Mayday (2013) s10e06 Episode Script
Who's in Control?
WOMAN: (OVER PA) Ladies and gentlemen, we're beginning our descent towards Amsterdam Schiphol Airport.
You'll have to turn that off for now, sir.
Thank you.
NARRATOR: Turkish Airlines Flight 1951 is preparing to land in Amsterdam.
Amsterdam, Turkish 1951.
Descending to 7,000.
Speed - 250.
(BEEPING) The crew is flying a state-of-the-art Boeing 737.
Flaps - 15.
Localiser alive.
Localiser capture.
In the final moments of the flight, the landing turns into a catastrophe.
The plane falls like a rock.
The crash of Turkish Airlines Flight 1951 involves the most popular plane on earth.
With nearly 1.
5 million passengers boarding 737s every day, investigators need to figure out if the problem was with the plane or with the pilots flying it.
Mayday, mayday.
On the morning on February 25, 2009, Turkish Airlines Flight 1951 becomes the first plane to crash at Amsterdam's Schiphol Airport in more than 10 years.
(SPEAKS FOREIGN LANGUAGE) TRANSLATION: It smashed into the ground really hard.
It made a tremendous noise.
The plane hits the ground in a muddy field, just north of runway 18 right.
Since the crash was so close to the airport, rescue workers arrive quickly.
Most of the passengers have survived but many are badly injured.
Survivors are taken straight to local hospitals.
Images of the Amsterdam accident quickly spread around the world.
This is the third crash of a passenger jet in the past six weeks.
The sudden nature of this accident adds to the mystery.
It doesn't take long for the Dutch Safety Board to arrive at the scene.
They will be investigating this accident but they won't be alone.
The crash involved an American-made plane, so the US National Transportation Safety Board sends Joe Sedor, one of its most experienced investigators, to Amsterdam.
When it's a non-US registered aircraft that crashes overseas, such as this Turkish Airlines, we are the state of manufacture and design of the airframe, and also, in this case, the engines.
Fuselage in three large pieces.
Engines forward of the main wreck site.
Flight 1951 was one of the most advanced aircraft in the skies, the Boeing 737-800 series.
It's designed to travel longer routes at higher altitudes.
The new generation 737 is still the best airplane Boeing ever built.
We developed an airplane that had an improved wing, improved avionics, simpler systems that required less maintenance.
Investigators know this isn't just any plane.
The 737 is the world's best-selling commercial jet.
Finding out why this one crashed is imperative.
There are more than 5,000 of them in the skies.
They carry about 1.
5 million passengers a day.
Investigators must quickly determine if there's a flaw with the plane that could cause another accident.
What they know so far is that Flight 1951 had been travelling from Istanbul, Turkey, to Amsterdam.
There were 128 passengers onboard the early morning flight, including four engineers from Boeing.
WOMAN: (OVER PA) Ladies and gentlemen, we're beginning our decent towards Amsterdam's Schiphol Airport.
Please raise your seat backs to the upright position, and stow away your tray tables.
There was no mention onboard of any kind of trouble in the cockpit.
The crash has killed nine people, including three members of the Boeing team and the pilots in the cockpit.
See if you can get me some aerials of the crash site.
There are eerie similarities to another recent accident involving a Boeing aircraft - British Airways Flight 38.
SEDOR: Approximately a year before this, there had been a 777 short landing at Heathrow, which had a duel engine flame-out.
In that accident, a Boeing 777 fell to the ground short of the runway.
The British Airways pilots reported that both their engines stopped delivering power just before landing.
At the time of the Turkish Airways crash, the cause of that accident hasn't been found.
Like in that case, investigators have plenty of clues to work with.
The plane and its engines are largely intact.
The flight data recorder and cockpit voice recorder are found in good condition.
There are also plenty of survivors to describe what happened.
According to the passengers, the landing had been routine but then suddenly, the plane simply dropped out of the sky and hit the ground.
But perhaps the biggest clue comes from the crash site itself.
The wreckage is not spread out.
It tells investigators that the plane could not have been travelling forward at high speed when it hit the ground.
MAN: The way the aircraft had crashed, it did appear to be some sort of a landing accident in which there was moderate control of some sort.
The pattern of debris and the passenger reports point investigators to an immediate suspect - the engines.
The engines issue was a very big issue in my thought process at the time, initially.
There's no evidence of fire on the fuselage.
In many crashes, fuel in the plane's tanks ignites on impact.
The lack of fire raises an obvious question - did the engines stop running because Flight 1951 had simply run out of gas? NANCE: That was one of the first thoughts that I had was, "Did this airplane have fuel aboard?" Because otherwise how does a 737 literally fall out of the sky on approach to an airport? But the location and condition of the plane's engines suggests that perhaps they didn't quit in flight.
Sure looks like it was running.
SEDOR: When we first looked at where the engines ended up, the initial impression was that they probably were producing thrust at impact, given that they were so far forward of the main wreckage.
That was just a very general conclusion.
Only the flight data recorder can tell investigators how much power the engines were generating in the seconds before the crash.
Lots of fuel.
Rules that out.
It doesn't take long to discover that there was plenty of fuel in the plane's tanks.
Flight 1951 definitely did not run out of gas.
Passengers report that in the final seconds before the crash, the plane hit what felt like turbulence.
It points investigators to a well-known culprit.
A microburst.
A microburst is a powerful column of air that shoots out of storm clouds.
It can literally slam a low-flying plane to the ground.
NANCE: If an airplane flies into that at approach speeds, you're not going to be flying anymore.
You're going to come out of the sky.
Certainly it was one of the things that all of us took a look at at the beginning - was there a microburst? In 1985, a Delta Airlines flight was caught in a microburst while landing at Dallas/Fort Worth Airport.
The plane hit the ground short of the runway.
137 people were killed.
Investigators learn that there were heavy clouds above the airport at the time of the accident.
A powerful gust of wind may well have accompanied them.
The flight data recorder will have recorded wind speeds outside the plane.
Investigators will need to analyse the flight data to prove the microburst theory.
In the meantime, the rescue has led to an unusual discovery.
SEDOR: There was three pilots in the cockpit, which is unusual - this is a two-crew cockpit.
So why was that third pilot there? Since none of the three pilots survived the accident, it's all the more urgent for investigators to retrieve the data from the cockpit voice recorder.
It records conversations in the cockpit.
They're in luck.
Because of the way that the aircraft crashed, access to the recorders and the condition of the recorders was excellent.
The reason for the third pilot is soon uncovered.
Flying standard arrival route.
Visibility, 3,500m, expected to decrease to 2,500m.
For First Officer Murat Sezer, this had been a training flight of sorts.
He was new to the airline and was being shown the intricacies of landing at Amsterdam's airport.
MAN: In the Jeppesen charts, which is what all pilots use to navigate to and from Schiphol, there's 102 pages of information on Schiphol alone, so there's dozens of approaches.
Runway 18 right has 3 high-speed exits to the left.
The captain, Hasan Arisan, was doing double duty.
He was training his first officer Make small corrections as we get close to the runway.
.
.
and was in command of Flight 1951.
It's because Captain Arisan was teaching that there was the third pilot in the cockpit.
Olgay Ozgur was a safety pilot.
He was there to keep an eye on the flight's progress during this training mission.
SEDOR: The purpose of that second set of eyes is to make sure that the captain and the first officer - if they're in a situation where it's a little bit of a training portion of the flight - that they don't miss something.
We've got a clean recording.
The voice recorder reveals that the three crew members began preparations for landing when still above 8,500ft.
Amsterdam, Turkish 1951, descending to 7,000.
Speed - 250.
But the voice recorder has picked up an unusual sound - a warning that makes no sense at this stage of the flight.
One of the investigators from Boeing was an engineering pilot that came and helped with the CVR and he's listening to see when Or are there any unusual sounds that can be heard that would not be normal? MAN: (ON RECORDING) Feed OK for ILS 18 right.
Descent to 4,000.
ILS 18 right.
(BEEPING) The alarm keeps sounding.
It's the landing gear configuration warning horn.
Captain Arisan continually dismisses it.
Turkish 1951 descend 4,000.
ILS 18 right.
Landing gear.
Is that the landing gear warning? They're 8,300ft here.
The warning is a solid clue.
But investigators can't yet see how it could possibly have caused a crash.
The crew of Turkish Airlines Flight 1951 got warnings to extend their landing gear while still thousands of feet in the air.
On the initial listen, we heard a gear warning horn occur as the aircraft was approaching, when it was still up at .
.
and it was coming in at about 10,000ft and below.
Investigators now turn to the flight data recorder to help solve some of the mysteries surrounding this flight.
The analysis of wind speeds outside the aircraft is completed.
It's clear - none are drastic enough to have brought down the plane.
No evidence of a microburst.
But the flight data recorder does provide some valuable insight into the cause of the landing gear warning.
One of the instruments that measures altitude had the plane already on the ground.
When we looked and saw the radio altimeter data on the recorder, it said about 8,000ft and then immediately it went down to about -8ft.
-8ft is an indication that the aircraft is on the ground but of course it's still at 2,000ft.
The Boeing 737 is equipped with two separate altimeters.
One measures air pressure to determine the plane's height above sea level.
That reading is displayed prominently in the cockpit on both pilots' flight display.
MAN: Falling to zero, Sezer.
The plane is also equipped with a radio altimeter.
It's made up of four antennas.
Two transmit signals to the ground and two others read the signal that bounces back to determine the plane's height.
It's precise.
It's very, very precise.
Pressure altimeters can sometimes be not as accurate and radio altimeters are 100% accurate, if they're working properly.
One antenna feeds the reading to the first officer's display.
The other feeds the captain's instruments.
In the case of Flight 1951, the captain's side was wrong most of the flight.
Investigators go back over the CVR and make a puzzling discovery.
MAN: (ON RECORDING) Amsterdam, Turkish 1951, descending to 7,000.
Speed - 250.
Radio altimeter.
Captain Arisan seems to have known that the landing gear warning was being caused by a faulty radio altimeter.
The airplane thought it was low to the ground and the gear was not down and the captain recognised that the problem was really in the radio altimeter, showing him that they should be on the ground, and he goes, "It's just the radio altimeter.
" Throughout much of the approach, the captain's radio altimeter had been displaying a reading of -8ft, triggering the warning to lower the gear.
They treated it like it was a nuisance.
Turkish 1951, descend to 2,000.
2,000.
1951.
Investigators dig for any other abnormalities.
They learn that with flight 1951 still about 17km from the airport, controllers directed the pilots to begin their final turn to line up with the runway.
Turkish 1951, left heading 210.
Clear to approach.
18 right.
Left 210, clear ILS.
Turkish 1951.
210 set, sir.
This turn puts flight 1951 in line with runaway 18 right.
It's equipped with an instrument landing system which sends out a signal outlining the ideal descent path to the foot of the runway.
The autopilot follows that glide path until the plane is a few hundred feet from the ground.
Then the pilot takes over.
It makes landing almost effortless.
The ILS is pretty easy to follow.
It's a video game.
My daughter has flown in a simulator and can land a 737 using the ILS.
The crew begins configuring their plane for landing, unfazed by the warning horn that's repeatedly triggered by the malfunctioning radio altimeter.
Flaps 15.
10km out, Flight 1951 picks up the ILS signal that will guide the plane to the runway.
Localiser alive.
Localiser capture.
The safety pilot, Olgay Ozgur, now reminds Captain Arisan about the failed altimeter.
We have radio altimeter failure, sir.
OK.
Turkish 1951, runway 18 right, clear to land.
Clear to land.
Thank you.
Investigators are stumped.
The crew knew about the malfunction and continued their approach.
How had it then caused them to crash? Clearly, there was more to this accident than a faulty altimeter.
NANCE: The whole premise of airline safety, the way we build the airplanes, the way we fly them, is based on the idea that we can have any number of failures and we should still be able to arrive safely.
The radio altimeter is just one instrument.
There's no way in the world that that one instrument, if it fails, should be a major cause of worry that we're going to have a crash.
Investigators wonder if the crew had been given proper guidance for their approach.
They turn to exchanges between the pilots and the controller who guided them in.
They carefully review every instruction.
TWO-WAY RECORDING: Turkish 1951, descend to 4,000.
AIR TRAFFIC CONTROL: Speed OK for ILS 18 right.
Turkish 1951, descend to 2,000.
Turkish 1951, left heading 210, clear to approach 18 right.
By following the controller's instructions, the crew made their final turn much too close to the runway.
So to intercept properly, they should be here.
International guidelines call for approaching planes to intercept the signal that guides them to the runway from below.
It's so pilots don't have to make any drastic last-minute course corrections to get to it.
To intercept here, they had to descend.
But Flight 1951 was given instructions that brought it to the threshold of the glide slope while still way above it.
It's a common practice at Schiphol because it gets planes to the runway faster.
SEDOR: Because they were so close, they had to capture the glide slope from above.
Although it is an unusual situation, it is one that can be handled by a flight crew if it is managed properly.
Approaching a glide slope from above is more difficult, mostly because the crew has to suddenly slow the plane and descend rapidly to intercept the signal.
HUFF: We've also called this a slam-dunk approach and some pilots like it, some pilots don't.
It's a little bit harder and things happen quicker when you're above the glide path trying to intercept from above.
And it's just a challenge for a lot of pilots.
The approach from above increased the crew's workload but it's standard practice at Schiphol Airport.
I've flown into Schiphol dozens of times and I expect it.
If the controller's instructions had somehow overtaxed this crew, their conversations would indicate it.
They're just 5km from the runway.
- 1,000.
- Check.
Flaps 14.
Speed break.
Speed break armed.
Green light.
One thing at a time.
Landing gear.
Gear down.
Three green.
Flaps? Flaps 40, green light.
- 500.
- All lights on.
Please warn the cabin crew.
Cabin crew, take your seats.
Then real trouble - a stall warning.
- Speed, sir! - I have control.
100 knots of speed.
Arisan fought to save his plane but just 400 feet above the ground at only 1.
5km from the runway, the Boeing 737 suddenly fell straight down.
It only took a few seconds for it to hit the ground.
AIR TRAFFIC CONTROL: Turkish 1951 The recording sheds light on the final minutes of the flight.
The crew was configuring their plane for landing well after it should have been done.
- Flaps? - Flaps, 40.
Green light.
Most airlines have regulations that call for a flight to be stabilised, to have all checklists completed by the time the plane hits 1,000 feet.
In the instrument conditions, you're required at 1,000 feet to have basically everything done.
The airplane is configured, you have slowed, you have run your before-landing check and you have received your landing clearance and from 1,000 feet on in, you just monitor the instruments and we're looking for the runway.
Please warn the cabin crew.
Cabin crew In fact, this crew was still running their checklist up to the moment the crisis hit, 460 feet above the ground.
This approach was not stabilised.
Because the aircraft was unstable, the flight crew was in a very high workload environment in the last 1,000 feet of flight.
The radio altimeter was malfunctioning, the aircraft was giving off warnings, the crew was assigned a challenging approach and they were executing a checklist late.
But none of this explains why Flight 1951 crashed.
In these type of accidents, you can never get inside the head of the pilot and that's a very frustrating type of accident.
But the flight data recorder does provide another intriguing clue.
Moments before Flight 1951 hit the ground, the plane's engines were at idle, hardly providing any power.
Perhaps this accident is a repeat of the Heathrow incident.
The engines, it was interesting to note, were at idle approximately the last two minutes of the flight until the very end when the thrust was increased again.
That was a big red flag right there.
The question is why was that the case? But then they spot something that's very different from the accident at Heathrow.
Retard flare mode.
For some reason, while still more than 1,000 feet above the ground, the plane's computer began preparing to touch down.
In retard flare mode, engine power is reduced to idle by the flight computer and the plane's nose automatically pitches up to the flare position.
Planes should only be in this configuration just before they touch the ground.
The autopilot raises the nose to break the descent.
The autothrottle brings the power back to flight idle and you touch down with the power either all the way in idle or just about to be in idle.
But Flight 1951 went into a slow nose-up position well before touchdown, causing the plane to fly slower and slower throughout its descent.
So why was Flight 1951 in landing mode? And why hadn't any of the three crew members noticed how slowly they were flying? So what else was going on when the engines went to idle? The trouble seems to start with the malfunctioning altimeter.
SEDOR: We had to look at the system as a whole and see how that -8 affected the other systems on the aircraft.
And that was a very big portion of this investigation.
We had to say - how did the autopilot use that data? More importantly, how did the autothrottle use that data? The computer that flies the plane consists of two main systems - the autopilot and the autothrottle.
The autothrottle determines how much power to ask the engines for, while the autopilot controls the plane's altitude and direction.
The two systems work independently of each other and only one of the radio altimeters provides information to the autothrottle.
In this case, I had to learn everything there was about radio altimeters and autothrottle systems, which I didn't know before.
The pieces of the puzzle begin coming together when they find the connection between the faulty radio altimeter and engine power.
The radio altimeter provides information to the autothrottle from the captain's side.
The only altimeter feeding information to the autothrottle was the captain's and it was wrong.
It showed -8 feet throughout most of Flight 1951's approach.
It's beginning to look like the faulty radio altimeter triggered the events that led to the crash.
Investigators need to know what went wrong with it.
On a 737, the transmitting and receiving antennas for both radio altimeters are lined up underneath the cockpit.
Three of the antennas were all but destroyed in the crash.
They can't be tested.
But one antenna from the captain's side is undamaged.
Investigators consider two possibilities - a failure of one of the components or some sort of interference that caused the faulty reading.
The only component that survived the crash checks out.
The computers that control the system also work.
But investigators do make a curious discovery about them.
They aren't the same ones that were installed on the plane when it was delivered to Turkish Airlines seven years ago.
This find changes the focus of the investigation.
The maintenance aspect of this accident aircraft was one that we looked at as deeply as we could.
When the plane's maintenance log is studied, investigators find that the radio altimeter on this plane had a problematic history.
We got additional data from Turkish Airlines and that data showed that on this one aircraft, of the past, I believe, over 1,000 flights, there was about 150 flights that had faulty radio altimeter systems.
The documents show that a little more than a year before the crash, both computers were replaced because of complaints they were causing faulty readings.
One of the incidents involved a radio altimeter reading of -8 feet.
So that was telling us that there was an issue that had been there - the issue did not just occur on this flight.
The faulty readings persisted.
Mechanics repeatedly swapped the computers and replaced the antennas to try to solve the problem.
It's determined that Turkish Airlines tried several ways to fix the altimeter but they couldn't find a repair that worked.
At the time of the accident, Turkish Airlines had a fleet of 52 Boeing 737-800 series airplanes.
It's on page 93.
When we reviewed the maintenance data, we found that radio altimeter problems had been written up several times on both the accident airplane and the fleet.
Investigators discover that in the year before the crash, Turkish Airlines dealt with 235 system faults with the radio altimeters on their 737s.
Fixes ranged from replacing and exchanging antennas, cleaning of the systems, exchanging and replacing the computers and installing gaskets to shield the system from possible water damage.
It's not like they weren't doing anything about it.
The Turkish Airlines maintenance personnel knew that the radio altimeter problem was one of their highest issues with regard to maintenance.
16 of those altimeter repairs were made to the plane that crashed in February 2009.
If the problem was so widespread, investigators wonder why it hadn't caused serious problems before this accident.
They don't have to dig too far back to find out that, in fact, it had, on this very same plane.
On two recent flights, they had the exact same problem.
Twice in the 48 hours leading up to the accident, the radio altimeter showed a negative reading, putting the plane into retard flare mode.
Both times, the crew noticed the problem, disengaged the autothrottles and brought the plane in for a safe landing.
You just disconnect it and fly the airplane.
In the months after the crash, other operators come forward with similar stories.
In Australia, in the Netherlands, in Canada, in Austria, pilots report their 737s going into retard flare mode when the left radio altimeter showed a faulty reading.
Each of those crews reacted the same way.
They disengaged the autothrottle and pushed the power back up manually.
They all landed safely.
HUFF: Things are gonna break on an airplane and usually you're able to identify that and take that out of, uh .
.
make it so that it's not a threat for the landing.
In 2008, Boeing received a whopping 2,569 reports of faulty radio altimeters on their latest 737s.
But very few of those cases involved the plane going into retard flare mode.
Hardly any reports at all.
Boeing also tried but couldn't find the cause of the failures.
They concluded that the radio altimeter problem was not a threat to safety because the 737 gives off enough warnings so that crews can intervene and land safely.
In fact, in every instance where the radio altimeter failed, crews were able to recover.
Turkish Airlines Flight 1951 seems to be the one exception.
Investigators still don't know why.
Finally, as investigators again revisit the last minutes of Flight 1951, the circumstances of the tragedy become clear.
They see a remarkable sequence of events that transpire to bring down this plane.
So, what was happening when the plane went into retard flare mode? They discover that the plane went into landing mode and pulled back power at the worst possible moment - exactly as the crew was descending to meet the glide slope.
It masked what was actually happening.
As the crew configured their plane to drop down to meet the glide slope, they expected the plane to slow down as part of that manoeuvre but the plane was actually slowing down because the computer was in landing mode.
That's why none of the three pilots said anything about the throttles moving to idle.
HUFF: It was insidious.
Where it first captured in the retard mode, it didn't hurt them at all because they were actually high and they were a little bit fast and the pilots actually wanted to power back anyway.
In fact, the throttles may have already been in the flight idle mode as they were trying to get down and slow down.
But the trouble starts here, at 8,300ft.
13 miles out from the airport, minutes before the crash.
Amsterdam.
Turkish 1951 descending 7,000.
Speed - 250.
MAN: (OVER RADIO) Turkish 1951, descend to 4,000.
Speed OK for ILS.
18 right.
Radio altimeter.
(MACHINE BEEPS) But would the crew have known that because of that radio altimeter they would've gone into a retard flare mode in the throttles? No.
It's a common problem at the airline but the crew couldn't see the risk it posed this flight.
We have an airplane that was malfunctioning in a very minor way but in a way that if not caught could and did metastasise into something much more virulent.
Turkish 1951, descend to 2,000.
2,000.
1951.
Turkish 1951, left heading 210.
Clear to approach.
18 right.
Left 201, cleared ILS.
Turkish 1951.
Left at 210 degrees maintaining 2,000ft brings the flight in right here.
5.
5 miles out.
They now have to intercept the glide slope from above.
At 2,000ft, with the glide slope below them, the pilots have to reduce their speed while descending steeply.
Speed - 140.
They believe the throttles are moving back for the descent to the glide slope.
In fact, the autothrottle is slowing the plane down because it's gone into landing mode.
It will continue to slow the plane until it stalls.
SEDOR: What we found is that when the flight crew was doing their before-landing checklist, each one of them was doing something while they should have been monitoring their air speed.
For the next 100 seconds, no-one notices what's happening.
Until it's too late.
Establish altitude set.
- 1,000.
- Check.
Flaps 40.
Speed set.
The experienced pilot recognised that the first officer was probably a little bit behind on the approach so he calls for flap 40 and moves the lever, informing the first officer that he has done so.
He's trying to help the first officer catch up to the position of the aircraft.
- Speed brake.
- Speed brake.
Green light.
One thing at a time.
Landing gear.
The plane is now 700ft from the ground.
- Gear down.
Three green.
- Flaps.
Flaps 40.
Green light.
In their haste to complete their checklist, none of the three crew members noticed the warnings that their speed is dropping dangerously.
First, a red bar appeared on their flight display.
Then, when the air speed continued to drop, a flashing box appeared around their actual air speed to draw the pilots' attention to it.
At this point, no-one sees either.
Cabin, report confirm.
The aircraft is now 600ft from landing.
When things start changing colours, it's a warning to you.
It's a caution to you that you're approaching the limits of the aircraft.
Missed approach.
Altitude set.
SEDOR: So, all these indications the crew has in front of them, saying that the aircraft is slowing down, during that time, they were still completing their checklists.
Of course, the aircraft is getting closer to the ground.
In fact, it's less than 500ft from touchdown.
And right before the stick shaker started, the captain told the safety pilot Please warn the cabin crew.
Cabin crew, take your seats.
- Speed, sir! - I have control! By the time they advanced the throttles to full power, it was unrecoverable.
They were too low for the engines to catch up.
And that's it.
It's now too late to save this plane.
They all knew about the altimeter problem but knowing didn't help.
Boeing also didn't foresee the potentially deadly effect of a faulty altimeter.
But on February 25, 2009, it triggered an unusual sequence of events that brought down a jetliner and killed nine people.
The official report into the accident blames it on a convergence of circumstances.
NANCE: There is never, ever just one cause to an airline accident.
It simply doesn't exist.
Maybe someday God will swat one out of the sky but until then, there's never one cause.
The Dutch report also points out that Boeing could have realised the problem with the radio altimeter system could have had an impact on safety.
Given that no-one foresaw how that failure might cause a crash, the Turkish Airlines accident raises a big question - are airplanes becoming too complex? Investigators have determined that Turkish Airlines Flight 1951 crashed mainly because the pilots didn't recognise the consequences of the warnings they were getting.
MAN: (OVER RADIO) Turkish 1951, descend to 4,000.
This is not the first plane to crash because the crew didn't understand what their plane was telling them.
Turkish Our problem is not the automation.
Our problem is the depth of the training and the ability of the human beings to recover from mistakes made in interfacing with the automation.
Mica Endsley studies the relationship between pilots and technology.
We haven't really designed the automation to take best advantage of what people do well and take away the parts that people don't do well.
In 1996, the crew of an Aeroperu 757 crashed when the pilots couldn't decipher contradictory warnings about their altitude and air speed.
The plane flew into the Pacific Ocean.
61 passengers and 9 crew members were killed.
In 1995, the flight management system on a Boeing 757 could and should have steered the plane to a safe landing in Cali, Colombia.
But a last-minute change to the flight plan meant the crew had to reprogram their computer.
They mistakenly entered a course that took them headlong into a 9,000ft mountain.
159 people died in the crash.
The cautionary tale here is that we can get this equipment, we can get these silicone-based units, if you will, so sophisticated that we can't talk to them effectively.
And when they go berserk or they have a problem or we mis-program them, we end up putting ourselves and our passengers in danger while we're trying to figure it out.
NASA is working on something called the Integrated Intelligent Flight Deck.
It's a project aimed at helping humans better use the technology that surrounds them in the cockpit.
They're looking at automated flight manual systems, voice recognition systems, they have developed new kinds of microphones that are bone-conducting chips.
They're developing new displays for understanding the ground environment for being able to detect where other aircraft are in the environment.
What some researchers are finding is that the best technology shouldn't replace pilots but work with them.
Really, integrating people with technology more effectively has to do with designing the displays so that you can really understand what it's doing and you can make it simpler to understand how to make it do what it is you want it to do.
You shouldn't have to push 16 buttons through 8 levels of menus to figure out what's going on with the system.
It should be integrated and presented effectively.
It should be as easy to communicate with as the person next to you.
Boeing and Airbus, the two largest manufacturers of passenger planes, take very different approaches to the relationship between humans and technology.
Airbus gives the flight computer much of the decision-making power in the cockpit.
In their view, this is the way to prevent a lot of human errors by making sure the airplane doesn't do something that's gonna cause a crash even if the humans want it to.
But Boeing has a different view.
Its philosophy is to provide information to pilots and have them make decisions.
Having more information is better for the pilot, having the pilot in the loop, in the equation, so to speak.
I kind of like that.
Airbus will argue vociferously and in continuous fashion that that view is archaic and incorrect.
I think the jury is still out.
The final report into the Turkish Airlines tragedy blames the crash of Flight 1951 partly on a failure of technology.
The erroneous radio altimeter data caused the autothrottle to go to an improper mode that is, of course, not a good situation.
The cause of the radio altimeter failure was never uncovered.
The Dutch investigation asks Boeing to improve the reliability of the system.
We learnt a lot about the radio altimeter system and how it affects the autothrottle system.
Boeing is currently working on system improvements to prevent this type of autothrottle event from occurring again.
But the report also faults the crew for not noticing their air speed was dropping dangerously low in spite of the fact that there were three pilots on board.
BETHUNE: Forget that you've got an autothrottle.
YOU look at the air speed and YOU look at the altitude.
You look out damn window if you want to but air speed is a crucial component of staying alive in an airplane and you always need to know what your air speed is.
But to Mica Endsley, the crew's failure to monitor instruments is entirely understandable and may be more the fault of the instruments than the crew.
It's actually very difficult for people to be monitors of automation.
One of the things that people don't do a good job of, actually, is monitoring.
We're very good on-the-spot decision makers, we're very good at coming up with creative solutions to problems but repetitive monitoring is the kind of thing that people aren't very good at, at all.
So, what we have to do is design automation to work with people in a way that keeps them more actively, cognitively involved and in the loop and not just monitoring a piece of automation to say, "Is it doing what it's supposed to be doing?" NANCE: Who's the ultimate computer? The pilot.
The individual who should be able to say, "I don't know what this thing is doing to me "but I'm punching it off and flying the airplane - fly the jet.
" What probably is the smartest thing we ever learn to say in our training in the airlines, "Fly the jet.
" Do that first or nothing else counts.
That's what they forgot to do.
Supertext Captions by Red Bee Media Australia
You'll have to turn that off for now, sir.
Thank you.
NARRATOR: Turkish Airlines Flight 1951 is preparing to land in Amsterdam.
Amsterdam, Turkish 1951.
Descending to 7,000.
Speed - 250.
(BEEPING) The crew is flying a state-of-the-art Boeing 737.
Flaps - 15.
Localiser alive.
Localiser capture.
In the final moments of the flight, the landing turns into a catastrophe.
The plane falls like a rock.
The crash of Turkish Airlines Flight 1951 involves the most popular plane on earth.
With nearly 1.
5 million passengers boarding 737s every day, investigators need to figure out if the problem was with the plane or with the pilots flying it.
Mayday, mayday.
On the morning on February 25, 2009, Turkish Airlines Flight 1951 becomes the first plane to crash at Amsterdam's Schiphol Airport in more than 10 years.
(SPEAKS FOREIGN LANGUAGE) TRANSLATION: It smashed into the ground really hard.
It made a tremendous noise.
The plane hits the ground in a muddy field, just north of runway 18 right.
Since the crash was so close to the airport, rescue workers arrive quickly.
Most of the passengers have survived but many are badly injured.
Survivors are taken straight to local hospitals.
Images of the Amsterdam accident quickly spread around the world.
This is the third crash of a passenger jet in the past six weeks.
The sudden nature of this accident adds to the mystery.
It doesn't take long for the Dutch Safety Board to arrive at the scene.
They will be investigating this accident but they won't be alone.
The crash involved an American-made plane, so the US National Transportation Safety Board sends Joe Sedor, one of its most experienced investigators, to Amsterdam.
When it's a non-US registered aircraft that crashes overseas, such as this Turkish Airlines, we are the state of manufacture and design of the airframe, and also, in this case, the engines.
Fuselage in three large pieces.
Engines forward of the main wreck site.
Flight 1951 was one of the most advanced aircraft in the skies, the Boeing 737-800 series.
It's designed to travel longer routes at higher altitudes.
The new generation 737 is still the best airplane Boeing ever built.
We developed an airplane that had an improved wing, improved avionics, simpler systems that required less maintenance.
Investigators know this isn't just any plane.
The 737 is the world's best-selling commercial jet.
Finding out why this one crashed is imperative.
There are more than 5,000 of them in the skies.
They carry about 1.
5 million passengers a day.
Investigators must quickly determine if there's a flaw with the plane that could cause another accident.
What they know so far is that Flight 1951 had been travelling from Istanbul, Turkey, to Amsterdam.
There were 128 passengers onboard the early morning flight, including four engineers from Boeing.
WOMAN: (OVER PA) Ladies and gentlemen, we're beginning our decent towards Amsterdam's Schiphol Airport.
Please raise your seat backs to the upright position, and stow away your tray tables.
There was no mention onboard of any kind of trouble in the cockpit.
The crash has killed nine people, including three members of the Boeing team and the pilots in the cockpit.
See if you can get me some aerials of the crash site.
There are eerie similarities to another recent accident involving a Boeing aircraft - British Airways Flight 38.
SEDOR: Approximately a year before this, there had been a 777 short landing at Heathrow, which had a duel engine flame-out.
In that accident, a Boeing 777 fell to the ground short of the runway.
The British Airways pilots reported that both their engines stopped delivering power just before landing.
At the time of the Turkish Airways crash, the cause of that accident hasn't been found.
Like in that case, investigators have plenty of clues to work with.
The plane and its engines are largely intact.
The flight data recorder and cockpit voice recorder are found in good condition.
There are also plenty of survivors to describe what happened.
According to the passengers, the landing had been routine but then suddenly, the plane simply dropped out of the sky and hit the ground.
But perhaps the biggest clue comes from the crash site itself.
The wreckage is not spread out.
It tells investigators that the plane could not have been travelling forward at high speed when it hit the ground.
MAN: The way the aircraft had crashed, it did appear to be some sort of a landing accident in which there was moderate control of some sort.
The pattern of debris and the passenger reports point investigators to an immediate suspect - the engines.
The engines issue was a very big issue in my thought process at the time, initially.
There's no evidence of fire on the fuselage.
In many crashes, fuel in the plane's tanks ignites on impact.
The lack of fire raises an obvious question - did the engines stop running because Flight 1951 had simply run out of gas? NANCE: That was one of the first thoughts that I had was, "Did this airplane have fuel aboard?" Because otherwise how does a 737 literally fall out of the sky on approach to an airport? But the location and condition of the plane's engines suggests that perhaps they didn't quit in flight.
Sure looks like it was running.
SEDOR: When we first looked at where the engines ended up, the initial impression was that they probably were producing thrust at impact, given that they were so far forward of the main wreckage.
That was just a very general conclusion.
Only the flight data recorder can tell investigators how much power the engines were generating in the seconds before the crash.
Lots of fuel.
Rules that out.
It doesn't take long to discover that there was plenty of fuel in the plane's tanks.
Flight 1951 definitely did not run out of gas.
Passengers report that in the final seconds before the crash, the plane hit what felt like turbulence.
It points investigators to a well-known culprit.
A microburst.
A microburst is a powerful column of air that shoots out of storm clouds.
It can literally slam a low-flying plane to the ground.
NANCE: If an airplane flies into that at approach speeds, you're not going to be flying anymore.
You're going to come out of the sky.
Certainly it was one of the things that all of us took a look at at the beginning - was there a microburst? In 1985, a Delta Airlines flight was caught in a microburst while landing at Dallas/Fort Worth Airport.
The plane hit the ground short of the runway.
137 people were killed.
Investigators learn that there were heavy clouds above the airport at the time of the accident.
A powerful gust of wind may well have accompanied them.
The flight data recorder will have recorded wind speeds outside the plane.
Investigators will need to analyse the flight data to prove the microburst theory.
In the meantime, the rescue has led to an unusual discovery.
SEDOR: There was three pilots in the cockpit, which is unusual - this is a two-crew cockpit.
So why was that third pilot there? Since none of the three pilots survived the accident, it's all the more urgent for investigators to retrieve the data from the cockpit voice recorder.
It records conversations in the cockpit.
They're in luck.
Because of the way that the aircraft crashed, access to the recorders and the condition of the recorders was excellent.
The reason for the third pilot is soon uncovered.
Flying standard arrival route.
Visibility, 3,500m, expected to decrease to 2,500m.
For First Officer Murat Sezer, this had been a training flight of sorts.
He was new to the airline and was being shown the intricacies of landing at Amsterdam's airport.
MAN: In the Jeppesen charts, which is what all pilots use to navigate to and from Schiphol, there's 102 pages of information on Schiphol alone, so there's dozens of approaches.
Runway 18 right has 3 high-speed exits to the left.
The captain, Hasan Arisan, was doing double duty.
He was training his first officer Make small corrections as we get close to the runway.
.
.
and was in command of Flight 1951.
It's because Captain Arisan was teaching that there was the third pilot in the cockpit.
Olgay Ozgur was a safety pilot.
He was there to keep an eye on the flight's progress during this training mission.
SEDOR: The purpose of that second set of eyes is to make sure that the captain and the first officer - if they're in a situation where it's a little bit of a training portion of the flight - that they don't miss something.
We've got a clean recording.
The voice recorder reveals that the three crew members began preparations for landing when still above 8,500ft.
Amsterdam, Turkish 1951, descending to 7,000.
Speed - 250.
But the voice recorder has picked up an unusual sound - a warning that makes no sense at this stage of the flight.
One of the investigators from Boeing was an engineering pilot that came and helped with the CVR and he's listening to see when Or are there any unusual sounds that can be heard that would not be normal? MAN: (ON RECORDING) Feed OK for ILS 18 right.
Descent to 4,000.
ILS 18 right.
(BEEPING) The alarm keeps sounding.
It's the landing gear configuration warning horn.
Captain Arisan continually dismisses it.
Turkish 1951 descend 4,000.
ILS 18 right.
Landing gear.
Is that the landing gear warning? They're 8,300ft here.
The warning is a solid clue.
But investigators can't yet see how it could possibly have caused a crash.
The crew of Turkish Airlines Flight 1951 got warnings to extend their landing gear while still thousands of feet in the air.
On the initial listen, we heard a gear warning horn occur as the aircraft was approaching, when it was still up at .
.
and it was coming in at about 10,000ft and below.
Investigators now turn to the flight data recorder to help solve some of the mysteries surrounding this flight.
The analysis of wind speeds outside the aircraft is completed.
It's clear - none are drastic enough to have brought down the plane.
No evidence of a microburst.
But the flight data recorder does provide some valuable insight into the cause of the landing gear warning.
One of the instruments that measures altitude had the plane already on the ground.
When we looked and saw the radio altimeter data on the recorder, it said about 8,000ft and then immediately it went down to about -8ft.
-8ft is an indication that the aircraft is on the ground but of course it's still at 2,000ft.
The Boeing 737 is equipped with two separate altimeters.
One measures air pressure to determine the plane's height above sea level.
That reading is displayed prominently in the cockpit on both pilots' flight display.
MAN: Falling to zero, Sezer.
The plane is also equipped with a radio altimeter.
It's made up of four antennas.
Two transmit signals to the ground and two others read the signal that bounces back to determine the plane's height.
It's precise.
It's very, very precise.
Pressure altimeters can sometimes be not as accurate and radio altimeters are 100% accurate, if they're working properly.
One antenna feeds the reading to the first officer's display.
The other feeds the captain's instruments.
In the case of Flight 1951, the captain's side was wrong most of the flight.
Investigators go back over the CVR and make a puzzling discovery.
MAN: (ON RECORDING) Amsterdam, Turkish 1951, descending to 7,000.
Speed - 250.
Radio altimeter.
Captain Arisan seems to have known that the landing gear warning was being caused by a faulty radio altimeter.
The airplane thought it was low to the ground and the gear was not down and the captain recognised that the problem was really in the radio altimeter, showing him that they should be on the ground, and he goes, "It's just the radio altimeter.
" Throughout much of the approach, the captain's radio altimeter had been displaying a reading of -8ft, triggering the warning to lower the gear.
They treated it like it was a nuisance.
Turkish 1951, descend to 2,000.
2,000.
1951.
Investigators dig for any other abnormalities.
They learn that with flight 1951 still about 17km from the airport, controllers directed the pilots to begin their final turn to line up with the runway.
Turkish 1951, left heading 210.
Clear to approach.
18 right.
Left 210, clear ILS.
Turkish 1951.
210 set, sir.
This turn puts flight 1951 in line with runaway 18 right.
It's equipped with an instrument landing system which sends out a signal outlining the ideal descent path to the foot of the runway.
The autopilot follows that glide path until the plane is a few hundred feet from the ground.
Then the pilot takes over.
It makes landing almost effortless.
The ILS is pretty easy to follow.
It's a video game.
My daughter has flown in a simulator and can land a 737 using the ILS.
The crew begins configuring their plane for landing, unfazed by the warning horn that's repeatedly triggered by the malfunctioning radio altimeter.
Flaps 15.
10km out, Flight 1951 picks up the ILS signal that will guide the plane to the runway.
Localiser alive.
Localiser capture.
The safety pilot, Olgay Ozgur, now reminds Captain Arisan about the failed altimeter.
We have radio altimeter failure, sir.
OK.
Turkish 1951, runway 18 right, clear to land.
Clear to land.
Thank you.
Investigators are stumped.
The crew knew about the malfunction and continued their approach.
How had it then caused them to crash? Clearly, there was more to this accident than a faulty altimeter.
NANCE: The whole premise of airline safety, the way we build the airplanes, the way we fly them, is based on the idea that we can have any number of failures and we should still be able to arrive safely.
The radio altimeter is just one instrument.
There's no way in the world that that one instrument, if it fails, should be a major cause of worry that we're going to have a crash.
Investigators wonder if the crew had been given proper guidance for their approach.
They turn to exchanges between the pilots and the controller who guided them in.
They carefully review every instruction.
TWO-WAY RECORDING: Turkish 1951, descend to 4,000.
AIR TRAFFIC CONTROL: Speed OK for ILS 18 right.
Turkish 1951, descend to 2,000.
Turkish 1951, left heading 210, clear to approach 18 right.
By following the controller's instructions, the crew made their final turn much too close to the runway.
So to intercept properly, they should be here.
International guidelines call for approaching planes to intercept the signal that guides them to the runway from below.
It's so pilots don't have to make any drastic last-minute course corrections to get to it.
To intercept here, they had to descend.
But Flight 1951 was given instructions that brought it to the threshold of the glide slope while still way above it.
It's a common practice at Schiphol because it gets planes to the runway faster.
SEDOR: Because they were so close, they had to capture the glide slope from above.
Although it is an unusual situation, it is one that can be handled by a flight crew if it is managed properly.
Approaching a glide slope from above is more difficult, mostly because the crew has to suddenly slow the plane and descend rapidly to intercept the signal.
HUFF: We've also called this a slam-dunk approach and some pilots like it, some pilots don't.
It's a little bit harder and things happen quicker when you're above the glide path trying to intercept from above.
And it's just a challenge for a lot of pilots.
The approach from above increased the crew's workload but it's standard practice at Schiphol Airport.
I've flown into Schiphol dozens of times and I expect it.
If the controller's instructions had somehow overtaxed this crew, their conversations would indicate it.
They're just 5km from the runway.
- 1,000.
- Check.
Flaps 14.
Speed break.
Speed break armed.
Green light.
One thing at a time.
Landing gear.
Gear down.
Three green.
Flaps? Flaps 40, green light.
- 500.
- All lights on.
Please warn the cabin crew.
Cabin crew, take your seats.
Then real trouble - a stall warning.
- Speed, sir! - I have control.
100 knots of speed.
Arisan fought to save his plane but just 400 feet above the ground at only 1.
5km from the runway, the Boeing 737 suddenly fell straight down.
It only took a few seconds for it to hit the ground.
AIR TRAFFIC CONTROL: Turkish 1951 The recording sheds light on the final minutes of the flight.
The crew was configuring their plane for landing well after it should have been done.
- Flaps? - Flaps, 40.
Green light.
Most airlines have regulations that call for a flight to be stabilised, to have all checklists completed by the time the plane hits 1,000 feet.
In the instrument conditions, you're required at 1,000 feet to have basically everything done.
The airplane is configured, you have slowed, you have run your before-landing check and you have received your landing clearance and from 1,000 feet on in, you just monitor the instruments and we're looking for the runway.
Please warn the cabin crew.
Cabin crew In fact, this crew was still running their checklist up to the moment the crisis hit, 460 feet above the ground.
This approach was not stabilised.
Because the aircraft was unstable, the flight crew was in a very high workload environment in the last 1,000 feet of flight.
The radio altimeter was malfunctioning, the aircraft was giving off warnings, the crew was assigned a challenging approach and they were executing a checklist late.
But none of this explains why Flight 1951 crashed.
In these type of accidents, you can never get inside the head of the pilot and that's a very frustrating type of accident.
But the flight data recorder does provide another intriguing clue.
Moments before Flight 1951 hit the ground, the plane's engines were at idle, hardly providing any power.
Perhaps this accident is a repeat of the Heathrow incident.
The engines, it was interesting to note, were at idle approximately the last two minutes of the flight until the very end when the thrust was increased again.
That was a big red flag right there.
The question is why was that the case? But then they spot something that's very different from the accident at Heathrow.
Retard flare mode.
For some reason, while still more than 1,000 feet above the ground, the plane's computer began preparing to touch down.
In retard flare mode, engine power is reduced to idle by the flight computer and the plane's nose automatically pitches up to the flare position.
Planes should only be in this configuration just before they touch the ground.
The autopilot raises the nose to break the descent.
The autothrottle brings the power back to flight idle and you touch down with the power either all the way in idle or just about to be in idle.
But Flight 1951 went into a slow nose-up position well before touchdown, causing the plane to fly slower and slower throughout its descent.
So why was Flight 1951 in landing mode? And why hadn't any of the three crew members noticed how slowly they were flying? So what else was going on when the engines went to idle? The trouble seems to start with the malfunctioning altimeter.
SEDOR: We had to look at the system as a whole and see how that -8 affected the other systems on the aircraft.
And that was a very big portion of this investigation.
We had to say - how did the autopilot use that data? More importantly, how did the autothrottle use that data? The computer that flies the plane consists of two main systems - the autopilot and the autothrottle.
The autothrottle determines how much power to ask the engines for, while the autopilot controls the plane's altitude and direction.
The two systems work independently of each other and only one of the radio altimeters provides information to the autothrottle.
In this case, I had to learn everything there was about radio altimeters and autothrottle systems, which I didn't know before.
The pieces of the puzzle begin coming together when they find the connection between the faulty radio altimeter and engine power.
The radio altimeter provides information to the autothrottle from the captain's side.
The only altimeter feeding information to the autothrottle was the captain's and it was wrong.
It showed -8 feet throughout most of Flight 1951's approach.
It's beginning to look like the faulty radio altimeter triggered the events that led to the crash.
Investigators need to know what went wrong with it.
On a 737, the transmitting and receiving antennas for both radio altimeters are lined up underneath the cockpit.
Three of the antennas were all but destroyed in the crash.
They can't be tested.
But one antenna from the captain's side is undamaged.
Investigators consider two possibilities - a failure of one of the components or some sort of interference that caused the faulty reading.
The only component that survived the crash checks out.
The computers that control the system also work.
But investigators do make a curious discovery about them.
They aren't the same ones that were installed on the plane when it was delivered to Turkish Airlines seven years ago.
This find changes the focus of the investigation.
The maintenance aspect of this accident aircraft was one that we looked at as deeply as we could.
When the plane's maintenance log is studied, investigators find that the radio altimeter on this plane had a problematic history.
We got additional data from Turkish Airlines and that data showed that on this one aircraft, of the past, I believe, over 1,000 flights, there was about 150 flights that had faulty radio altimeter systems.
The documents show that a little more than a year before the crash, both computers were replaced because of complaints they were causing faulty readings.
One of the incidents involved a radio altimeter reading of -8 feet.
So that was telling us that there was an issue that had been there - the issue did not just occur on this flight.
The faulty readings persisted.
Mechanics repeatedly swapped the computers and replaced the antennas to try to solve the problem.
It's determined that Turkish Airlines tried several ways to fix the altimeter but they couldn't find a repair that worked.
At the time of the accident, Turkish Airlines had a fleet of 52 Boeing 737-800 series airplanes.
It's on page 93.
When we reviewed the maintenance data, we found that radio altimeter problems had been written up several times on both the accident airplane and the fleet.
Investigators discover that in the year before the crash, Turkish Airlines dealt with 235 system faults with the radio altimeters on their 737s.
Fixes ranged from replacing and exchanging antennas, cleaning of the systems, exchanging and replacing the computers and installing gaskets to shield the system from possible water damage.
It's not like they weren't doing anything about it.
The Turkish Airlines maintenance personnel knew that the radio altimeter problem was one of their highest issues with regard to maintenance.
16 of those altimeter repairs were made to the plane that crashed in February 2009.
If the problem was so widespread, investigators wonder why it hadn't caused serious problems before this accident.
They don't have to dig too far back to find out that, in fact, it had, on this very same plane.
On two recent flights, they had the exact same problem.
Twice in the 48 hours leading up to the accident, the radio altimeter showed a negative reading, putting the plane into retard flare mode.
Both times, the crew noticed the problem, disengaged the autothrottles and brought the plane in for a safe landing.
You just disconnect it and fly the airplane.
In the months after the crash, other operators come forward with similar stories.
In Australia, in the Netherlands, in Canada, in Austria, pilots report their 737s going into retard flare mode when the left radio altimeter showed a faulty reading.
Each of those crews reacted the same way.
They disengaged the autothrottle and pushed the power back up manually.
They all landed safely.
HUFF: Things are gonna break on an airplane and usually you're able to identify that and take that out of, uh .
.
make it so that it's not a threat for the landing.
In 2008, Boeing received a whopping 2,569 reports of faulty radio altimeters on their latest 737s.
But very few of those cases involved the plane going into retard flare mode.
Hardly any reports at all.
Boeing also tried but couldn't find the cause of the failures.
They concluded that the radio altimeter problem was not a threat to safety because the 737 gives off enough warnings so that crews can intervene and land safely.
In fact, in every instance where the radio altimeter failed, crews were able to recover.
Turkish Airlines Flight 1951 seems to be the one exception.
Investigators still don't know why.
Finally, as investigators again revisit the last minutes of Flight 1951, the circumstances of the tragedy become clear.
They see a remarkable sequence of events that transpire to bring down this plane.
So, what was happening when the plane went into retard flare mode? They discover that the plane went into landing mode and pulled back power at the worst possible moment - exactly as the crew was descending to meet the glide slope.
It masked what was actually happening.
As the crew configured their plane to drop down to meet the glide slope, they expected the plane to slow down as part of that manoeuvre but the plane was actually slowing down because the computer was in landing mode.
That's why none of the three pilots said anything about the throttles moving to idle.
HUFF: It was insidious.
Where it first captured in the retard mode, it didn't hurt them at all because they were actually high and they were a little bit fast and the pilots actually wanted to power back anyway.
In fact, the throttles may have already been in the flight idle mode as they were trying to get down and slow down.
But the trouble starts here, at 8,300ft.
13 miles out from the airport, minutes before the crash.
Amsterdam.
Turkish 1951 descending 7,000.
Speed - 250.
MAN: (OVER RADIO) Turkish 1951, descend to 4,000.
Speed OK for ILS.
18 right.
Radio altimeter.
(MACHINE BEEPS) But would the crew have known that because of that radio altimeter they would've gone into a retard flare mode in the throttles? No.
It's a common problem at the airline but the crew couldn't see the risk it posed this flight.
We have an airplane that was malfunctioning in a very minor way but in a way that if not caught could and did metastasise into something much more virulent.
Turkish 1951, descend to 2,000.
2,000.
1951.
Turkish 1951, left heading 210.
Clear to approach.
18 right.
Left 201, cleared ILS.
Turkish 1951.
Left at 210 degrees maintaining 2,000ft brings the flight in right here.
5.
5 miles out.
They now have to intercept the glide slope from above.
At 2,000ft, with the glide slope below them, the pilots have to reduce their speed while descending steeply.
Speed - 140.
They believe the throttles are moving back for the descent to the glide slope.
In fact, the autothrottle is slowing the plane down because it's gone into landing mode.
It will continue to slow the plane until it stalls.
SEDOR: What we found is that when the flight crew was doing their before-landing checklist, each one of them was doing something while they should have been monitoring their air speed.
For the next 100 seconds, no-one notices what's happening.
Until it's too late.
Establish altitude set.
- 1,000.
- Check.
Flaps 40.
Speed set.
The experienced pilot recognised that the first officer was probably a little bit behind on the approach so he calls for flap 40 and moves the lever, informing the first officer that he has done so.
He's trying to help the first officer catch up to the position of the aircraft.
- Speed brake.
- Speed brake.
Green light.
One thing at a time.
Landing gear.
The plane is now 700ft from the ground.
- Gear down.
Three green.
- Flaps.
Flaps 40.
Green light.
In their haste to complete their checklist, none of the three crew members noticed the warnings that their speed is dropping dangerously.
First, a red bar appeared on their flight display.
Then, when the air speed continued to drop, a flashing box appeared around their actual air speed to draw the pilots' attention to it.
At this point, no-one sees either.
Cabin, report confirm.
The aircraft is now 600ft from landing.
When things start changing colours, it's a warning to you.
It's a caution to you that you're approaching the limits of the aircraft.
Missed approach.
Altitude set.
SEDOR: So, all these indications the crew has in front of them, saying that the aircraft is slowing down, during that time, they were still completing their checklists.
Of course, the aircraft is getting closer to the ground.
In fact, it's less than 500ft from touchdown.
And right before the stick shaker started, the captain told the safety pilot Please warn the cabin crew.
Cabin crew, take your seats.
- Speed, sir! - I have control! By the time they advanced the throttles to full power, it was unrecoverable.
They were too low for the engines to catch up.
And that's it.
It's now too late to save this plane.
They all knew about the altimeter problem but knowing didn't help.
Boeing also didn't foresee the potentially deadly effect of a faulty altimeter.
But on February 25, 2009, it triggered an unusual sequence of events that brought down a jetliner and killed nine people.
The official report into the accident blames it on a convergence of circumstances.
NANCE: There is never, ever just one cause to an airline accident.
It simply doesn't exist.
Maybe someday God will swat one out of the sky but until then, there's never one cause.
The Dutch report also points out that Boeing could have realised the problem with the radio altimeter system could have had an impact on safety.
Given that no-one foresaw how that failure might cause a crash, the Turkish Airlines accident raises a big question - are airplanes becoming too complex? Investigators have determined that Turkish Airlines Flight 1951 crashed mainly because the pilots didn't recognise the consequences of the warnings they were getting.
MAN: (OVER RADIO) Turkish 1951, descend to 4,000.
This is not the first plane to crash because the crew didn't understand what their plane was telling them.
Turkish Our problem is not the automation.
Our problem is the depth of the training and the ability of the human beings to recover from mistakes made in interfacing with the automation.
Mica Endsley studies the relationship between pilots and technology.
We haven't really designed the automation to take best advantage of what people do well and take away the parts that people don't do well.
In 1996, the crew of an Aeroperu 757 crashed when the pilots couldn't decipher contradictory warnings about their altitude and air speed.
The plane flew into the Pacific Ocean.
61 passengers and 9 crew members were killed.
In 1995, the flight management system on a Boeing 757 could and should have steered the plane to a safe landing in Cali, Colombia.
But a last-minute change to the flight plan meant the crew had to reprogram their computer.
They mistakenly entered a course that took them headlong into a 9,000ft mountain.
159 people died in the crash.
The cautionary tale here is that we can get this equipment, we can get these silicone-based units, if you will, so sophisticated that we can't talk to them effectively.
And when they go berserk or they have a problem or we mis-program them, we end up putting ourselves and our passengers in danger while we're trying to figure it out.
NASA is working on something called the Integrated Intelligent Flight Deck.
It's a project aimed at helping humans better use the technology that surrounds them in the cockpit.
They're looking at automated flight manual systems, voice recognition systems, they have developed new kinds of microphones that are bone-conducting chips.
They're developing new displays for understanding the ground environment for being able to detect where other aircraft are in the environment.
What some researchers are finding is that the best technology shouldn't replace pilots but work with them.
Really, integrating people with technology more effectively has to do with designing the displays so that you can really understand what it's doing and you can make it simpler to understand how to make it do what it is you want it to do.
You shouldn't have to push 16 buttons through 8 levels of menus to figure out what's going on with the system.
It should be integrated and presented effectively.
It should be as easy to communicate with as the person next to you.
Boeing and Airbus, the two largest manufacturers of passenger planes, take very different approaches to the relationship between humans and technology.
Airbus gives the flight computer much of the decision-making power in the cockpit.
In their view, this is the way to prevent a lot of human errors by making sure the airplane doesn't do something that's gonna cause a crash even if the humans want it to.
But Boeing has a different view.
Its philosophy is to provide information to pilots and have them make decisions.
Having more information is better for the pilot, having the pilot in the loop, in the equation, so to speak.
I kind of like that.
Airbus will argue vociferously and in continuous fashion that that view is archaic and incorrect.
I think the jury is still out.
The final report into the Turkish Airlines tragedy blames the crash of Flight 1951 partly on a failure of technology.
The erroneous radio altimeter data caused the autothrottle to go to an improper mode that is, of course, not a good situation.
The cause of the radio altimeter failure was never uncovered.
The Dutch investigation asks Boeing to improve the reliability of the system.
We learnt a lot about the radio altimeter system and how it affects the autothrottle system.
Boeing is currently working on system improvements to prevent this type of autothrottle event from occurring again.
But the report also faults the crew for not noticing their air speed was dropping dangerously low in spite of the fact that there were three pilots on board.
BETHUNE: Forget that you've got an autothrottle.
YOU look at the air speed and YOU look at the altitude.
You look out damn window if you want to but air speed is a crucial component of staying alive in an airplane and you always need to know what your air speed is.
But to Mica Endsley, the crew's failure to monitor instruments is entirely understandable and may be more the fault of the instruments than the crew.
It's actually very difficult for people to be monitors of automation.
One of the things that people don't do a good job of, actually, is monitoring.
We're very good on-the-spot decision makers, we're very good at coming up with creative solutions to problems but repetitive monitoring is the kind of thing that people aren't very good at, at all.
So, what we have to do is design automation to work with people in a way that keeps them more actively, cognitively involved and in the loop and not just monitoring a piece of automation to say, "Is it doing what it's supposed to be doing?" NANCE: Who's the ultimate computer? The pilot.
The individual who should be able to say, "I don't know what this thing is doing to me "but I'm punching it off and flying the airplane - fly the jet.
" What probably is the smartest thing we ever learn to say in our training in the airlines, "Fly the jet.
" Do that first or nothing else counts.
That's what they forgot to do.
Supertext Captions by Red Bee Media Australia