The crash of Air Inter flight 148: Analysis
On the 20th of January 1992, Air Inter flight 148 departed Lyon, France, on a domestic flight to Strasbourg. But during final approach, flying at night in bad weather, the Airbus A320 slammed into the heavily wooded flank of Mount Ste.-Odile, killing 87 people. Nine survivors banded together to keep each other alive through the freezing cold night while rescuers struggled for hours to find the plane.
In the aftermath of the crash, the French BEA launched one of the most exhaustive investigations in its history, studying the disaster from every possible angle. What they discovered was that the crash occurred as a result of the convergence of a poorly designed user interface, inexperienced pilots, and the fundamental nature of human psychology.
Air Inter was France’s leading domestic airline for several decades, during which it was sometimes the only airline serving certain major domestic routes. The airline was a quasi-public entity, with several government-run enterprises holding a large stake, but it acted as a private corporation. Its stated goals were low fares and punctuality, both of which helped it cater to business travelers and put it into direct competition with France’s TGV high speed rail network. According to former Air Inter pilots, the airline actually paid crews more if they flew faster, as part of an effort to stay ahead of the TGV. Air traffic controllers knew this and would sometimes “help out” Air Inter crews by giving them shortcuts or expedited approaches.
Flight 148 was Air Inter’s regular commuter flight from Lyon in southeastern France to Strasbourg in the northeast, near the German border. On the evening of the 20th of January 1992, 90 passengers and 6 crew boarded this flight, operated by a state-of-the-art Airbus A320. In command of the flight were Captain Christian Hecquet and First Officer Joël Cherubin. Although both were experienced pilots, they were new to the A320: Captain Hecquet had 162 hours on the type, while First Officer Cherubin had just 61 hours and first flew passengers on the A320 less than a month earlier. Hecquet was described as a slow learner, but competent once he acquired the necessary skills, while Cherubin was described as “an average pilot, with no exceptional abilities or marked shortcomings.” However, those who knew him noted that he was academically adept and passionate about his job.
The pilots’ experience level is significant considering that the Airbus A320 was easily the most technologically advanced airliner of its time, and employed a novel philosophy surrounding the role of the pilots and the role of automation. The A320 was the first fully fly-by-wire aircraft, in which pilot inputs are fed to computers that then move the control surfaces. It was also the first airliner to have side stick controls and had several never-before-seen manually selectable autopilot modes. Air Inter’s decision to upgrade to the A320 from its previous fleet of 1960s-era Dassault Mercure and Caravelle 212 aircraft represented a massive technological leap, management of which required considerable investment from the airline. To facilitate the transition, Mercure and Caravelle pilots were retrained in large numbers to fly the A320, resulting in a glut of relatively inexperienced pilots. The computer system that paired pilots on regular flights did not take their experience level into account, occasionally resulting in dubious pairings like Hecquet and Cherubin.
There were other factors that made them a problematic pair besides their inexperience, however. Other pilots who interacted with Hecquet and Cherubin on the day of the flight noted that they had opposite personalities and did not get along well. Hecquet had a reserved air, liked to take things slow, and leaned heavily on routines. Cherubin was more outgoing, excelled at improvisation, and appeared more at ease with the A320. Cherubin was also not a passive subordinate; he frequently corrected his captain’s mistakes and carried out procedures without having been told to do so, both traits that Hecquet found annoying. The result was that Cherubin and Hecquet engaged in as little discussion as possible, even when it came to important aspects of the flight.
Air Inter flight 148 left Lyon at 5:20 p.m. and headed north on the short, 50-minute flight to Strasbourg. Weather in Strasbourg was overcast with rain, transitioning to snow at higher elevations. The first part of the flight proceeded normally until it neared Strasbourg and the crew began to plan their descent. Strasbourg Airport had two runways: runway 05, which ran southwest to northeast, and runway 23, which ran the opposite way. Only runway 23 had an instrument landing system (ILS) that could automatically guide the plane all the way down to its threshold; runway 05 lacked such a system. However, at that time the active landing runway was 05, which required a so-called “non-precision approach” in which the pilots would need to manually work out the descent profile. Pilots at Air Inter rarely flew non-precision approaches, and Hecquet was clearly not comfortable doing one. Instead, he planned to begin the approach as though he were landing on runway 23 so he could use the ILS, then loop around to land on runway 05 once he had the airport in sight.
However, Cherubin assumed that they would execute the non-precision approach into runway 05 and began configuring the computer accordingly. Only after Cherubin had done this did Hecquet mention that he wanted to start with an ILS approach to runway 23, at which point he overrode Cherubin’s previous inputs. Cherubin responded by saying, “I don’t know why you don’t try a 05 VOR DME [non-precision approach].” Hecquet responded that such an approach would require a loop back away from the airport that would add ten minutes to the flight, although this was not in fact true.
The air traffic controller also assumed that flight 148 would be performing a non-precision approach into runway 05 and began giving them instructions accordingly. The controller cleared the flight to proceed to the ANDLO waypoint, which is in line with runway 05 and is not part of the runway 23 ILS procedure. Captain Hecquet exclaimed, “ANDLO, now they’re starting to mess me about,” but he complied anyway, since an approach to runway 23 didn’t necessarily conflict with an instruction to overfly ANDLO.
When they reached this waypoint, the controller cleared flight 148 for a non-precision approach to runway 05. However, because the crew never intended to land straight onto runway 05, they were way too high to actually perform such an approach. Only now did First Officer Cherubin inform the controller of their actual intention. The controller replied that if flight 148 performed an ILS approach to runway 23 followed by a visual approach to runway 05, they would have to hold because of traffic waiting to take off on runway 05. In light of this information, Captain Hecquet abandoned his plan and agreed to loop back around to ANDLO for a non-precision approach to runway 05, as both Cherubin and the controller had originally suggested. This would allow flight 148 to get in before the departing planes.
Under the controller’s guidance, flight 148 made a swooping 180-degree left turn short of the airport. While heading back away from the airport towards ANDLO, the crew prepared for the approach by re-programming the computer and calculating the appropriate descent profile. Captain Hecquet calculated that once aligned with the runway, they would need to descend at an angle of 3.3 degrees.
At the end of the outbound leg, the controller instructed flight 148 to turn left onto a heading of 090 degrees, or due east, followed quickly by a command to turn to 051 degrees (northeast) in line with the runway. These commands were somewhat hasty. By obeying them immediately, Captain Hecquet completed the turn too early, passing ANDLO on the left instead of overflying it.
First Officer Cherubin noticed this and said, “We’re going… you’re turning inside, look.”
Hecquet didn’t respond, so Cherubin repeated his comment. “You’re turning inside, look! You should have rolled out on 070 [degrees].”
“Yeah, yeah,” said Hecquet, switching to a more easterly heading to get back in line with the runway radial. Flight 148 was now headed toward the airport but remained slightly to the left of the approach path, a course which would take it over the cloud-shrouded top of 2,710-foot Mount Sainte-Odile. The normal approach skirted slightly to the right of this high terrain.
At 6:19 p.m., the crew began the next stage of their descent down to the runway. To start descending, Captain Hecquet had to enter his previously calculated flight path angle of -3.3 degrees into the flight management system. To understand what happened next, a brief explanation of the Airbus A320’s descent modes is necessary.
The A320 flight computer had two navigation modes. The first, called Heading/Vertical Speed, or HDG/VS, allowed pilots to control the plane by entering a compass heading and a desired rate of descent (in hundreds of feet per minute). The second mode, called Track/Flight Path Angle, or TRK/FPA, allowed pilots to create a specific vector, or track, along which they wanted the plane to fly. Up to this point, Hecquet had been using HDG/VS mode to steer the plane onto the various compass headings given to him by air traffic control. Now, he needed to put the plane onto a vector that would take it to the runway threshold, including a flight path angle of -3.3 degrees. This required pressing the mode change button to switch to TRK/FPA mode, then entering -3.3 degrees using the input knob.
In what would prove to be a catastrophic error, Captain Hecquet forgot to press the mode change button, leaving the flight management system in HDG/VS mode. He then used the input knob to enter -3.3 degrees. But because the system was in vertical speed mode, his entry of “-33” was read as “-3,300 feet per minute” instead of “-3.3 degrees.” A small “V/S” indication near the input value revealed the current descent mode, but it could be hard to see from certain angles. Neither pilot noticed the mistake, even though -3,300 feet per minute is a blistering rate of descent rarely seen during normal flight.
An astronomically unlikely coincidence completed the sequence and sealed the fate of flight 148. If a pilot requests a change in descent rate larger than a certain threshold, the A320’s computer will incrementally work up to the requested change instead of responding instantaneously. This creates a smoother flight profile and prevents large, jerky movements. However, if the plane is climbing and the pilot commands a descent, or vice versa, the computer interprets this as an emergency situation and gives the pilot greater authority by using a higher threshold and larger increments, resulting in more rapid compliance with the desired rate of descent.
At the exact moment that Captain Hecquet selected the descent rate of -3,300 feet per minute, a pocket of turbulence caused the plane to climb for about half a second. Because the plane was technically climbing at the exact instant that Hecquet requested this steep descent, the computer interpreted this as an emergency and pushed the plane toward the selected rate more quickly than it otherwise would have.
Three factors came together in that moment. The plane was off course, the captain selected a descent rate that was much too steep, and an obscure safety system reduced the delay in reaching said descent rate. The combination of these three elements put flight 148 on a collision course with Mount Ste.-Odile, while the absence of any of them would have allowed the plane to clear the mountain safely.
Still, a crash was not inevitable. At any moment, one of the pilots could have noticed that the descent mode was wrong, that their altitude was too low, or that their descent rate was abnormally high. However, several factors conspired to prevent the pilots from noticing the problem. The first was simply their inexperience on the Airbus A320. They were still adjusting to the A320’s advanced digital cockpit displays and had not yet had time to develop an instinctive understanding of what “looked right” and what didn’t. This meant that monitoring the automation required the pilots to not merely look at the values displayed on their instruments, but consciously perform mental assessments of their significance that become instantaneous and automatic only once a pilot has accrued several hundred hours on the type. In the high-workload environment at the beginning of the steep descent, the subconscious prioritization of tasks pushed this sort of monitoring aside, in a phenomenon known as task saturation. The human brain can only keep track of so many tasks at once, and in a moment where the pilots were focusing on lining up with the runway, extending the landing gear, setting the flaps, and communicating with air traffic control, monitoring took a back seat.
Over the next 60 seconds, both pilots missed all the aforementioned clues that they were descending too fast. First Officer Cherubin also failed to make the required altitude callouts as the plane descended. This is probably because he was focused on the plane’s lateral position. The pilots knew they were off course, and Cherubin was worried that Hecquet was not working hard enough to get them back on track. By preoccupying himself with the lateral position, he neglected to think about their vertical position. 40 seconds after the descent began, Hecquet commented, “We’ll have to watch it doesn’t descend.” However, before he could take this thought any farther, Cherubin once again redirected his attention to their lateral position. Flight 148 continued to descend like a bat out of hell, headed straight for the fog-bound summit of Mount Ste.-Odile.
At this moment a ground proximity warning system could have saved everyone, but flight 148 didn’t have one. France was one of the only countries that didn’t require such systems, and Air Inter was one of the last major airlines that hadn’t installed them. As a result, the crew received no warning whatsoever that they were about to crash. When the mountaintop loomed out of the fog a split second before impact, there wasn’t even enough time to scream.
Flight 148 plowed headlong into the forested mountainside, careening through dense pines for several hundred meters. The cockpit disintegrated, the wings broke off and caught fire, and the front of the fuselage peeled like a banana and folded under the plane. The remainder broke apart and tumbled through the snowy forest until it finally came to a stop, surrounded by flames. The crash instantly killed both pilots and most of the passengers, but remarkably some people did survive. One man seated in row 14 was thrown out of the plane and regained consciousness still strapped into his seat, embedded in the snow. Several more passengers and a flight attendant at the back of the plane escaped the worst of the impact, finding themselves with their seats still attached to the floor even though the roof and walls had been torn away. Fire immediately bore down on what remained of the passenger cabin, forcing the passengers to flee. One man recalled grabbing a 9-year-old boy and jumping out of the plane with him. Two seriously injured passengers were unable to move, and they burned alive as the fire overtook them.
Eleven people initially survived the crash, including a 2-year-old girl, who escaped any sort of injury and was found wandering nearby. Most of the survivors gathered at the burning tail section, using the fire to keep warm amid sub-freezing temperatures. They expected help to arrive in 30 minutes to an hour — but it did not.
Meanwhile, as soon as the controller realized the plane was missing, an emergency plan sprang into action. A search began near the plane’s last reported location, and 1,000 people set out on foot to find it. A group of amateur radio enthusiasts was recruited to search for the signal from the plane’s emergency beacon, but the transmitter had been destroyed in the crash and did not broadcast. Furthermore, no one had been out on the top of Mount Ste.-Odile to witness the accident. Emergency services, journalists, and curious civilians all descended on the area in a fruitless search for the plane.
As the survivors grew weaker, some of them set out to get help. Among them was Nicolas Skourias, a graduate student who had suffered only minor injuries in the crash. On his way down the mountainside he ran into a television crew who were out looking for the plane. He guided them back to the crash site, whereupon they alerted the authorities to its location, more than four hours after the crash. Emergency services scrambled to evacuate the injured while reporters interviewed survivors at the crash site on live TV.
Two of those evacuated from the site soon died of their injuries, bringing the death toll to 87 with 9 survivors, including one flight attendant and two young children. All but one of these were seated at the back of the plane. As the survivors recovered in the hospital, investigators launched an inquiry into the crash, while the police began a parallel judicial investigation. The police, suspicious of the civilian investigators, initially barred them from entering the site until they had carefully documented everything. As a result, the black boxes were exposed to an intense fire for a long period of time. By the time investigators got their hands on them, the flight data recorder had been destroyed, and the cockpit voice recorder had suffered some minor damage. Fortunately, investigators were able to retrieve some flight data using the quick access recorder, a device normally used by maintenance workers for diagnostic purposes. Although it was not designed to withstand a crash, it survived mostly intact, and was able to stand in for the ruined FDR.
What investigators found was an unfortunate set of freak coincidences and human errors that sent flight 148 into a mountain. The controller’s error put the plane off course. Then, with the help of a safety system, the captain accidentally sent the plane into a steep descent by using the wrong computer mode. However, the systemic causes of the crash went far deeper than these discrete events. Investigators discovered that mix-ups involving Vertical Speed mode and Flight Path Angle mode on the A320 were very common, especially during training, and to a lesser extent in the course of normal line operations. There had been several incidents at Air Inter in which pilots accidentally selected a steep descent rate after using the wrong descent mode, only to discover their mistake when they broke through the bottom of the cloud base. In the case of flight 148, the only difference was that the mountain was higher than the cloud base.
This represented a major problem with the ergonomics of the A320 cockpit. The mode change button was not located near the associated display or the input knob, making it easier to forget. Once the wrong mode had been selected, it was difficult for pilots to detect this mistake directly. To make matters worse, the infrequency of non-precision approaches used by Air Inter meant that its pilots rarely had to switch from VS mode to FPA mode.
Other factors also contributed to their failure to detect the error. First of all, the pilots were under a high workload, particularly in the lateral plane, because they repeatedly changed their approach plan. This was exacerbated by imprecise commands from the controller and a lack of communication between Hecquet, Cherubin, and the controller surrounding each individual’s intentions. The lateral position of the aircraft therefore consumed the pilots’ attention, to the detriment of other tasks. Secondly, the relationship between the pilots played a role. Due to their mutual dislike of one another, they spoke as little as possible, causing each man to perform his duties without telling the other what he was doing. This also contributed to Cherubin’s failure to make the required altitude callouts. And third, Cherubin’s comfort in the A320 likely allowed him to place too much trust in the plane’s automated systems, leading him to deprioritize monitoring tasks. He fully trusted the computer to handle their vertical profile and did not expect a mistake by Hecquet to compromise its ability to do so.
Investigators also identified deficiencies in the outlook of Air Inter and in French aviation more broadly. Air Inter’s emphasis on punctuality may have pushed the pilots into accepting an approach that dramatically increased their workload, and quite probably affected the controller’s decision to offer this approach to them in the first place. Investigators also found that Air Inter pilots generally did not make all the required callouts, a sign that the airline’s cockpit culture had been allowed to drift away from standard operating procedures. Its pilots were also not well trained to handle non-precision approaches. And finally, Air Inter had no system capable of aggregating pilots’ reports of in-flight incidents and analyzing trends in this data to find areas that should be improved. This allowed repeated incidents of incorrect descent mode selections to go largely unnoticed.
Major systemic flaws also allowed Air Inter’s fleet to operate without ground proximity warning systems. After an accident in the US state of Virginia in 1974, the United States was the first country to mandate such systems. In 1979, the International Civil Aviation Organization observed the success of this policy in the US and mandated it globally. However, ICAO doesn’t have legal supremacy over sovereign countries, and France did not bring its regulations into compliance with the ICAO rule, citing the number of false alarms produced by the systems. France’s regulatory authorities also disliked that Sundstrand had a monopoly on the technology. However, in the early 1990s France finally mandated that all airplanes have GPWS technology with a deadline in 1992. In December 1991, Air Inter still had not begun the process of installing GPWS on its planes, prompting a strongly worded letter from the head of the French Aircraft Operations Bureau asking why they had not done so. Air Inter never responded to the letter.
The investigation looked into the abnormally long time taken to find the crash site and identified several deficiencies there as well. The entire search and rescue operation was predicated on the assumption that it would be easy to find a crashed airliner in a populated area — an assumption which proved not to be true. Initially, authorities received the specific location where the plane was last seen on radar, then expanded the search area to 21 square kilometers around this point, only to reduce it back to the original relatively small area later. This meant that a large portion of the manpower searching for the plane was allocated to areas that probably could have been ruled out from the beginning. A more effective strategy is to start at the flight’s last known location and work outward until the plane is found. Investigators noted that if medical assistance had arrived sooner, two additional passengers might have survived.
In its final report, the commission of inquiry made a long list of revolutionary recommendations to ensure that such an accident never happens again. These recommendations included the following:
1. Operators should have methods to ensure that two inexperienced pilots are not placed together. As a result of this recommendation, Air Inter forbade pilots with less than 300 hours on type from flying together.
2. Pilots should undergo a minimum amount of training on non-standard approaches greater than that which they currently receive.
3. When introducing a new aircraft with a new design philosophy, operators should teach pilots about that philosophy and how it will affect crew communication and allocation of tasks. As a result of this recommendation, Air Inter began training pilots on these topics.
4. Pilots should be trained in crew resource management (which was not previously required in France).
5. All aircraft in France with capacity for 31 or more passengers should have a ground proximity warning system, effective ASAP. Air Inter installed such systems on all its planes by the end of 1992.
6. Regulators should look into the possibility of installing warnings at airports that will alert controllers when a plane descends below the minimum descent altitude.
7. Regulators should look into the possibility of making flight recorders even more fire resistant.
8. There should be a recording device that records pictures of the cockpit instruments in flight, to be used in the investigation of accidents. (A novel and interesting idea, but it has never been implemented.)
9. Investigators should have the legal authority to ensure the safety of flight recorders immediately, without waiting for police to give them permission.
10. There should be clear and obvious differences in the display when in different descent modes. The input values should be displayed with their associated units of measurement, instead of simply a number. (The A320 display was modified so that a descent rate of -3300 shows up as “-3300” instead of simply “-33.”)
11. Presentation of information regarding an aircraft’s vertical plane should be just as prominent as that regarding its lateral plane, to encourage pilot awareness of altitude and descent rate.
12. Certification rules regarding the ergonomics of cockpit instruments and automation should be improved.
13. Emergency locator beacons should be stronger.
14. Emergency response plans should be updated to ensure that radar track information is presented to the relevant authorities within half an hour of the initiation of the search for a missing aircraft.
15. Operators should have programs in place to analyze flight data for trends affecting safety, and should inform regulatory authorities of any significant findings.
The crash of Air Inter flight 148 represented a watershed moment in French aviation safety. The accident spurred French regulators to finally implement numerous policies that other countries had long since adopted. It also brought about a re-imagination of how airlines ought to handle the giant technological leap to the Airbus A320, and forced Airbus to rework how the A320 presented information to pilots.
On the surface level, it is easy to narrow one’s understanding of the accident solely to Captain Hecquet’s failure to change the descent mode. But the reason this article is 4,000 words long and not 1,000 is because the Air Inter disaster is a microcosm of all the factors affecting human error in aviation. It is the perfect case study of how humans interact with their environment in ways that the creators of that environment, and indeed the human actors themselves, never intended. It touches on fundamental human weaknesses that all too often were not taken into account when designing procedures and systems. It teaches important lessons about information presentation, task saturation, and trust in automation. For these reasons, the crash of Air Inter flight 148 remains highly relevant to this day.
________________________________________________________________
Join the discussion of this article on Reddit!
Visit reddit.com/r/admiralcloudberg to read 100+ similar articles.
