- [Petter] This incident sequence
lasted a total of 31 seconds, about the same time that it takes
to drink a tall glass of water. During these 31 seconds,
years of delayed legislation, a deceitful resume, personal shortcomings
and an incredible chain of events came together in the worst possible way. Stay tuned. - [Radio Altimeter] 100, 50, 40, 30, 20, 10. - [Petter] Atlas Air Flight 3591
was scheduled as a domestic cargo flight going from Miami International Airport to George Bush Airport
in Houston, United States. The flight took place
on the 23rd of February, 2019, but the complicated circumstances that led up
to what ultimately happened to the aircraft had started more than a decade earlier. Because in response
to several accidents and incidents that had happened during the 1990s,
caused in part by pilot performance issues, a law called
the Pilot Records Improvement Act or PRIA had been implemented
in the United States back in 1996. This law forced all airlines to review
the training and employment history of any pilot they wanted to hire. And the background check needed to cover
at least the preceding five years prior to the start of the new employment. The idea with this law
was to highlight any training deficiencies that might have been noticed
during the pilot's initial training or with their previous employers. But it soon became clear
that the law lacked some important details. After the horrible accident
of Colgan Air Flight 3407 in February of 2009, where the pilot flying
made some very basic handling errors, it was found that he had been struggling
during his initial training and that he had suffered
multiple checkride failures. Now, this was exactly what PRIA
was designed to pick up, but the background check
had failed to do so. This led the National Transportation Safety Board
to recommend that PRIA should be amended by the implementation
of something that they call the Pilot Records Database or PRD. In this database, the training information
of all pilots should be automatically uploaded and easily searchable with the help
of the pilot's license number, making the background check easier to do
and much more accurate. The United States Congress
approved this suggestion and they mandated the FAA
to implement the PRD within a reasonable time frame. But that time frame
just kept getting longer and longer without the implementation of the law
getting any closer. Finally, a deadline for the implementation
was set to the 30th of April, 2017. But sadly, that date came and went
without anything happening. 40 days after this deadline had expired,
the pilot who was going to become the pilot flying in this story
was hired by Atlas Air as a first officer on their Boeing 767 fleet. So, why is this important? Well, the first officer
was 44 years old at the time and had a total flying experience
of just over 5,000 hours. He had a very troublesome
training history behind him, and as an example, he had resigned
from two of his previous airlines, Air Wisconsin and CommuteAir
because he was unable to get through their initial training. But in his application to Atlas Air,
he had conveniently failed to include those airlines in his resume,
stating that he had instead been studying for a college degree during those years,
where his employment had a noticeable gap. This was partially true. He had been studying, but he had also been failing
his initial training in both of these airlines
during that time. And one of those employments
was well within the five-year scope of the PRIA legislation. He did disclose a failure of an oral exam
with Trans State Airlines, but he had missed to disclose the subsequent failure of his ERJ145
air transport license upgrade. And he also later failed
a line check that wasn't mentioned. But what should have been
maybe the biggest red flag in his training history on his resume, which he, by the way,
also failed to disclose, was his unsatisfactory attempt to upgrade to a captain
on the ERJ175 with Mesa Airlines, which happened just a few months before he was doing
his interview with Atlas Air. These later discovered training records
showed several very worrying tendencies, including that when he was faced
with a situation that he wasn't expecting, he would sometimes react by just randomly starting to push buttons
and make inputs on the flight controls in order to be seen as doing something
rather than taking the time to properly evaluate the situation
together with his colleague. And that's worth remembering. Now, this all pointed
towards an individual with personality traits
not suitable for a commercial pilot. But he was able to mask
these deficiencies on interviews mainly because of the PRIA legislation
which was built on the honesty of the applicant. It was very hard for the airline
to find any records that the pilot didn't voluntarily disclose
and that was exactly why the Pilot Records Database
was designed to sort out. But there was, however, one note
in the documentation from the pilot that should have stood out
to the recruitment team of Atlas Air. That note did come from his employment
at Mesa Airlines, and it stated that the pilot was returning to first officer
after his command upgrade course had finished. This meant that the first officer
managed to get through the interview, which by the way, didn't include
a simulator assessment. And he then started his training
on the Boeing 767 with Atlas Air in the summer of 2017. Immediately, there were signs of trouble
as he was struggling to understand the performance calculations
and airplane systems during his ground school training. He received some extra training for that
and eventually passed his oral exam, which enabled him to continue
to the fixed base part of the type rating. The type specific training
that pilots have to go through generally involves a ground school course
and then a fixed base simulator training which is heavily dedicated to learning
scan flows and normal procedures. And then finally, a full flight part
in a moving simulator where emergency and handling scenarios are practiced. In the fixed base training,
his struggles continued as he couldn't get a grasp
of the normal procedures which led to further extra training
before he was allowed into the full flight part. But only after a few sessions
in the full flight, his sim partner had had enough
and he complained to the instructors that he was being held back
by the first officer. This led to his full flight simulator training
being paused and then he restarted it from the very beginning again
a few weeks later with a different partner. Unfortunately, after he had started again,
the training was then disrupted by the effects of a hurricane
that lasted for a few days. And after that, he failed
his first type rating skill test on the 767. The reason stated for his failure
by the examiner was unsatisfactory performance
on crew resource management, threat and error management,
non-precision approaches, steep turns and just general judgment. The type rating examiner did say though, that the first officer
was very nervous on the checkride. And after another retraining session,
he finally passed the type rating course and began his flight training
in the real aircraft. So, should this lackluster performance
have raised any red flags then? Well, the training manager of the airline
had looked over the first officer's file, but concluded that he didn't need to be put
under specific performance surveillance, partly because of that hurricane interruption
which had disrupted his training. At the time, it was thought
that this disruption might have been the reason
for his checkride failure. Now after these initial issues,
the recurrent training and check-in during the subsequent two years
had gone quite well with satisfactory results up until the day of Flight 3591. So that was the story of the first officer
but what about the captain then? Well, the captain was 60 years old
and had been working for Atlas since 2015. He was quite experienced,
with just under 11,200 hours of total time and 1,252 of those hours
were flown on the Boeing 767. His training history was much better,
but he had been under performance observation after a failed training day when he was still a first officer
back in 2015. This performance observation protocol
was finished about a year and a half later with satisfactory results. And he was then successful in his command upgrade training
in August, 2018. Now, it should be said very clearly here
that a single training failure can happen to anyone
for a variety of reasons. Things like personal circumstances
can give anyone a bad day. So having failed a single checkride
doesn't necessarily mean anything. But repeated failures
is a completely different matter. In any case, these were the two pilots who were scheduled
to fly Atlas Air Flight 3591 together on the 23rd of February, 2019. They had flown together before
on the previous day, during which, they had been flying a night duty. But this flight was scheduled
as a nice 1100 hour local departure with a flight time
of only 1 hour and 45 minutes. So, it was potentially,
a quite relaxing day out in other words. The pilots met up before the flight and started looking through
the pre-flight documentation as the cargo that they were going to be carrying
was being loaded onto the aircraft. There were no particular NOTAMs
that were planned to affect them, but the weather around Houston
looked a little bit challenging. A cold front was forecasted
to be passing over the area from the northwest. And when colder air moves in
towards a warmer, humid air mass, it wedges itself under the warmer air
and forces it upwards. This can cause the formation
of some really heavy rain showers and even thunderstorms. But the good news here
was that the worst weather should have passed by the time
that they were planning to land. They discussed this and then decided
on the final fuel that they would carry and then walked out towards the aircraft. Everything was being loaded
according to standard operating procedures and the weight and balance of the aircraft
was well within limits. On the walk out towards the aircraft,
the crew was joined by a third pilot who wanted to travel with them
on the jump seat over towards Houston. This was obviously not a problem
as there was plenty of jump seats available on the cockpit of the 767,
so he was welcome to tag along. The first officer
was going to be pilot flying for the flight, so, after the captain had verified
the technical status of the aircraft which was completely clear of faults,
the captain went outside to complete the walk around. And the first officer started working
on the cockpit setup and prepared for the departure. The aircraft that they were going to fly
was a 27-year old Boeing 767-300 which had been converted into a freighter
two years before the flight. It was in good condition
and had, like I mentioned before, no technical problems recorded on it. Once the first officer was finished
with his cockpit setup, he checked the time
on his nice, large wristwatch and saw that they were closing in
on their scheduled departure time of 11:00 and he let the captain know about this. The final checklist was completed
and the aircraft pushed back, started its engines
and taxied out for departure. At time 11:08, Atlas Air Flight 3591
took off normally from Runway 09 in Miami. The initial climb and cruise sections
of this flight was completely uneventful and the aircraft leveled off
at its planned cruising level of flight level 400 or 40,000 feet. Like I mentioned earlier,
this was a relatively short flight, so when they started to get closer
to their top of descent, the captain left the active ATC frequency in order to listen and record
the latest weather for Houston. The ATC information indicated
that they could expect one of the westerly-facing runways,
most likely Runway 26 Left or 27, which would give them
the quickest taxi route in towards the stand. But the actual landing runway
would be given to the crew later when they were in contact
with the approach ATC sector. The weather was OK, with some light to moderate rain showers
in the area and some light winds from a westerly direction. Once this information had been received, the first officer handed over the controls
to the captain and started setting up the FMC for the expected arrival and approach. They discussed using auto brake setting 2
and landing flaps of 25 degrees. And with reported winds
of 320 degrees at 14 knots, the crew decided to use a fly speed
of 135 knots for the landing. As they were discussing this,
air traffic control came in and advised them that after they had passed
the an-am point called GIRLY, they could proceed with the LINKK ONE
arrival route towards Houston. And this matched very well
what they had prepared to do. The problem though,
was that that cold front that had been forecasted to pass through was now situated
between the aircraft and the airport. The frontline was full of active storm clouds
and this would likely force the pilots to have to ask for radar vectors later on
away from the arrival route in order to make sure that they wouldn't have to cross
those storms. And this will become very important
very soon. At time 12:07, the Houston controller
called the aircraft up and gave them their first descent clearance
to flight level 350. This was promptly read back by the captain
and then the aircraft started descending. During the following minutes,
the pilots discussed several waypoint speed and altitude restrictions. And it seemed to be some confusion as to how these restrictions
should be interpreted. Sometimes a certain restriction
is only applicable if the aircraft is following a specific arrival,
but not if it's following another arrival. And this discussion took quite some time
during the initial descent. Air traffic control continued to give
further descent clearances. And at time 12:12,
the first officer handed over the controls to the captain
and started his approach briefing. And speaking about briefings,
here comes one from my sponsor. I just want to tell you about
a fantastic streaming platform that I'm sure that most aviation enthusiasts
among you will absolutely love and that's Curiosity Stream,
today's sponsor. For a really affordable price,
Curiosity Stream gives you access to thousands of award-winning documentaries
and series anytime and anywhere. With Curiosity Stream,
you can dive into topics like the evolution of the jet engine,
the secrets of the SR-71 Blackbird, and the story
of the Wright Brothers' first flight. You can also stream
thousands of other shows, covering everything
from science and technology to history, culture and sport. With new content added every week,
I find that there's always something new to discover on the platform. I recently watched
an incredible documentary which takes you on a journey
through the history of flight, featuring some rare archive footage
and interviews with aviation pioneers. So, if you are ready to take your love
for aviation to new heights, then head over
to curiositystream.com/mentourpilot or scan this QR code on the screen here. If you use the promo code mentourpilot,
you will save 25% off your purchase, making it crazy good value. Thank you Curiosity Stream.
Now, let's continue the story. As the aircraft was descending down
towards Houston, the crew was told to switch the radios
over to Houston Center on frequency 133.8. The captain read back this new frequency,
but waited with calling up the next controller until after the first officer
had finished his approach briefing, taken back the controls and the descent checklist
had been completed. The new controller
confirmed the arrival route and the captain asked
which runway they could expect. Now, the controller responded
that he didn't have this information and that seemed
to slightly annoy the captain who was now trying to plan
his descent down towards the airport. Knowing which runway to expect
is quite important since approaching different runways
will mean a different amount of track miles and that will have an effect on how quickly
an aircraft needs to descend. Descent planning in large aircraft
is all about energy management. The more track miles you have to fly,
the slower you need to descend to get rid of the energy. But if you plan
with having a lot of track miles and then you suddenly get a shortcut
to a closer runway, you might have to descend much quicker,
potentially even having to use the speed break or even flaps to achieve this. The other issue with not knowing
what runway and approach to expect is that the pilot monitoring might potentially
have to go into the FMC CDU at a later stage and reprogram it to a different approach
than what they had previously briefed. If that happens,
this will take the pilot monitoring away from his task
of monitoring the flight progress, and this will soon become important. During the next couple of minutes, the pilots were discussing different speeds
to be used during the descent. The captain wanted to fly quicker,
as we often do, but the first officer was pointing out
that there were speed restrictions as part of the arrival route. So an increase in speed
might not be a great idea as they would have
to decelerate soon anyway. These are all quite standard discussions
that you would hear in any cockpit. And it seems like the CRM
was working quite well at this stage, with just some standard banter going on. At time 12:30:13, the aircraft was told
to switch over to the Houston approach controller. Now, they would finally be told
which landing runway they would get. The controller welcomed them
to the new frequency and then gave them the name of the current ATIS information
as well as the altimeter pressure setting. He then told the crew
to fly the published transition route for Runway 26 Left. As soon as the captain heard this,
he dived into the FMC CDU and started setting up the final part
of the routing to the approach. Once that was done, the crew set their altimeters
to the current pressure setting and the captain started actuating
the approach checklist. When that checklist was completed,
the first officer started complaining about how he didn't think
that the virtual navigation or VNAV of the autopilot
was doing a great job. He was expecting it to stay a bit higher
than they currently were and to start decelerating earlier. And these comments
led into a discussion between the pilots of how little they trusted the automatics. It is likely that VNAV was actually doing
what it could at this point. But every time
that the pilots updated routing or set another speed restriction,
VNAV would have to reprogram its calculated descent path,
leading to fluctuations on how it flew. At time 12:34:08, the approach controller
called the crew up and advised them about a line of heavy precipitation
coming up just in front of them. This was the cold front
that they had seen on the planning stage. And the controller told them to advise him
if they needed vectors to avoid. Basically, turning around the storms
in front of them. And it is at this point
that the actions that would eventually lead up to the accident really start. Because as the crew were now looking
at their weather radar, trying to decide which way to turn
to avoid the storms, they were now handed over
to the next approach controller. When the captain checked in,
the new controller wanted to know if they needed vectors
towards the east of the weather in front, which would give them
a significantly longer route, or if they had any other intentions. The captain told the controller
to stand by. And now the first officer
suddenly makes a statement about his instruments. It wasn't possible to hear
on the cockpit voice recorder exactly what the problems
with his instruments were, but the captain responded with a quick laugh
and that was then followed by the first officer handing over
the controls to the captain in an attempt to try and quickly fix
whatever that problem was. As soon as the captain
had taken the controls, the first officer reached over and pushed
the Electronic Flight Instrument switch on the right side of his instrument panel. This switch was used to change
the source used for the instrumentation on his side from a primary
to an alternate source. Apparently, it was a well-known fix
on the Boeing 767 that if the displays would start acting up
for whatever reason, pushing the EFI switch momentarily
would reset the fault and then the switch could go back to the normal position again
with everything sorted. As the first officer was doing this,
they continued to discuss which the best route for them would be
around the storms. The first officer was handling the radios
as he was trying to fix his instruments and he now asked the controller
if they could turn to the west to avoid the clouds that way. This would give them
the shortest track to fly and likely save them
quite a few minutes. The controller came back and said
that the only problem with that vector was that there was
a lot of departing traffic in that sector
which they would have to avoid. The pilots discussed this
and they then responded that it was fine with them
to go to the east instead because they had loads of fuel
on board anyway. But the controller that had heard
their first request had now figured out a way for the aircraft to actually take that vector
to the west instead. But it would require them to keep
a higher descent rate down to 3,000 feet in order to keep out of the way
for the departing traffic. The first officer read this back
and the captain then set 3,000 feet on the mode control panel and selected level change
as the descent mode for the autopilot and the autothrottle. The EFI button had now been pushed
and it apparently fixed the problem because the captain laughed shortly
and the first officer responded that the EFI always fixes the problem. They now received an instruction
to turn left onto heading of 270 degrees. And after having set that, the captain handed back the controls
to the first officer. As soon as the first officer took over,
he remembered what the controller had said about keeping a high descent rate,
so he reached over and extended the speed breaks
into the flight E10 position. This raised the spoilers on the wings,
creating more drag and therefore,
increased the descent rate. Now, every time that the speed brake
is in use, the standard operating procedure at Atlas Air was for the pilot flying
to keep one hand on it, as a reminder to stow it
when it was no longer needed. The speed brake is situated
on the left side of the throttle quadrant. So, in order for the first officer
to hold it, he needed to reach under
and past the thrust levers. The captain was now heads down
into the FMC CDU again, trying to connect the route
in the most sensible way to get descent guidance from VNAV. And like I mentioned earlier,
this also meant that he likely wasn't looking at the instruments
at this point. He asked if he should connect the route
via a point called GRIEG on the approach and the first officer agreed and at the same time,
asked for flaps 1 to be extended to further increase the drag
and the descent rate. Flaps 1 was selected,
the routing towards GRIEG was executed. And the captain then called LNAV available,
which was not the right thing to do here since the aircraft
was on a righter heading and shouldn't follow that new route
that he had created. But the first officer
still tried to connect the LNAV, but since the aircraft was not
on an intercept heading at that point, it didn't engage. This caused a bit of confusion
before they both realized the mistake that they had made. And this is a tiny little thing, but it shows some level
of degraded situational awareness here, possibly from the added stress
of avoiding the storms, departing traffic, and planning the routing
towards the approach. The aircraft was now descending through
around 7,000 feet and had just entered some clouds
associated with the front that they were just about to pass. This meant that the pilots were now flying
in instrument conditions without any outside references,
and at the same time, some light to moderate turbulence
started rocking the aircraft. At time 12:38:31
when the aircraft descended through 6,300 feet, a distinct click could be heard
on the cockpit voice recorder. Immediately after this click,
the thrust levers started moving forward from their idle position
towards almost full thrust and the nose of the aircraft
started slowly pitching up from an attitude of around minus 1 degree
to 4 degrees nose up. So, what was going on? Well, remember that big wristwatch
that the first officer was wearing? It is very likely that since he was holding
the speed brake lever with his left hand when the turbulence started,
that wrist watch came into contact with the TO/GA button, which on the 767 is situated on the lower side
of the thrust levers. That click that was heard
on the cockpit voice recorder was, therefore, likely the TO/GA button
being activated by mistake. Now, if the toggle switch was activated,
that would send signals to the autopilot to keep its current track,
maintain the speed and start to climb with at least 2,000 feet per minute. In any case, the TO/GA activation
was likely not noticed by either pilot at this stage because none of them
mentioned anything. Instead, the first officer
started to retract the speed break and move it into the down
and arm position, possibly as a reaction
to the thrust levers moving forward as that would have been the reaction to do
in case the aircraft was leveling off at the cleared altitude. Now, in order to understand
what is going to happen during the coming 31 seconds,
we need to examine the way that the human body reacts
to certain accelerations and movements. We, humans, use
basically three different ways to determine orientation
and movement within a space. The first way is with the help
of the vestibular system, which is situated in the inner ear
where the otolith organs senses direction and accelerations. The second are the nerves in the skin,
muscles and joints who feels their position based on gravity,
general feeling, and sounds. And finally, we use our eyesight
that uses visual inputs to make sense of all those other feelings. The problem is that the vestibular system
cannot distinguish between acceleration caused by gravity
and other types of accelerations, like that of an accelerating
or maneuvering aircraft, for example. And with the visual cues gone,
like in this case, where the aircraft was inside of a cloud,
a type of sensory illusion known as somatogravic illusion
can therefore start affecting the pilots. And it can happen very quickly
if we're not careful. This illusion is caused
by the effect of added thrust to the aircraft, which led
to a sudden acceleration. That acceleration
would have forced the pilots gently backwards into their chairs. And at the same time,
the aircraft started pitching up, causing exactly the same type of feeling. Inside the vestibular system,
both of these accelerations would have been added to each other,
and without a visual reference point, the brain of the first officer
would have interpreted these two accelerations
as one much more abrupt, sudden pitch up. This is actually exactly
the same type of illusion that we use in full flight simulators
to simulate accelerations and decelerations. Now, the correct way
to stop this illusion from progressing would have been to concentrate
on the artificial horizon in front of him, which would have accurately told him
that the pitch up was much smaller than the one that he was sensing. But remember, this pilot had a history
of reacting immediately under stress without taking the time to carefully evaluate
the situation that he was in. So, what happened now
was that the first officer started pitching the aircraft forward,
down towards the ground below. The autopilot was still engaged, but the force
of the first officer's pitch down command was enough to override it and the aircraft started
an accelerated descent. This then led to the next problem
which was a decrease in the aircraft's vertical acceleration
and a rapid increase in the longitudinal forward acceleration,
which likely, just made the somatogravic illusion even worse, causing a sensation
of tumbling backwards. A very nasty sensation. This feeling is known
as the inversion illusion, since the feeling is completely opposite
to what's actually happening. This is likely why the first officer
now cried out "Oh" and "whoa," as he started pitching forward even more. A sudden shock like this can trigger a fight or flight response
in the human brain, which will favor a strong physical reaction
but severely hamper the brain's ability to accurately assess
what really needs to be done. This often leads to impulsive
and incorrect actions known as a startle response,
which is exactly what happened here. As the pitch was now decreasing, the aircraft descent rate
kept increasing rapidly. An owl beeper warning
went off in the cockpit, likely caused by the autopilot warning
that manual input was overriding it. And it was followed by the first officer
calling out, "Where is my speed?" And, "We're stalling!" Now, there was nothing in this scenario
that should have led the first officer into believing
that the aircraft was stalling. The speed was well above
the stall onset speed and it was increasing rapidly. There was no stick shaker
and the red and black barber's poles on the speed display
were not anywhere near the speed that they were flying. The only possible explanation
for this call out was that the first officer felt
like the aircraft was violently pitching up and thus, possibly stalling. Again, a quick instrument scan
would have shown clearly what was going on and there is no indication
that there was anything wrong with his instrument at this time. Now, the combined forces and acceleration
caused by the sudden pitch forward, combined with the increasing acceleration
of the aircraft, would have now caused a force vector
acting on the first officer's inner ear similar to a vertical drop with the aircraft attitude
at the wings level position. Likely, a truly terrifying feeling. Now, if the first officer thought
that it was stalling, the correct initial action would have,
indeed, been to unload the wings by pitching forward. But the continuation of that procedure
should have been to recover the aircraft back towards wings level
once the stall indications, stick shaker or buffeting,
would have disappeared. But in this situation,
there were no actual stall indications other than what was
in the first officer's head. All of the things that have happened
up until this point, after the inadvertent TO/GA activation
had only taken around 13 seconds, which is about the same time
it has taken me to say this sentence. It all happened really quickly. But what about the captain then? What was he doing,
and why didn't he intervene? Well, when the first officer
inadvertently activated the TO/GA button, the captain was still working
on the route in the FMC CDU, and he also had just received a message from air traffic control
that he was responding to. This likely meant
that during those crucial first 13 seconds, he didn't monitor the instruments
and he didn't notice the extreme inputs that the first officer
was suddenly making. When the owl warning went off,
he would have looked up, trying to make sense of what was going on, but the same sensory inputs
that the first officer was experiencing would have affected him as well,
probably slowing down his assessment and leading to some initial
real confusion. If he would have been paying attention
to the instruments when this all started, he would likely have been able
to take over the controls and disconnect the automatics much sooner. But from the position that he was now in,
it would have all been sudden chaos and very hard
to initially correctly assess. The fact that he was head down
when this all started would have also explained
why he didn't notice that the Flight Mode Annunciator
had gone into go-around mode as soon as the TO/GA button
was accidentally pushed. In any case, at time 12:38:51 seconds, the aircraft had entered
into a terrifying dive with an attitude
of 46.1 degrees pitch down, and with that, a rapidly accelerating airspeed
passing 359 knots. That is above the max speed
of the aircraft and around 100 knots faster
than the maximum speed with flaps 1 extended. And just to give you some reference
of how serious this aircraft situation was, an aircraft upset situation in pitch, which we pilots are trained
to always try and recover from, is defined as the aircraft exceeding
either 25 degrees nose up pitch or 10 degrees nose down. In this case, the aircraft
had a nose attitude 36 degrees lower than what would be considered an upset,
descending through 3,500 feet with a vertical speed
of around 9,000 feet per minute. Here, the captain must have gained
his bearing slightly because some pitch up force was recorded
coming from the left control column. And he could be heard shouting,
"What's going on?" The thrust lever
was now also moved temporarily into the idle
and then pushed back up again, but we don't know which one of the pilots
who actually did that. But the fact that the captain
now started to try to pull up was not accompanied by a formal
my controls command, which would have been
the right thing to do. Instead, the two pilots
were now pushing and pulling in opposite directions. The first officer was unable to respond
to the captain's question about what was going on. Instead, he continued making
panicked call outs, which he continued to do
for the rest of the flight. Now, the reason that we know that the two pilots
were working against each other at this point
was because the flight data recorder started recording different angles
of the left and right elevator at the back of the horizontal stabilizer. Because there is a function that in case
one of the controls gets jammed will allow the other control
to override the clutch and control the elevators separately
to enable control of the aircraft in spite of the jam. This requires quite some force,
and that's exactly what was now happening. Unfortunately,
the captain's opposite commands were not enough to stop
the downward pitch that the first officer
had already put in motion. The vertical speed increased
to over 9,000 feet per minute as the aircraft descended
through 3,000 feet. So, why didn't the captain
just take over the controls? Well, in a rapidly developing emergency
like this, it would have been very hard for the captain to actually understand what was happening
and what was causing it. He would have seen that the first officer
was not incapacitated and that he was still inputting
on the controls. And because of that,
it would have been very hard for him to judge if the first officer
was trying to solve the issue or was actually causing it. Severe startle and surprise works opposite to logic
and clear decision making. And for the captain to go
from programming the FMC into grasping that the first officer was suddenly pushing
the nose of the aircraft down towards the ground
in just a few seconds was likely too much to ask for. So his delayed response was likely due to a combination
of startle, confusion and shock. In any case, the two pilots continued
to input opposite controls for a total of around 10 agonizing seconds before the pilot
on the jump seat called, "Pull up!" This likely happened as the aircraft
passed around 3,000 feet, which corresponded with the cloud base and below the clouds,
the visibility was good. This would have made
a full terrifying picture of their attitude and descent trajectory
instantly clear for both pilots. And from this point,
the first officer control inputs changed to full pitch up. Together with the captain's inputs,
they managed to get the nose up to around 16 degrees below the horizon,
causing a wing loading of around 4G's as they pulled with full force and with the overspeed warning
blaring behind them in the cockpit. Unfortunately,
this was all too little, too late. There just wasn't enough altitude left
to recover the aircraft. And at time 12:39:03, only 31 seconds after the TO/GA button
had been inadvertently pushed, the aircraft crashed into the marshlands
on the outskirts of Trinity Bay with a vertical speed
of 12,850 feet per minute and a total airspeed of 434 knots. The aircraft was immediately destroyed and scattered over an area
bigger than 12 acres of mostly shallow water
and muddy marshlands. And all three pilots immediately perished. Now, some of you might be asking, why weren't there any GPWS warnings
sounding in the cockpit? With that kind of a trajectory,
that warning should have activated early and might have pulled the first officer
out of his somatogravic illusion. Well, it turns out that the aircraft was equipped
with a Honeywell Enhanced GPWS system that was functioning well
during the accident. But the problem was that the angle
and extreme speed that the aircraft had when it got closer to the ground
caused the radio altitude data that was fed into the system to be flagged by the computer
as being unrealistic. That caused the computer to disregard
the radio altimeter for three seconds, which stopped the system
from ever issuing a warning. But even if it would have worked,
it wouldn't have made any difference. It was just too late at that point. The rescue work very quickly turned
into a recovery operation, since it was clear that it wouldn't be any survivors
after this crash. The aircraft had been reduced
to millions of pieces of very small debris that was spread out over a huge area. And it took specialized machines
several weeks to recover most of it and bring it into a hangar,
where the pieces were placed together as accurately as possible. Both the cockpit voice recorder
and the flight data recorder were eventually found. And even though the audio quality
wasn't great from the voice recorder, a picture of what had caused this crash
quickly emerged. The National Transportation Safety Board
concluded that the crash had been caused by the inadvertent activation
of the aircraft's TO/GA button, which had then led to a change
in pitch and acceleration that went initially undetected. These changes caused the first officer,
who was pilot flying, to experience somatogravic illusion and due to his documented
personality weaknesses in stress management
and situational awareness and just general pilot suitability, he was unable to recover
from the illusion and instead, pushed the aircraft into a dive
from which they were unable to recover. The final report were very critical
to both the hiring procedures in Atlas Air, who had failed to notice
the first officer's command upgrade failure just months before the recruitment. But most of all, the critique landed hard
on the FAA's failure to implement the Pilot Records Database
at the time of the accident. The first officer had clearly concealed part of his training history
with two other airlines. And had this database been in place,
a simple search on his license number would have revealed both the failures
and his attempt to conceal them. And that likely would have stopped him
from being hired in the first place. This was an accident
that really shouldn't have happened. And it highlights just how important it is
for airlines to vet their pilots carefully and for important
safety critical legislation to be implemented as quickly as possible. The Pilot Records Database
has since been implemented. And this accident
also led to some research into the possibility of implementing
the more advanced military GPWS system into civilian airliners. A system like that
would have been able to predict the trajectory of the aircraft earlier,
which would likely have been helpful. Now, check out this video next. Or if you're interested in seeing
more incidents and accident videos, then check out this playlist. Please consider supporting my work
by sending a super thanks, buying some awesome merch, or join my fantastic Patreon crew. All support is hugely appreciated. Have an absolutely fantastic day
and I'll see you next time, bye-bye.
