Figure: Flight 587's Flight path and Key Events, from NTSB Report, Figure 3.

Eddie Sez:

Most pilots think of this as a wake turbulence incident, it was not. The wake turbulence encounter occurred at 1,700 feet and simulator tests confirmed it was easily survivable. The crash was caused by overly aggressive rudder inputs by a pilot who did not understand the aircraft's rudder limiter system, proper unusual attitude recovery procedures, and the nature of the aircraft's design maneuvering speed. These misunderstandings may have been caused by the excellent American Airlines Advanced Maneuvering Program (AAMP). The AAMP is what I think of as a doctoral level course in airmanship, one that could have easily been misunderstood by pilots not schooled in the more primary aspects of maintaining aircraft control.

While it is true the first officer was the cause of this crash, the captain could have saved the day had he been more proactive. Part of good Crew Resource Management is knowing when to step in. The first officer was out of his league during the unusual attitude recovery and the captain needed to say "I've got it," and then simply centered the rudders and rode it out.

What follows are quotes from the relevant regulatory documents, listed below, as well as my comments in blue.

Accident Report


[NTSB Report, page 2]

[Aeronautical Information Manual, ¶7-3-9] Because of the possible effects of wake turbulence, controllers are required to apply no less than specified minimum separation for aircraft operating behind a heavy jet and, in certain instances, behind large non heavy aircraft (i.e., B757 aircraft).

  1. Additionally, appropriate time or distance intervals are provided to departing aircraft:
    1. Two minutes or the appropriate 4 or 5 mile radar separation when takeoff behind a heavy/B757 jet will be: (i) From the same threshold.

Figure: Rudder Pedal Movements During the Second Wake Encounter, from NTSB Report, Figure 1.

[NTSB Report, page 3]

  • The National Transportation Safety Board's airplane performance study for this accident determined that flight 587 started its takeoff roll about 0913:51 and lifted off about 0914:29, which was about 1 minute 40 seconds after the Japan Air Lines airplane.
  • ATC's takeoff clearances were indeed 2 minutes apart, but it appears the Japan Airlines crew took more time to get to the lift off point.

  • Japan Air Lines flight 47 and American Airlines flight 587 were separated at all times by at least 4.3 nm horizontally and 3,800 feet vertically.

  • About 0914:43, the local controller instructed the flight 587 pilots to turn left, fly the bridge climb, and contact the New York TRACON departure controller. About 5 seconds later, the captain acknowledged this instruction. Radar data indicated that the airplane climbed to 500 feet above mean sea level (msl) and then entered a climbing left turn to a heading of 220°. About 0915:00, the captain made initial contact with the departure controller, informing him that the airplane was at 1,300 feet msl and climbing to 5,000 feet msl. About 0915:05, the departure controller instructed flight 587 to climb to and maintain 13,000 feet msl, and the captain acknowledged this instruction about 5 seconds later. About 0915:29, the CVR recorded the captain's statement "clean machine," indicating that the gear, flaps, and slats had all been retracted.

  • About 0915:35, flight 587 was climbing through 1,700 feet msl with its wings approximately level. About 1 second later, the departure controller instructed flight 587 to turn left and proceed direct to the WAVEY navigation intersection (located about 30 miles southeast of JFK). About 0915:41, the captain acknowledged the instruction. The controller did not receive any further transmissions from flight 587.

  • FDR data indicated that, about 0915:36, the airplane experienced a 0.04 G drop in longitudinal load factor, a 0.07 G shift to the left in lateral load factor, and about a 0.3 G drop in normal (vertical) load factor. The airplane performance study found that these excursions were consistent with a wake turbulence encounter. Between 0915:36 and 0915:41, the FDR recorded movement of the control column, control wheel, and rudder pedals. Specifically, the control column moved from approximately 0° (neutral) to 2° nose up, 2° nose down, and back to 0°; the control wheel moved a total of seven times, with peaks at 18° right, 30° left, 37° right, 34° left, 5° left, 21° left, and 23° right, before moving to between 5° and 6° left; and the rudder pedals moved from about 0.1 inch left (the starting point for the pedals) to about 0.1 inch right and 0.2 inch left before moving to 0.1 inch left. The airplane performance study indicated that, during this time, the rudder moved from 0° (neutral) to about 2° left, about 0.6° right, and back to 0°.
  • These control inputs appear to be appropriate, maintaining aircraft attitude primarily with aileron and very little rudder input.

  • During the wake turbulence encounter, the airplane's pitch angle increased from 9° to 11.5°, decreased to about 10°, and increased again to 11°. The airplane's bank angle moved from 0° (wings level) to 17° left wing down, which was consistent with the turn to the WAVEY navigation intersection.

  • At 0915:44.7, the captain stated, "little wake turbulence, huh?" to which the first officer replied, at 0915:45.6, "yeah." At 0915:48.2, the first officer indicated that he wanted the airspeed set to 250 knots, which was the maximum speed for flight below 10,000 feet msl. At that point, the airplane was at an altitude of about 2,300 feet msl.

  • FDR data indicated that, about 0915:51, the load factors began excursions that were similar to those that occurred about 0915:36: the longitudinal load factor dropped from 0.20 to 0.14 G, the lateral load factor shifted 0.05 G to the left, and the normal load factor dropped from 1.0 to 0.6 G. The airplane performance study found that these excursions were also consistent with a wake turbulence encounter. According to the FDR, the airplane's bank angle moved from 23° to 25° left wing down at 0915:51.5, the control wheel moved to 64° right at 0915:51.5, and the rudder pedals moved to 1.7 inches right at 0915:51.9.
  • While it appears this second encounter was similar to the first, the first officer's inputs were different. The aircraft was at a higher speed, 240 knots, and his rudder inputs were much higher. Whle 1.7 inches doesn't seem like a lot, at this speed with the aircraft's rudder control system, it would result in maximum deflection. More about this below, Rudder Control System.

  • At 0915:51.8, 0915:52.3, and 0915:52.9, the CVR recorded the sound of a thump, a click, and two thumps, respectively. At 0915:54.2, the first officer stated, in a strained voice, "max power." At that point, the airplane was traveling at 240 knots. About 0915:55, the captain asked, "you all right?" to which the first officer replied, "yeah, I'm fine." One second later, the captain stated, "hang onto it. Hang onto it." The CVR recorded the sound of a snap at 0915:56.6, the first officer's statement "let's go for power please" at 0915:57.5, and the sound of a loud thump at 0915:57.7. According to the airplane performance study, the vertical stabilizer's right rear main attachment fitting fractured at 0915:58.4, and the vertical stabilizer separated from the airplane immediately afterward. At 0915:58.5, the CVR recorded the sound of a loud bang. At that time, the airplane was traveling at an airspeed of about 251 knots.

  • The CVR recorded, at 0916:00.0, a sound similar to a grunt and, 1 second later, the first officer's statement, "holy [expletive]." At 0916:04.4, the CVR recorded a sound similar to a stall warning repetitive chime, which lasted for 1.9 seconds. At 0916:07.5, the first officer stated, "what the hell are we into...we're stuck in it." At 0916:12.8, the captain stated, "get out of it, get out of it." The CVR recording ended 2 seconds later. The airplane was located at 40° 34' 37.59" north latitude and 73° 51' 01.31" west longitude. The accident occurred during the hours of daylight.


The Captain

[NTSB Report, page 10]

The First Officer

[NTSB Report, page 11]

Figure: Rudder Control System, from NTSB Report, Figure 6.

The Rudder Control System

[NTSB Report, page 18]

  • The rudder control system includes (1) the rudder pedals, the rudder trim actuator, the yaw damper actuator, and the yaw autopilot actuator, which command the rudder to move; (2) pushrods, bellcranks, a tension regulator, and cables (also referred to as linkages), which transmit rudder commands; (3) three servo controls (upper, middle, and lower), which operate the rudder; (4) a rudder travel limiter system, which provides a variable stop that limits rudder pedal travel with increasing airspeeds; and (5) a differential unit, which is a mechanical device that sends the rudder servo controls a command that is the sum of a pilot or an autopilot input and a yaw damper input. The maximum rudder deflection is 30o either left or right, the maximum rate of rudder movement (with no loads) is 60° ±5° per second, and the maximum rudder pedal displacement is 4 inches.

  • At the public hearing for this accident, the vice president of Airbus' flight control and hydraulic department stated that the rudder was not normally used during cruise flight to control roll. The vice president of training for Airbus North America customer services stated that the ailerons and spoilers were used to control roll. This Airbus vice president also stated that the rudder was used to control yaw and sideslip and that the rudder "is not a primary flight control to induce roll under any circumstances unless normal roll control is not functional." He further stated that, if pilots were to experience a roll for any reason, "they will intuitively try and counter the roll with their normal roll control. If they exhaust their normal roll control, they will then go to rudder to try and induce a roll." He added that it would be "a long path to get down to that level of degradation to where a pilot would be exposed to using rudder."
  • This agrees with most of my experiences in jet aircraft, except those that are aerobatic in nature. I can see that a pilot brought up in light aircraft may not have the same "intuitive" understanding.

  • Regarding the rudder travel limiter system, American Airlines' A300 fleet standards manager stated that, before the flight 587 accident, he thought that pilots knew "quite a bit" about the rudder limiter system but that, after the accident, it became apparent that pilots, as well as the aviation industry as a whole, "didn't know much about rudder limiter systems and in fact possibly had wrong perceptions." The A300 fleet standards manager also stated the following:
    • Most pilots think that a limiter on some system will protect...the pilot from exceeding whatever parameter that limiter is limiting. And in this case...and it's not unique to Airbus aircraft...the pilots think that the rudder limiter will protect the aircraft structurally, and if it can't...they think...that there would be a limitation or a warning or caution or a note that would indicate...that the rudder limiter couldn't protect [the aircraft] structurally.
  • Regarding the rudder pedals, the A300 fleet standards manager stated that, before the flight 587 accident, American Airlines did not teach its pilots during training that rudder pedal movement would become restricted as airspeed increased. The fleet standards manager also stated that he did not know that the rudder pedal movement would become restricted because the pedals are not normally pushed to the stop in flight. In addition, the fleet standards manager stated that, before the flight 587 accident, he did not think that any pilot would have thought that full rudder could be gained from about 1 1/4 inch of pedal movement and 10 pounds of pressure (above the breakout force) at an airspeed of 250 knots.
  • This is an unusual system and runs contrary to most rudder limiter systems. In the G450, for example, available rudder travel is indeed reduced at higher speeds.

Simulator Tests

[NTSB Report, page 18]

American Airlines Advanced Maneuvering Program (AAMP)

[NTSB Report, page 18]

Misunderstanding VA

A pilot's over-aggressive use of the rudder may have been abetted by the knowledge they were below design maneuvering speed. The flight of American Airlines 587 provides a good case study:

[NTSB Report, ¶2.5.3]

Design Maneuvering Speed, VA, is an aircraft certification number of the manufacturer's choosing. It changes with weight, configuration, and altitude. Unless you have a complete chart in front of you, it may be meaningless in most specific instances. You should never manipulate a flight control faster than you can perceive the changes that result.

More about this: Technical / VA - Maneuvering Speed.

Probable Cause

Figure: Shear, bending, and torsion, from NTSB Report, Figure 9.

[NTSB Report, ¶3.1] The National Transportation Safety Board determines that the probable cause of this accident was the in-flight separation of the vertical stabilizer as a result of the loads beyond ultimate design that were created by the first officer's unnecessary and excessive rudder pedal inputs. Contributing to these rudder pedal inputs were characteristics of the Airbus A300-600 rudder system design and elements of the American Airlines Advanced Aircraft Maneuvering Program.

  • Flight 587's cyclic rudder motions after the second wake turbulence encounter were the result of the first officer's rudder pedal inputs.

  • Flight 587's vertical stabilizer performed in a manner that was consistent with its design and certification. The vertical stabilizer fractured from the fuselage in overstress, starting with the right rear lug while the vertical stabilizer was exposed to aerodynamic loads that were about twice the certified limit load design envelope and were more than the certified ultimate load design envelope.

  • The first officer had a tendency to overreact to wake turbulence by taking unnecessary actions, including making excessive control inputs.

  • The American Airlines Advanced Aircraft Maneuvering Program ground school training encouraged pilots to use rudder to assist with roll control during recovery from upsets, including wake turbulence.

  • The American Airlines Advanced Aircraft Maneuvering Program excessive bank angle simulator exercise could have caused the first officer to have an unrealistic and exaggerated view of the effects of wake turbulence; erroneously associate wake turbulence encounters with the need for aggressive roll upset recovery techniques; and develop control strategies that would produce a much different, and potentially surprising and confusing, response if performed during flight.

  • Before the flight 587 accident, pilots were not being adequately trained on what effect rudder pedal inputs have on the Airbus A300-600 at high airspeeds and how the airplane's rudder travel limiter system operates.

  • The Airbus A300-600 rudder control system couples a rudder travel limiter system that increases in sensitivity with airspeed, which is characteristic of variable stop designs, with the lightest pedal forces of all the transport-category aircraft evaluated by the National Transportation Safety Board during this investigation.

  • The first officer's initial control wheel input in response to the second wake turbulence encounter was too aggressive, and his initial rudder pedal input response was unnecessary to control the airplane.

  • There is a widespread misunderstanding among pilots about the degree of structural protection that exists when full or abrupt flight control inputs are made at airspeeds below the maneuvering speed.

See Also:


Aeronautical Information Manual

NTSB Aircraft Accident Report, AAR-04/04, In-Flight Separation of Vertical Stabilizer, American Airlines Flight 587, Airbus Industrie A300-605R, N14053, Belle Harbor, New York, November 12, 2001