process," says Carl Vogt, an NTSB member who is overseeing the painstaking investigation of Flight 427. "We've got people out there on their hands and knees going through the grime, and people with laptops in Washington working on this."
Call it forensic aviation. Drawing on about 10 scientific fields ranging from meteorology to propulsion, it has enabled the NTSB to determine the proximate causes of all but four of the thousands of crashes it has investigated since the agency was established in 1967. Sometimes the clues are straightforward. After an Air Florida jet crashed into the icy Potomac River in 1982, the NTSB determined that ice blocked sensors that read the engine thrust; the crew saw an artificially high reading, and the pilot failed to throttle up the plane enough, sending the Boeing 737 into a stall. Even when a plane seems to have vaporized on impact, a clue as small as a light-bulb filament can be revealing. When officials found one intact in the debris of the 1986 crash of a commuter plane in east Texas, they quickly ruled out failure of the on-board electrical system: if there had been no power, the light bulb would have been off and the filament cold, making it shatter on impact. And in what was arguably the most elaborate and challenging crash investigation ever, Scotland Yard combed 845 square miles of debris from the 1988 crash of Pan Am Flight 103 over Lockerbie, Scotland. Two fingernail-size plastic fragments allowed the team to trace the crash to a bomb inside a radio-cassette player.
Often, "the board operates by eliminating, not pinpointing," says former NTSB member Lee Dickinson. In Pittsburgh last week, for instance, the go team quickly found that fans from the two engines were bent in such a way as to suggest the engines were running at impact. If other clues match, that would rule out total engine failure as a cause of the disaster. And flames at the crash site diminished the possibility that the plane had run out of fuel. Beyond those hints, clues to the cause of any crash run the gamut from the "oh, s--t" on a cockpit recorder to the molten metal splattered against an engine housing:
The data recorder is so precise that investigators can often use it to create a computer simulation of the last minutes of a flight, including a reproduction of the indicator dials the pilot would have seen. If the data show significant acceleration or deceleration within a few seconds, that may suggest that the plane flew into a deadly kind of wind shear called a microburst. A microburst forms when a storm creates a downdraft that spreads out in a starburst, typically .25 to 2.5 miles across, as it hits the ground. A plane passing through a microburst will first encounter head winds, which make the craft pitch up; the pilot compensates by throttling back. Then the plane abruptly meets tail winds. With the engine throttled back just seconds earlier, the craft may not have enough airflow to keep it aloft. When the flight-data recorder from a 1985 Delta Air Lines crash in Dallas showed the plane losing 60 knots of airspeed in 17 seconds, it was a solid clue that a microburst brought it down. One hundred thirty-seven people died when the craft, approaching the runway, bounced across a freeway and into water tanks, where it burst into flames.
Engine propellers can kss provide what crash sleuths call "the Easter egg" of clues. If a piece of propeller lacks certain scratches, it means the blades were not turning at the time of the crash, says metallurgist David Levinson. But the turbine blades have to do more than turn, since simply moving forward through air, like a child's pinwheel, can spin the blades around; spin alone doesn't mean they're providing any forward thrust. For that, the blades have to be "turnin' an' burnin'," as jet jocks put it: not merely rotating in the airflow but igniting fuel for the thrust that keeps a plane aloft. How do investigators tell? An engine producing thrust is hot. At impact, the turbine blades grind against containment rings; the tips of the hot blades melt and splatter inside the rings. Lack of splatter is an unambiguous sign that the engine wasn't working at impact.
Planes can fly without engines, at least for a while. "A jet glides remarkably well," says aerospace engineer Hubert Smith of Pennsylvania State University. "It doesn't just drop out of the sky." But it can't fly without the hydraulic system, which controls the ailerons, elevator and rudder. So in 1989, when a United Airlines pilot on approach to Sioux City, Iowa, suddenly lost control of his DC-10 and crashed into a cornfield, hydraulics were the first suspect. Sure enough, the go team investigating that crash discovered a small fatigue crack in a compressor disc (found by a farmer in his cornfield). They deduced it had spun out of the engine and cut the hydraulic lines, rendering the plane unflyable. The pilot managed a relatively "safe" crash landing, and 184 out of 296 passengers survived. But the crew of an American Airlines DC-10 taking off from Chicago in 1979 wasn't so lucky. A pylon ripped off a wing, tore through the hydraulic lines "and prevented the pilots from having any control," says Jack Enders, former president of the Flight Safety Foundation. The plane yawed left just after takeoff, rolled over and plunged into the ground in a holocaust, killing all 271 peo-ple on board and two more on the ground.
If this field investigation is typical, it will be finished by next week. But a final report on the tragedy will be at least another nine months in coming, as the NTSB examines every possible crash scenario. Investigators have one unusual clue, however, as they piece together the causes of last week's disaster: it was hauntingly similar to the crash of a United Airlines 737 near Colorado Springs, Colo., in 1991. As that jet approached the airport, it suddenly pitched up, swung sharply right and nose-dived into a park, killing all 25 aboard. The flight recorder suggests USAir 427 did much the same -- rolling left and plunging straight into the ground at high speed. But the similarities are little comfort to investigators. The Colorado crash is one of the four that the NTSB has never resolved.