For almost five years, one of the world's largest jetmakers knew that the 27-foot-tall tail fin on one of its jets had almost snapped off in flight.
Officials with manufacturer Airbus understood that losing a tail fin would prove catastrophic. Even so, they kept their concerns to themselves until after a tail fin did break off one of its jets, causing the second-worst aviation disaster in U.S. history.
Not until after American Airlines Flight 587 crashed in 2001 -- a catastrophe investigators say was caused when the tail fin broke off the A300 jet -- did Airbus disclose its findings from an incident in 1997 to government safety officials, a USA TODAY investigation has found.
Had federal regulators known earlier how easily tail fins could break in flight, the crash of Flight 587 -- and the deaths of 265 people -- might have been prevented, according to some accident investigators and aviation safety experts. The jet crashed Nov. 12, 2001, just after takeoff from New York's John F. Kennedy International Airport.
The nation's aviation safety system is designed to prevent crashes by learning lessons from close calls. In this case, the system broke down.
''When I heard (about what Airbus knew), it made me sick,'' says Bernard Loeb, who headed the National Transportation Safety Board (news - web sites)'s (NTSB (news - web sites)) aviation division until January 2001. ''People are kicking themselves.''
The system is dependent on airlines and jetmakers sharing their knowledge and experience with federal regulators. The NTSB, an agency with limited resources, cannot function without this help, current and former investigators agree. But those are the very entities the NTSB investigates.
Airbus officials say they did nothing wrong. They say that the crash of Flight 587 was due to mistakes by the pilots and that there is nothing they could have done to prevent it. Attempts to find fault with the company are ''projected wisdom in hindsight,'' spokesman Clay McConnell says.
USA TODAY reviewed hundreds of pages of NTSB documents and interviewed more than a dozen government officials and knowledgeable sources. The review found that on May 12, 1997, American Airlines pilots on Flight 903 from Boston to Miami lost control of the jet. In response, they made a series of radical maneuvers that placed extreme stress on the tail fin, nearly snapping it off. Eventually, they landed safely.
Within days, federal investigators knew that the jet had nearly crashed, but they did not realize how close it came to breaking apart in midair. Then last year, a closer examination of the tail revealed that it had cracked during the incident. If federal regulators had known this information, they say, they could have warned pilots to avoid the nearly identical maneuvers that caused the crash two years ago.
It's impossible to know for certain whether a more thorough investigation in 1997 could have prevented the 2001 crash, several senior investigators say. Those officials still do not know why the co-pilot aboard Flight 587, leaving New York for the Dominican Republic, suddenly began his maneuvers. A final report on the accident is expected early next year.
Even so, said NTSB board member Carol Carmody, ''It's enormously frustrating that we didn't have this information.''
Carmody was acting chairwoman of the agency when Airbus revealed its findings last fall. ''I feel that we missed an opportunity,'' she said in an interview.
The stakes are also enormous in lawsuits stemming from the 2001 crash. Airbus and American Airlines are battling over who should pay damages for the crash.
The close call
Flight 903 from Boston had been routine, if a little bumpy. The pilots, waiting for storms over Miami to clear, reminded passengers to buckle their seatbelts. Controllers ordered them to hold at 16,000 feet near West Palm Beach.
At 3:29:14 p.m. May 12, 1997, as the jet began a turn, it suddenly banked hard to the right, back to the left, then to the right again.
''It was horrifying,'' Michelle Singh, 36, who was seated in row 16, recalled in an interview. ''There were no words to explain. People crying. People hurt. People scared. I was ready to die.''
Passengers clung to each other as the gyrations tossed them from side to side. Anything not strapped down -- shoes, briefcases and passengers themselves -- flew about the cabin, according to NTSB records. After 12 seconds, the jet began to plunge. It fell 3,000 feet in 18 seconds. Melanie Joison's baby flew out of her arms. Joison unbuckled her belt to grab her baby and crashed face-first into the ceiling. The blow knocked her unconscious and broke four of her ribs. Other passengers safely caught the baby.
''The terror and the screams were more than I have ever experienced,'' Scott Stow, an American pilot sitting in the passenger section, told investigators then.
Capt. Mark Eberle and co-pilot Donald Rescigno told investigators it seemed that a mysterious force -- perhaps a powerful downdraft -- had blown the jet out of control.
Initially, the jet banked 56 degrees to the right -- twice as steep as a passenger jet ever gets in a normal flight. The co-pilot tried to level the jet with the control wheel, which activates panels on the wings. It had no effect.
Rescigno tried the rudder, a large vertical panel at the rear of the tail fin. His left foot stomped on one of two pedals that move the panel. The rudder swung left.
A jet's rudder has tremendous power. It keeps the jet flying straight if one of the two engines fails. The control wheel, which looks similar to a car's steering wheel, is the preferred way to level a jet's wings, but the rudder accomplishes the same thing in a clumsy, overpowering way.
Rescigno held the rudder pedal down for four seconds. By then, the jet was rolling back to the left so quickly he could not control it. The left wing dropped sharply. Rescigno responded by slamming on the right rudder pedal, which began the cycle again. The jet banked back to the right even more steeply, to 65 degrees.
Overall, the jet banked left or right nine times within 40 seconds. In the most severe bank, the jet tilted at 83 degrees -- its right wing pointed nearly at the ground.
Eventually, the pilots increased the speed of the jet enough to regain control. Thirty minutes later, the jet landed in Miami, the cabin a mess of upended food carts, luggage and trembling passengers.
The pilots' rudder movements were nearly identical to those on Flight 587. The co-pilot on the later flight whipped the jet's rudder left or right five times.
On its last swing, the tail fin snapped off the fuselage.
Pilots caused 'stall'
Soon after Flight 903, NTSB investigators discovered that the incident had little to do with a gust of wind. The pilots had made a series of errors that caused the wild ride.
The pilots allowed the jet to slow too much as it entered the holding pattern, the jet's data recorder showed. The pilots apparently put the engines in idle as they descended and neglected to add power after leveling off at 16,000 feet. As a result, the jet essentially stopped flying, a condition known as a ''stall.'' When the pilots began the initial bank to the right, the stall caused the right wing to drop more than the pilots intended.
The discovery that pilots could forget to maintain enough speed disturbed the NTSB. The agency focused on that failing as the central problem.
A yearlong probe began.
Federal investigators had no idea that within a month of the incident, Airbus engineers in Europe found an additional problem: The jet's gyrations had put massive strain on the tail fin.
In an internal memo on June 12, 1997, an unidentified Airbus official wrote that his department ''urgently'' recommended additional inspections of the jet because the forces on it had apparently exceeded the ''design limit.'' That meant that the wind and jostling on the tail fin had exceeded the greatest forces it had been expected to experience in its lifetime.
Then, on June 19, a more complete analysis showed that the forces not only had gone above the design limit, they also apparently had reached the ''ultimate limit.''
When engineers build a jet, they compute the greatest forces likely to hit surfaces, such as wings and the tail. These are known as design limits. For a safety margin, international aviation regulations mandate that they make those surfaces 50% stronger. These are the ultimate limits, above which a tail fin or other surface is expected to fail.
It's extremely rare for commercial aircraft to reach a design limit in flight. It's almost unheard of for one to reach an ultimate limit. That would mean it had come dangerously close to breaking apart. Such a finding would get immediate attention from federal investigators -- if they learn about it.
Had Airbus pressed the matter in 1997, its findings would have been worse. Flight 903's tail fin reached or exceeded its ultimate limit three times during the incident, Airbus itself calculated last year. A Federal Aviation Administration (news - web sites) official testified at an NTSB hearing in October that a conservative estimate found the stress on the tail fin went well above the ultimate limit -- to within 1% of the force required to break it off.
In June 1997, Airbus requested that American Airlines perform another inspection of the jet to ensure it was not damaged. American inspectors, following Airbus' instructions, examined the tail fin. But they did not use methods that would have allowed them to see inside the tail fin. They saw no damage from their visual inspection, and the jet continued to fly for nearly five years.
Only last March, as part of the Flight 587 investigation, did Airbus conduct an ultrasound inspection of the tail fin on the jet involved in the 1997 incident. The inspection found two crescent-shaped cracks at one of the points where the tail fin attaches to the fuselage. The fin was replaced. Airbus says the tail fin was still strong enough to meet regulatory requirements.
Airbus engineers weren't the only ones who expressed worries about the incident. A high-ranking American Airlines pilot wrote a memo to a senior official in May 1997 warning that the pilots' use of rudder had nearly caused major structural damage to the jet. Paul Railsback, flight operations managing director, urged immediate changes in American's pilot training, says the memo obtained by USA TODAY.
Again, the NTSB has no record that Railsback's concerns were brought to its attention. American says it altered its training to reflect his comments.
Airbus officials deny withholding data or hindering the probe of Flight 903. ''I am convinced that my company made a good faith effort to raise these concerns,'' says McConnell, Airbus' spokesman.
Officials also say they did not realize that the tail fin on the jet had nearly broken loose. In part, that's because Airbus built the tail fin 25% stronger than regulations required. By insisting on an additional inspection of the tail, Airbus felt confident that it had not been damaged, McConnell says.
Furthermore, he says, the 1997 assessments were imprecise and Airbus did not calculate the actual stress on the tail until last year. The NTSB and American had access to the same data and failed to raise concerns, he says.
At the NTSB hearing in October on Flight 587, an American Airlines official was permitted to put questions to Airbus officials. With the two firms feuding over who is at fault in the 2001 crash, the questioning quickly grew tense. Airbus' Michel Curbillon said the company had shared its concerns about Flight 903 with federal investigators and others. ''This has been known within the company and was also informed to everybody,'' Curbillon said.
His questioner, American's Tim Ahern, who had worked on the Flight 903 investigation, disputed him. ''Frankly, as a party to that event, sir, this information was just (released) this year,'' Ahern said.
Curbillon pointed to a report submitted to the NTSB in August 1998, on the Flight 903 investigation. Airbus' flight safety director, Yves Benoist, wrote that pilots needed to be trained better on how to use the rudder. ''Using too much rudder in a recovery attempt can lead to structural loads that exceed the design strength of the fin,'' the report said.
Airbus officials say they underscored their concerns in 1997 by raising repeatedly the broader issue of pilot training on rudder use. For example, an August 1997 letter sent to American and written by Airbus, Boeing and the FAA warned that pilots could damage a jet with too much rudder.
But the NTSB has no record that Airbus disclosed what it had learned about the stress put on Flight 903's tail fin. The submission to the NTSB contained only a general warning about rudder use. The letter to American did not mention Flight 903 and was not sent to the NTSB until last year.
NTSB and American officials say the Airbus warnings were so vague that they did not attract attention in the Flight 903 probe.
When Loeb, who has retired from the NTSB, learned in October that Airbus knew in 1997 about the high stress on Flight 903, he was dumbstruck.
He had overseen the earlier investigation and never heard a word about the high stress on the tail fin, he says. Even if the calculations were preliminary, Airbus had a duty to share them, he says.
''That's a significant, significant issue,'' Loeb says. The jet involved in the incident should have been grounded until regulators were certain it was not damaged by the extreme forces, he says.
Officials who were part of the probe in 1997 say if they had known of Airbus' findings and the damage to the fin, it would have changed the focus of the investigation. Rather than focusing on the pilots' actions, the NTSB would have looked more closely at the rudder itself, they say. ''I think the answer is most assuredly we would have done something more on 903 if we had known this, if we had heard this from Airbus,'' the NTSB's Carmody says.
What about safety board?
Among the small circle of senior accident investigators, there is also debate about whether the NTSB itself should have seen the clues back in 1997 that rudder use could damage a tail fin.
According to former NTSB chairman James Burnett, the answer is yes. The agency's job is to turn the jumble of aviation incidents into nuggets of common sense that prevent accidents. By definition, it should have unraveled the puzzle and acted, Burnett says.
But, with only 60 aviation accident investigators and engineers to work cases involving airlines, it does not surprise him that the NTSB occasionally misses things. This is especially true during busy periods such as 1997, when the board was occupied with two crashes that had occurred the prior year: TWA Flight 800 and the ValuJet crash in the Florida Everglades. ''But it should be an occasion for taking stock,'' Burnett says.
Such concerns are not new. A Rand Corp. report in 1999 warned that the agency was too dependent on the airlines and aircraft manufacturers it investigates. The report urged greater use of independent experts to supplement NTSB investigations.
Loeb says he wishes now that, in the midst of perhaps the busiest period in the agency's history, he had spent more time on Flight 903.
''People are saying, 'We should have done more.' But the fact is, these are people who are working 60 hours, 70 hours a week and getting paid for 45 or 50. Yeah, they are going to miss things and feel bad about it,'' he says.
''On the other hand,'' Loeb adds, ''if Airbus did the calculations and didn't bother to tell us, shame on them.''