The Co-Pilot Held The Stick Back While The Captain Screamed “Descend”

The Co-Pilot Held The Stick Back While The Captain Screamed “Descend”

---

The stick was pulled back. Not once. Not twice. For four minutes, it stayed against the stop. The nose pointed at the stars. The ocean rushed up. The alarms screamed “STALL” forty-seven times. The co-pilot, young, terrified, did not understand why the plane would not climb. The other pilot, more experienced, shouted “Descend, descend, you need to descend.” The captain, returning from his rest, walked into a cockpit where the horizon was spinning and the altimeter was unwinding like a clock counting down to zero. None of them saw the simple truth. The plane was not falling because the engines failed. It was falling because the man holding the controls had been pulling backward for three full minutes, and no one had stopped him.

May 31, 2009. Rio de Janeiro’s Galeão International Airport. The evening sun painted the tarmac gold. At gate, an Airbus A330-203, delivered to Air France in 2005, sat waiting. It was one of the safest aircraft ever built. Zero crashes in commercial service. Sophisticated fly-by-wire systems that prevented pilots from making catastrophic errors. Computerized protections that would not allow a stall. A machine designed to be unflyable into danger.

The plane had undergone a major inspection just weeks earlier, in April 2009. It had accumulated nearly 19,000 flight hours. Two General Electric CF6 engines, reliable as sunrise. On board: 216 passengers from nearly 30 nationalities—Brazilians, French, Germans, and others. Plus 12 crew members. The flight from Rio to Paris usually took eleven hours. A routine red-eye across the Atlantic.

Three pilots sat in the cockpit, one more than usual. The long flight required a relief crew. Captain Marc Dubois, 58 years old, veteran of 11,000 flight hours, 1,750 of them on the A330. He had joined Air France almost eleven years ago. First Officer Pierre-Cédric Bonin, 32 years old, the youngest and least experienced—only 3,000 total hours, 800 on the A330. Most of those had been accumulated in the previous year. He had worked for Air France for five years. Relief First Officer David Robert, 37 years old, joined Air France in 1999, with 6,500 hours, 4,500 of them on the A330. The most experienced of the three on this specific aircraft.

The crew had arrived in Rio from Paris on May 28. They had a three-day layover before returning. Some had brought family members. They rested, visited the city, prepared for the journey home.

Takeoff was smooth. No complications. The weather in Rio was perfect.

The route from Rio to Paris crosses the Atlantic Ocean, beginning in Brazil and traveling over the North Atlantic before approaching the European continent. On paper, it was a straight line across the water. In reality, it crossed one of the most volatile weather systems on Earth: the Intertropical Convergence Zone. A region where trade winds from the northern and southern hemispheres collide, pushing warm, moist air upward. The result is a permanent belt of thunderstorms, towering cumulonimbus clouds that reach 50,000 feet, turbulence severe enough to tear wings off smaller aircraft.

At 7:29 PM local time, Flight 447 lifted off. Minutes later, cruising at 20,000 feet along the Brazilian coast, the relief pilot David Robert retired for his scheduled rest. He would sleep for about three hours. The captain and Bonin remained at the controls.

Around 1:30 AM, at 35,000 feet, they entered oceanic airspace. The flight was normal. Then a message arrived from Air France operations: strong thunderstorms ahead on the route. The pilots discussed options. Avoid the turbulence? Climb above it? Descend below it? The screens showed weather cells building like mountains in the dark.

Bonin, the young first officer, began to worry. The storm clouds were approaching faster than expected. He dimmed the cockpit lights to improve visibility through the windshield.

The Airbus began to enter the clouds.

Through the windshield, everything went black.

The ride became bumpy. Light turbulence at first, then stronger. The pilots debated climbing to flight level 370—37,000 feet—to top the storm. They were at flight level 350. It seemed they were scraping the highest clouds. But when they checked, they realized FL 370 was the maximum altitude the aircraft could reach. Climbing to the limit would leave no engine margin. At higher altitudes, the air is thinner. Engines produce less thrust. If something went wrong, they would have no reserve.

Captain Dubois mentioned St. Elmo’s fire—electrical discharges dancing on the windshield, a sign of building electrical charge. Another indication they were too close to the storm.

Around 2:00 AM, relief pilot David Robert returned from his rest. Captain Dubois briefed him on the situation: turbulence, inability to climb, embedded in cloud layer. Robert took the right seat where the captain had been sitting. Dubois unbuckled and left the cockpit to take his own rest period. He gave no specific instructions on how to handle the coming storm. No warnings about the danger of manual flight in icing conditions. No reminder about unreliable airspeed procedures.

He walked back to the crew rest area, lay down, and closed his eyes.

The cockpit was now in the hands of two first officers.

Robert adjusted his weather radar sensitivity to maximum. This made the radar more detailed but also more cluttered. He suggested turning left twelve degrees to avoid the worst of what appeared ahead. Bonin agreed. They did not inform air traffic control of the course change. A minor deviation. Routine.

Minutes later, the aircraft encountered an updraft. The sound changed—not rain hitting the fuselage, but ice crystals. Tiny frozen projectiles. The cabin temperature rose suddenly. An unusual smell filled the cockpit. Ozone. Robert recognized it. The smell of electricity mixing with air. Common in thunderstorms. The sudden temperature and humidity change meant the engines had ingested significant water or ice, temporarily overloading the air conditioning system.

Bonin reduced speed from Mach 0.82 to Mach 0.80, the recommended turbulence penetration speed. The autothrottle responded by reducing engine power to 84 percent N1.

Then the pitot tubes began to freeze.

Pitot tubes are small, heated probes mounted on the outside of the aircraft. They measure airspeed by sensing the pressure of oncoming air. They have heaters to prevent ice buildup. But when ice crystal concentrations are too high, the heaters cannot keep up. The ice accumulates faster than it can melt.

The crystals impacting the fuselage began to overwhelm the heating elements.

At 2:10 AM, the autopilot disconnected abruptly. Followed immediately by the autothrottle. An audible alert filled the cockpit. Simultaneously, a strong wind gust rolled the aircraft sharply to the right.

Bonin grabbed the controls. “I have the controls,” he said.

He was now flying an Airbus A330 manually at 35,000 feet, inside a thunderstorm, at night, with unreliable instruments. He had 3,000 total flight hours. Only 800 of them on this aircraft.

The aircraft was reasonably level, though developing a slight right roll. Then Bonin’s primary airspeed indicator dropped from 275 knots to 60 knots. An impossible reading. The backup indicator showed the same. The pitot tubes were blocked. The computers had lost reliable airspeed data.

Bonin reacted instinctively. He pulled the stick left to correct the roll. He pulled back to maintain altitude. This was the beginning of the sequence that would condemn the flight.

He overcorrected. Left, right, left, right. The aircraft began to porpoise. The pitch—nose-up angle—increased rapidly. The vertical speed shot upward. They were climbing too fast.

Alerts began to sound. A brief stall warning. Then silence. Then more warnings. Bonin said, “We don’t have any good speed readings.”

Robert, who was supposed to be managing the ECAM (Electronic Centralized Aircraft Monitor) checklist, said, “We’ve lost the speeds.” He began reading ECAM messages aloud, but too fast, too garbled. In just 18 seconds, ten error messages appeared on the screen, each with its own audible alert. The tension was palpable.

The aircraft pitch reached 12 degrees nose-up. Climbing at 6,900 feet per minute. Seven times the normal rate.

Then the flight directors reappeared on Bonin’s screen. But instead of being in altitude-hold mode, they had activated in vertical-speed mode. A distraction. Unhelpful.

Robert looked away from the ECAM screen and finally saw the pitch attitude. “Bonin, watch your speed,” he said. “Watch your speed.”

Bonin responded, “OK, OK, I’m descending.” He pushed the stick forward slightly. The climb rate decreased—but the aircraft continued climbing.

Robert: “Stabilize. According to the instruments, we’re climbing. You need to descend.”

Bonin: “I’m descending.”

Robert: “We’re climbing. According to this, we’re climbing. So descend.”

Bonin reduced pitch. The climb rate dropped to about 4,000 feet per minute. The aircraft reached its maximum calculated altitude of 37,000 feet. The flight directors disappeared again. The left-side airspeed indicator showed a correct reading of 223 knots. They had lost about 50 knots from before the pitot freeze. Speed was still decreasing.

The stall warning stopped, briefly. It seemed they might recover control.

Then the flight directors reappeared—showing a command to climb at 10,000 feet per minute. The same rate the aircraft was already climbing. A feedback loop of confusion.

Bonin said, “We’re climbing. We’re climbing.”

Robert, realizing the gravity, pressed the call button to summon Captain Dubois from his rest. He pressed it repeatedly. No response. “Where is the captain?” he asked. “Why isn’t he coming?”

The stall warning resumed. It would sound continuously for the next 45 seconds.

Bonin reacted by pulling back on the stick. Increasing the nose-up pitch to 16 degrees.

Sixteen degrees. The normal nose-up attitude for takeoff. At 37,000 feet. In a thunderstorm. With unreliable speed.

The horizontal stabilizer at the tail began moving to the maximum nose-up position. This made recovery even more difficult. The aircraft no longer had enough engine thrust to maintain the requested climb rate. At this altitude, above the maximum performance envelope, the engines could not produce sufficient power. Not even to maintain speed. Certainly not to accelerate.

Bonin pushed the throttle to TOGA—Takeoff/Go-Around power. Maximum thrust. The setting used for emergency climbs from runways. At 35,000 feet, it did nothing. The engines gasped in the thin air. The aircraft could not obey.

The nose pitched to 18 degrees up. The engines, finally unable to generate enough thrust, surrendered. The Airbus entered an aerodynamic stall. It began to descend.

The ice in the pitot tubes melted. The airspeed indicators returned. They showed 183 knots. They had lost 90 knots from the onset of the crisis.

The aircraft entered a steep dive. The stall alarm did not stop.

Robert, disoriented, said, “We still have the engines. What is happening? I don’t understand what is happening.”

The aircraft rolled right. Bonin moved his stick left and up, but the controls seemed to have no effect.

“I have no control over the aircraft,” Bonin said.

“Keep your stick left,” Robert told him.

Then Robert, the more experienced first officer, tried to take control. He pushed his stick forward to lower the nose. But the controls were in “dual input” mode. The system averaged his command with Bonin’s—and Bonin was still pulling back. The nose did not lower.

Robert said to Bonin, “Descend. Descend. Descend.”

Bonin replied, “But I’ve been pulling back the whole time.”

The words hung in the air.

Captain Dubois walked back into the cockpit. He had been in the crew rest area for perhaps ten minutes. The scene before him was chaos. The stall alarm was silent—momentarily—but the aircraft was descending at an incredible vertical speed. They had just passed 35,000 feet. The same altitude they had been at when he left.

The pitch oscillated between 9 and 15 degrees nose-up and nose-down. Ten ECAM warnings on the screen. Master Warning lights flashing. Severe buffeting shaking the airframe.

Dubois asked, “What are you doing?”

Both Bonin and Robert answered. They had lost control. Neither understood what was happening.

Instead of taking the controls, Dubois sat in the jump seat behind the pilots. He could not see that both control sticks were being pulled back—the opposite of what was needed to recover from a stall.

The aircraft continued diving. As the nose pitched down, speed increased. Angle of attack decreased slightly. The stall warning system, detecting valid data again, began to sound. This created a cruel irony: the alarm sounded when the nose dropped (correct indication) and stopped when the nose rose (incorrect indication). The pilots heard the alarm stop when they pulled back—reinforcing the wrong action.

The aircraft descended between 10,000 and 15,000 feet per minute. The nose oscillated between 8 degrees down and 15 degrees up. Robert moved the throttle to TOGA again. The aircraft stopped rolling right but began oscillating left and right.

Bonin continued trying to correct with his stick.

Robert asked Dubois: “What do you think? What do you think? What should we do?”

Dubois, who had missed the entire sequence, could only answer: “I don’t know. We’re falling.”

All three pilots discussed the situation. The aircraft passed through turbulence. Dubois urged Bonin to level the wings. Bonin ignored him.

In an astonishing exchange, the pilots argued over whether they were climbing or descending. They finally agreed they were descending.

Robert tried again to take control. He pushed his stick forward. But because Bonin still had his stick pulled back, the system averaged the inputs. No effect.

Robert said to Bonin, “Climb. Climb. Climb. Climb.”

Bonin responded: “But I’ve been pulling back the whole time.”

Dubois shouted: “No, no, no. Don’t climb. No, no.”

Robert said: “Then descend. Give me the controls. Give me the controls.”

Robert took control and finally lowered the nose. The aircraft began to gain speed. Essential for recovery. But without telling his colleagues, Bonin pulled his stick back again. Robert could not understand why the aircraft was still descending.

Robert changed strategy. He pulled back on his stick as well. The GPWS—Ground Proximity Warning System—began chanting: “Sink rate. Pull up. Sink rate. Pull up.”

The aircraft was at 2,500 feet. Robert changed his mind again. “Go forward. Forward.”

Bonin repeated: “Go forward. Forward.”

Both pilots pulled back to the maximum. The nose rose 16.2 degrees. The aircraft rolled slightly left. Speed: 107 knots horizontal. 107 knots vertical. The plane was falling almost straight down.

At 2:14:22, Bonin pressed and held his priority takeover button, locking out Robert’s control inputs. An alert announced “Priority right.” Bonin, in panic, screamed: “We’re going to crash! This can’t be happening! But what happened?”

Dubois ordered firmly: “Ten degrees of pitch.”

The stall alarm and GPWS played their death chorus in the background.

At 2:14:28, the Airbus A330 impacted the Atlantic Ocean. The angle: 45 degrees nose-down. Vertical speed: 10,912 feet per minute. Fifty-one knots vertical. One hundred and seven knots horizontal. The aircraft shattered into thousands of fragments. All 228 people on board died instantly.

No air traffic control centers responsible for monitoring the flight noticed anything wrong during the first hours. HF radio communication problems were not uncommon over the Atlantic. But by 1:53 AM, it became clear there were no signals from the flight. The crew had not reported their positions at required waypoints. All air traffic control centers coordinated to check if anyone had news. The last radar contact had been at 1:33 UTC, detected by radar over the Fernando de Noronha archipelago. The aircraft was at 35,000 feet—below its normal cruising altitude of 37,000.

Other aircraft on the same route tried to establish contact. No response. Satellite communication attempts failed. ACARS messages from Air France went unanswered.

The Brazilian Air Force reported that radar at the Cape Verde islands had not detected the aircraft. Search aircraft were dispatched. French reconnaissance aircraft joined.

On June 6, debris from Flight 447 was confirmed floating on the Atlantic. On June 7, more wreckage was found. The recovery of the black boxes seemed impossible due to ocean depth and complicated seabed geography.

Almost two years later, on April 2, 2011, the wreckage was finally located on a relatively flat area of the ocean floor at a depth of approximately 12,800 feet, six and a half miles from its last transmitted position. A ship equipped with two mini-submarines was sent to track the black boxes’ locator beacons, which had a battery life of about a month—long expired by then. But the search team had the coordinates from the wreckage.

On April 27, 2011, one of the flight data recorders was found at 13,000 feet. The memory unit was missing. It was found separately on May 1, in apparently good condition. On May 3, the cockpit voice recorder was found. The black boxes had finally spoken.

The final report concluded that no aircraft component had failed. The pitot tubes iced over, causing the autopilot to disconnect. This was a known vulnerability, but it should not have led to a crash. The standard procedure for unreliable airspeed is clear: set thrust to 85 percent N1, set pitch to 5 degrees nose-up, and wait for reliable data to return.

The crew did not apply this procedure.

Instead, the young first officer, startled by the autopilot disconnection and the roll, pulled back on the stick. This reduced speed further. He then continued pulling back for nearly four minutes, despite repeated stall warnings, despite the other pilot telling him to descend, despite the captain returning and asking what was happening.

The stall warning sounded 47 times. Forty-seven times, a computer screamed “STALL” at men who did not understand what the word meant in this context. They had been trained to trust the aircraft’s protections. They did not believe an Airbus A330 could stall. But the protections had degraded with unreliable data. The aircraft was in a stall from 37,000 feet all the way to the ocean.

David Robert finally understood the problem too late. His corrections were insufficient. Captain Dubois returned too late. And Pierre-Cédric Bonin, the young pilot who meant well, who only wanted to keep the plane in the air, held the stick back until the very end.

The accident led to sweeping changes in aviation: mandatory stall recovery training for all pilots, new unreliable airspeed procedures, improved pitot tubes, and the installation of cockpit cameras on some aircraft.

But none of that brought back the 228 souls who fell for three minutes and thirty seconds, in the dark, over the cold Atlantic, while the stall alarm screamed and the co-pilot asked, “What’s happening?”