Continental Flight 1404

Departure into a crosswind

Continental Airlines Flight 1404 was preparing to leave Denver International Airport for Houston on the evening of 20 December 2008. The Boeing 737-500, registered N18611, carried 110 passengers, two pilots and three flight attendants. The flight was a scheduled domestic passenger service operated under Part 121. Visual weather conditions prevailed, with strong, gusty winds from the west. The captain was flying the airplane, while the first officer handled radio calls and monitored the takeoff.

The pilots arrived at about 17:00 and completed their preflight preparations. The captain inspected the airplane outside while the first officer performed cockpit checks. After pushback approval at about 18:04, they taxied toward runway 34R. The departure information they had received reported winds of 11 knots. Runway 34R was a dry, grooved concrete runway, 12,000 feet long and 150 feet wide. A postaccident inspection and friction test confirmed a bare, dry surface with normal or good friction.

At 18:17:26, the tower cleared the flight for takeoff and reported wind from 270 degrees at 27 knots. This produced a crosswind component of 26.6 knots, below Continental's dry-runway takeoff guideline of 33 knots. The crew had noticed clouds moving rapidly and knew there was a crosswind. Given the wind information supplied to them, the NTSB considered their decision to proceed reasonable.

Rudder inputs and a sudden left turn

The captain advanced the thrust levers and concentrated on the runway ahead. The first officer watched the engine instruments until takeoff power was set, then monitored airspeed for the standard calls. At 18:18:04 he reported that power was set at 90.9 percent. As the airplane accelerated, the captain applied increasing right rudder to counter the left crosswind. The rudder became increasingly effective with speed, taking over the directional-control task from nosewheel steering.

The recorded control movements then became much larger. One right rudder input reached about 88 percent of available pedal travel, followed by a reduction to about 15 percent. The nose first moved right, then turned left again. The captain applied a second substantial right rudder input, reaching about 72 percent. This slowed the left turn briefly. The NTSB performance study estimated that the crosswind component increased from about 30 knots to more than 40 knots during this second cycle, reaching a peak of about 45 knots at 18:18:12.

As the airplane began turning left rapidly, the captain relaxed the right pedal to neutral. After a small, brief further input, he stopped applying rudder. He turned the control wheel right and reached for the nosewheel steering tiller. These actions were ineffective for controlling the airplane at that speed. The NTSB found that the sudden threat, perceived loss of control and extreme time pressure likely produced acute stress that led to these control choices. The airplane departed the left side of runway 34R at 18:18:17, travelling about 110 knots.

Across the ground and out of the cabin

The 737 left the pavement about 2,600 feet from the runway's approach end. It crossed taxiway WC and an airport service road, traversing uneven ground west of the runway. The captain began reducing power at about 18:18:20 and called for a rejected takeoff about a second later. Braking and reverse-thrust commands followed. Engine damage prevented the requested increase in reverse power. The airplane reached about 120 knots before slowing; the recorders stopped at 18:18:27 while it was still moving at about 90 knots.

Ground scars showed where the landing gear crossed the taxiway and then lost contact with the falling terrain. When the airplane struck the ground again, its engines also left scars. Both main landing gear assemblies separated, the nose gear folded aft, and the fuselage broke near the wheel wells. The right-wing fuel tank was breached, and a fuel-fed fire developed on the right side. Both pilots sustained back injuries, and the captain suffered multiple spinal fractures.

The flight attendants saw the fire and promptly ordered an evacuation. They used the forward, overwing and aft exits on the left side. An exit-row passenger opened the overwing exit, and two deadheading pilots helped the cabin crew direct passengers and keep them away from the right-side exits. All occupants escaped before the fire entered the cabin. People with less serious injuries helped others up the hill to fire station 4 for assessment and treatment. Six people were seriously injured, 41 had minor injuries and 68 were uninjured. Nobody died.

Darkness and lower terrain initially concealed the accident from the tower. Calls from other pilots alerted controllers, who notified the airport rescue crews. The first location passed to responders was incorrect, and some vehicles changed direction after receiving the corrected position. The nearest fire station had its garage doors facing away from the wreckage. Firefighters reached the site about five minutes after the accident and extinguished the exterior and cabin fires. The NTSB concluded that the initial confusion had not made emergency-response timing a significant issue in the accident.

Reconstructing the winds and the controls

Investigators examined the flight recorders, runway marks, wreckage and airplane systems. They found no evidence of a preaccident failure in the engines, structure or systems, including nosewheel steering. The steering cable had broken under overstress after the airplane left the runway. Flight control movements matched the recorded inputs, and the airplane responded appropriately. Performance calculations showed that the rudder could have generated enough force to counter the weathervaning effect of the winds encountered.

The wind varied sharply across Denver airport. Analysis by the National Center for Atmospheric Research found mountain wave activity producing localized, intermittent surface gusts. The NTSB estimated the winds along the accident takeoff path from recorder data and airplane performance. These estimates gave winds of 30 to 45 knots from the west. The airport's wind sensors sampled different places and heights, and their recordings could miss peaks between samples. The investigation linked the differences between the earlier 11-knot report and the tower's 27-knot report to observation times, sensor locations and the changing local wind field.

The tower's display contained more adverse information than the departure-end wind passed to the crew. Airport wind information showed 35 knots with gusts to 40 knots. The controller followed the established practice of issuing departure-end winds, and the NTSB found no clear requirement to supply the most adverse displayed conditions. It concluded that the stronger information would likely have prompted the captain to delay departure or seek another runway. Denver's runway-selection policy also failed to account clearly for crosswind components.

Training for gusts at runway level

The captain had about 13,100 flying hours, including 6,300 in the 737. His records showed satisfactory training and checking, including simulator takeoffs and landings in a 35-knot crosswind. During the investigation, however, Continental discovered that its 737-500 simulator allowed only steady wind below 50 feet. Instructors had been unaware that gust inputs made at the controls of the simulator were not applied at those heights. The training therefore failed to reproduce the changing surface winds that challenged the captain in Denver.

Airline operating data also showed that strong crosswind encounters were rare. Among 940,000 recorded departures, Continental identified only 62 with crosswind components of at least 30 knots, measured seven seconds after takeoff. Four involved 737-500s. In the NTSB's simulator work, releasing rudder at 90 knots in a steady 35-knot crosswind led to a runway excursion within about five seconds. Resuming corrective rudder after two seconds enabled participants to remain on the runway; waiting three seconds made recovery inconsistent. The NTSB concluded that immediate renewed significant right rudder could have prevented this excursion.

The Board adopted its final report on 13 July 2010. It identified the captain's cessation of right rudder about four seconds before the excursion as the probable cause, in a gusting crosswind beyond his training and experience. The air traffic control system for disseminating wind information and inadequate crosswind training due to deficient simulator gust modelling contributed. The Board recommended research into mountain winds, transmission of maximum adverse wind components, explicit crosswind criteria for runway selection and realistic surface-gust profiles for simulator training.

Crew seats and unsecured galley equipment

The impact sequence exposed weaknesses in crew seating. Both pilots' seat height-adjustment mechanisms failed downward, and the front restraint attachment points fractured upward. The seats had been approved to older static-load standards. A later requirement for improved dynamic-impact protection applied to passenger and flight attendant seats in newly manufactured airplanes, but cockpit seats were exempted. The NTSB concluded that seats meeting the improved crashworthiness requirements would likely have reduced or prevented the pilots' injuries. It recommended applying those standards to cockpit seats in newly manufactured airplanes whose types had been certificated before 1988.

The forward flight attendant's jumpseat also failed. Laboratory examination found a manufacturing defect and preexisting fatigue cracks in its pivot plates. The cracks weakened the structure and extended during impact, allowing the seat pan to collapse. The flight attendant sustained minor injuries and remained able to help evacuate passengers. The Board recommended periodic inspections for fatigue cracks in this jumpseat model, with replacement when cracks were found.

In the aft galley, investigators found a drawer loose on the floor beside an unoccupied flight attendant jumpseat. Its compartment latch plate had been attached solely by adhesive. The Board found that this attachment method was inadequate because it could deteriorate with age and exposure to the environment. B/E Aerospace had issued a service bulletin in September 2009 describing mechanical attachment of the latch plate on Continental 737-500 galleys. Compliance was voluntary, and the manufacturers could not identify how widely similar attachments were used. The NTSB recommended mechanical fasteners for adhesive-only galley latch installations on transport-category airplanes.