TK 1951: The 737 Radio Altimeter Failure That Pulled the Thrust to Idle

November 5, 2026

A single wrong number on one radio altimeter, minus 8 feet, was enough to pull the thrust levers of a Boeing 737-800 back to idle on short final into Amsterdam, and nobody on the flight deck noticed in time.

Imagine you are on the ILS at Schiphol. You were kept high and close in by the approach controller, you are catching the glideslope from above, the gear is going down and the landing checklist is starting. The autothrottle is engaged. Everything looks normal.

It is not. Somewhere behind the panel, the aircraft has just decided it is about to land, at roughly 2,000 feet.

Turkish Airlines flight 1951 (TC-JGE) is one of the most instructive accidents in modern 737 operations, because no single failure caused it. It was a faulty sensor, an automation mode the crew could not see coming, a compressed approach and a stall warning that came too late.

Short answer: why did Turkish Airlines 1951 crash? The left radio altimeter fed an erroneous height of -8 ft to the autothrottle, which switched to its landing-flare (retard) mode at about 1,950 ft and reduced thrust to idle. The approach was not stabilised, the crew did not notice the speed decay until the stick shaker activated, and the recovery came too late. The Dutch Safety Board (DSB) published its final report on 6 May 2010.

The flight

On 25 February 2009, a Turkish Airlines Boeing 737-800, registration TC-JGE, was flying from Istanbul to Amsterdam Schiphol. It was on approach to the Polderbaan runway (18R) when it crashed short of the runway. Nine people died and dozens were injured among the 135 on board. The accident was investigated by the Dutch Safety Board, assisted by the NTSB, Boeing advisers and Turkish Airlines.

The DSB found that the left radio altimeter system had begun to provide erroneous data shortly after takeoff from Istanbul. The only indication on the flight deck was a -8 ft readout on the left primary flight display. Investigators could not determine why the fault occurred, nor why the radio altimeter computer did not flag it.

As a captain, I find this accident a textbook example of how much of a modern jet's behaviour hangs on sensors we never look at.

Preparing for the B737 NG?B737 NG Bundle →

Minute by minute: from vectors to stall

The sequence below follows the DSB report as summarised by the sources listed at the end. I give rounded values where I could not check the exact figure against the original text.

Vectors. Approach control vectored the aircraft to join the localizer at roughly 6 NM from touchdown and around 2,000 ft. The glideslope was therefore intercepted from above.
Descending through about 1,950 ft. The left radio altimeter value suddenly changed to -8 ft. The autothrottle, which takes its height data from the left radio altimeter, treated this as the landing flare and moved to retard, bringing the thrust levers to idle.
Autopilot. The approach was flown with the right autopilot engaged, so the autopilot kept following the glideslope while the speed fell.
Not stabilised at 1,000 ft. The flight did not meet the operator's stabilised approach criteria at 1,000 ft above the ground. A go-around was the expected response. The crew was busy with the landing checklist at the time.
Speed decay and pitch-up. With idle thrust, the autopilot held the glideslope by increasing pitch, airspeed dropped and the pitch attitude rose. According to the DSB this went unnoticed by the crew until the stick shaker activated at about 490 ft.
Recovery attempt. Thrust was advanced, but the DSB found the approach-to-stall recovery was not executed correctly. The aircraft stalled and struck the ground short of the runway.

How the 737NG radio altimeter and autothrottle interact

On the 737NG, the radio altimeter (RA) measures the real height above the ground and feeds several systems: the flight displays, the GPWS, the autoflight system and the autothrottle. In this event the autothrottle used the left RA, while the right autopilot was engaged.

Close to the ground, the autothrottle normally retards the thrust levers for the flare. In the accident, the system received "-8 ft" and applied that landing logic at about 1,950 ft. The DSB noted that the crew had no information on the relationship between a left RA failure and the autothrottle behaviour.

Understanding which sensor feeds which system is exactly what type rating examiners probe. If you are building that picture for a 737 course, the B737 NG Systems Study Notes go through the autoflight and radio altimeter architecture system by system.

Threat and Error Management analysis

Element	In this event
Threats	Short, high vectors; glideslope capture from above; latent RA fault with only a -8 ft readout; busy flight deck during the landing checklist
Errors	Idle thrust and speed decay not detected; approach continued although not stabilised at 1,000 ft; stall recovery not executed correctly
Undesired aircraft state	Low energy, high pitch attitude on final, then stick shaker
Countermeasures that would have worked	Monitoring airspeed and thrust (the FMA and speed tape), a go-around at 1,000 ft, calling the RA discrepancy, early full-thrust stall recovery

The Swiss cheese: HFACS layers

Unsafe acts: not noticing the idle thrust and falling airspeed; continuing an unstabilised approach; recovery not as trained.
Preconditions: workload during a compressed approach; the RA fault was invisible apart from the -8 ft on one PFD; no documented knowledge of the RA-autothrottle interaction.
Unsafe supervision: the DSB discussed the operator's training of crews, including the first officer's line training with a safety pilot on board.
Organisational influences: the DSB noted that Turkish Airlines' safety assurance programme lacked a system for risk identification and management. On the manufacturer side, the autothrottle design allowed a single RA to command retard in the air.

What the investigators concluded

The DSB concluded that the accident resulted from a combination of circumstances, not a single cause: the erroneous left RA reading, the autothrottle retard mode, the compressed approach, the crew not recognising the speed decay in time and the stall recovery. The Board recommended that Boeing improve the reliability of the radio altimeter system and the autothrottle logic.

Boeing issued a bulletin to 737 operators stressing the importance of monitoring the primary flight instruments and advising against using the autopilot or autothrottle for landing when there are radio altimeter discrepancies. Please read the report itself for the full list of recommendations.

With hindsight: what has changed since

After the accident, the FAA moved to address the altimeter and autothrottle interface on 737NG aircraft through a proposed airworthiness directive, and Boeing told operators to monitor the primary flight instruments closely. I could not confirm the current status of every directive, so check the latest AD and your operator's FCOM before relying on this summary.

In 2020, press reports claimed that comments from US parties influenced how the DSB report treated technical factors. That is a reported dispute about the report, and it is a reminder to read the original findings and not only the summaries.

Lessons for today's line pilot

Know which sensor feeds which system. A bad input fails quietly in places you do not look.
Monitor the energy state, not only the path. Speed on the tape and thrust lever position are part of every scan, even with the autopilot flying.
Treat an unusual RA value as a real discrepancy. A -8 ft indication at 2,000 ft is a call-out, not noise.
Brief the high, close-in approach. If you are not stable by the gate, the plan is already a go-around.
Do not let checklists steal the monitoring. Assign who watches the airspeed while the landing checklist is read.
Practise approach-to-stall recovery. Add thrust, reduce the angle of attack, and do not wait.

If you want to drill the 737NG autoflight logic before your next simulator session, the B737 NG Type Rating Question Bank turns these interactions into questions, and 737ng.com is a useful reference for the aircraft itself.

Frequently asked questions

What caused the Turkish Airlines 1951 crash?

A faulty left radio altimeter reading of -8 ft made the autothrottle retard the thrust to idle during an unstabilised approach. The crew did not detect the speed loss until the stick shaker, and the aircraft stalled short of the runway.

Can a 737 radio altimeter failure make the autothrottle go to idle?

In this accident, yes. The autothrottle used the left radio altimeter and moved to retard at about 1,950 ft. Boeing then advised operators about monitoring and about not using the autopilot or autothrottle for landing when the radio altimeters disagree.

Was the crew to blame?

The DSB described a combination of factors. The approach was not stabilised and the speed decay was not noticed, but the crew had no information about the interaction between the radio altimeter fault and the autothrottle. A just-culture reading asks why it made sense to them at the time.

Where is the official report?

The Dutch Safety Board report is titled "Crashed during approach, Boeing 737-800, near Amsterdam Schiphol Airport, 25 February 2009" and was released on 6 May 2010.

Sources

Dutch Safety Board (Onderzoeksraad voor Veiligheid), "Crashed during approach, Boeing 737-800, near Amsterdam Schiphol Airport, 25 February 2009", final report, 6 May 2010. Copy: asn.flightsafety.org; summary: SKYbrary.
Flight Safety Foundation, Automation at Odds, AeroSafety World, June 2010.
Boeing bulletin to 737 operators, via SKYbrary.

Featured photo: Turkish Airlines B738 TC-JGE (the aircraft of Flight 1951, photographed in 2006) by Juergen Lehle (albspotter.eu), via Wikimedia Commons, licensed under CC BY-SA 3.0.

Captain Phil
Airline Captain and founder of Airline Pilot Study Notes