ATR72 Cognitive Bias Leads to Control Problems

ATR72 Cognitive Bias Leads to Control Problems

On 4 March 2016, ATR-72-212A (ATR72-500), G-COBO, operated Aurigny Air Services, the flag carrier of the Crown Dependency of Guernsey, suffered autopilot drop outs and unusually high forward pitch control forces on departure from Manchester Airport.  The aircraft diverted safely to East Midlands Airport.  The UK Air Accidents Investigation Branch (AAIB) became aware of the occurrence and recently issued an investigation report that discusses cognitive bias and the crew’s low exposure to icing conditions.

Left side of ATR72 tailplane assembly showing trim tab (circled) on left elevator (Credit: AAIB)

Left side of ATR72 tailplane assembly showing trim tab (circled) on left elevator (Credit: AAIB)

Pre-Flight Preparations

The AAIB say:

The aircraft arrived at Manchester Airport from Guernsey and remained on the ground for more than an hour, while it was snowing and the temperature was 0ºC.

While taxiing in, the crew’s perception was that light, wet snow was falling and melting on the taxiways, although some was lying on adjacent grass areas. In addition, the co-pilot could see no ice on the airframe or on the Ice Evidence Probe (IEP) when he actioned the after-landing checklist.

At 0814 hrs, the commander looked at the static air temperature gauge and observed that it was indicating between 1ºC and 2ºC. He then commented “it doesn’t appear to be sticking…so I think we can get away without de-icing”, adding that he would “have a good look”.

There was then a protracted delay before the aircraft could park, because another aircraft was being de-iced on their allocated parking stand. During this delay, the commander said he did not see any snow settling on the aircraft and suggested snow visible on other aircraft had probably accumulated overnight. The aircraft shut down at 0837 hrs, 23 min before it was scheduled to depart on a return flight to Guernsey.

Aircraft handling companies at Manchester stated that all other commercial aircraft which departed on the morning of 4 March 2016 sought de-icing/anti-icing before start-up. The de-icing providers had difficulty coping with the demand for their services and some flights were delayed while others were subsequently cancelled.

It was the commander’s belief that the air temperature was just above freezing (approximately 1ºC) and he thought the falling snow was melting on the aircraft. He could not see the top of the horizontal tailplane but, from the rear access step, he thought he saw most of the top of the wings and believed there was no frozen contamination. After completing his inspection he told the co-pilot, who had remained in the flight deck, that they could continue without de-icing. The co-pilot judged the commander’s inspection to be thorough because it took a long time and he accepted this decision.

The Incident Flight

The AAIB say:

During the takeoff, the commander exerted less aft pressure on the control column, to rotate the aircraft, than he expected and maximum nose-down pitch trim was then needed to maintain the appropriate climb attitude. The autopilot was engaged four times but on each occasion it disengaged, as designed, and the commander had to apply continuous forward pressure on the control column to retain the desired pitch attitude, as the climb proceeded.

Once at the cruising level, the commander decided he was having to exert excessive forward pressure on the control column and he elected to divert to East Midlands Airport (EMA). While descending, the aircraft flew out of icing conditions and the control difficulties dissipated.

The crew assessed that ice contamination had caused the problem and they made a normal landing.

The Initial Response

The same day an engineer inspected the aircraft and found no faults.  Subsequent investigation by AAIB did find that a microswitch that should have limited the elevator trim tab to +1.5º had failed.  This however worked so as to assist the pilot in this case with a max trim of +1.76º.

The operator downloaded flight data from the Flight Data Recorder (FDR) and from the Cockpit Voice Recorder (CVR) before the aircraft was returned to service.

Neither the commander nor the operator immediately considered that a serious incident had occurred and there was a three day delay before the AAIB was notified. The AAIB then assessed the occurrence as one ‘which could have caused difficulties controlling the aircraft’ so, in accordance with Regulation (EU) No 996/2010, it was classified as a serious incident.

Analysis of Aircraft Behaviour

The AAIB report that:

The manufacturer analysed the FDR data and concluded that the aircraft’s abnormal nose-up pitching tendency was consistent with the aerodynamic effects of upper surface icing on the horizontal tailplane.

It was concluded that the autopilot was disengaged by its internal monitoring circuitry, as a result of the load experienced by the autopilot pitch servomotor, or a manual pitch input.

Once the trim tab reached the stop, the commander had to apply an additional nose-down input on the control column to maintain the desired aircraft attitude. After levelling, he decided he would not be able to maintain the effort needed on the controls for the duration of the cruise, and made a decision to divert to EMA.

Analysis of Crew Performance & Training

The AAIB go on to say:

When flying the approach into Manchester in icing conditions, the crew saw little ice accreting on the aircraft. This may have led them to assume ice accretion was unlikely while on the ground. Also, the commander’s declaration of the temperature being a little above 0ºC may have reinforced this belief. He told the co-pilot the snow did not appear to be “sticking” and, before parking, he considered they could probably “get away without de-icing”.

The commander advised the co-pilot he would “have a good look” during his external inspection but his early pronouncement may have made both pilots susceptible to ‘confirmation bias’. From then on, they may have subconsciously tried to make the evidence available to them accord to the commander’s original assessment. This was apparent when the commander saw snow on other aircraft and declared it must have accumulated overnight.

Neither pilot seemed to consider the possibility there might be unseen ice on the upper surfaces after landing, nor that the skin temperature would probably have been colder than 0ºC. The large quantity of fuel that was added might have caused the skin temperature of the wings to warm above 0ºC but the tailplane temperature would have remained at or below 0ºC.

The AAIB comment that:

…it was apparent both ‘atmospheric icing conditions’ and ‘freezing conditions’ existed because the temperature was less than 3ºC, with visible moisture present.

In these conditions a de-ice/anti-ice and a subsequent check is expected.

The decision to depart without being certain the aircraft was free of ice, suggests the crew were affected by ‘optimism bias’ and only foresaw a positive outcome i.e. any snow ‘blowing off’.

On training the AAIB comment:

The crew was based in Guernsey, where snow seldom falls and de-icing is seldom necessary. However, the pilots had been trained for conditions like those at Manchester, so a contributory factor in deciding not to de-ice may have been that their training was less effective than it might have been.

Their conversion course syllabus specifically covered ground de-icing and the effect of improper aircraft treatment but this may not have been emphasised sufficiently during the training. Neither pilot had read the manufacturer’s ‘Cold Weather Operations’ booklet, even though this was referred to during the conversion course and circulated by the operator.

Ground de-icing procedures had been discussed with the commander as part of his recent conversion training and both pilots had acknowledged the operator’s extant FSM [Flying Staff Memo] ‘Winter Awareness 2015’. The operator’s recurrent winter training for its pilots relied on their self-study of this FSM and the reference material mentioned, in order to update their understanding of guidance in the OM and elsewhere. No recurrent classroom-based training was provided and the knowledge amassed from this self-study was not tested, other than through participation in recurrent simulator checks, which included a winter operations element.

The manufacturer’s ATO includes a scenario for poor de-icing of the tailplane in its conversion course but not all ATR simulators incorporate this profile, which only represents one potential consequence of inadequate ground de-icing.

On Crew Resource Management (CRM) issues the AAIB say:

The co-pilot appears to have been excluded from some of the commander’s decision making process. After the loadsheet had been checked for any CG issues, there was no further discussion of possible alternative reasons for the problem. No options were generated before actions were taken and there was no ongoing review of what had happened and what could be done. This was the co-pilot’s first diversion during commercial operations and his workload in support of the commander felt very high.

Safety Actions

The AAIB say:

  • The ATO responsible for the pilot’s type conversion training is adjusting its conversion course to align with the EASA Operational Suitability Data report.  This is being achieved by incorporating the manufacturer’s simulator profile for a badly de-iced tailplane.
  • The operator has enhanced its winter awareness training for pilots by purchasing a computer-based training module. All pilots will complete this before each winter season and their knowledge will be tested as part of the process.
  • The operator’s conversion courses are being extended through the inclusion of a ground training day, with a training captain, prior to the start of Line Flying Under Supervision. This is to ensure time is spent discussing, in detail, technical issues relating to line operations. Winter operations and de-icing/anti-icing will be among the topics covered.
  • The operator provided the co-pilot with additional training before he was allowed to resume line flying duties.
  • The operator is reviewing its requirements for aircraft inspections following use of thickened de-icing fluids.
  • The operator intends to provide better guidance for the aftermath of a serious incident by making changes to its Operations Manual (Part A). This is likely to include a recommendation for a group debrief to take place as a matter of course, so the crew can discuss what happened and what they have learnt.
  • The manufacturer has stated it will contact all operators prior to the start of the next European winter, to promote awareness of the circumstances which led to this serious incident.

AAIB Conclusions

The investigation concluded that ice contamination affected the tailplane and caused pitch control difficulty after the aircraft rotated, on departure. The evidence indicated that this would have been avoided if the aircraft had been de-iced/anti-iced and then inspected carefully before flight.

The crew considered, before parking, that de-icing was probably going to be unnecessary. It may then have become difficult for them to change their assessment because of ‘confirmation bias’, even though they were in freezing conditions and snow was falling. A contributory factor may have been the crew’s lack of experience operating aircraft in such conditions.

The commander optimistically thought that lying snow would blow off the aircraft before rotation; an assessment that was flawed and a possible reflection on the training the pilots had received for such winter conditions. The operator has recognised that recurrent winter training for pilots may have been over-reliant on self-study and has taken remedial action.

Our Comments

It would have been too easy just to say what the crew should have done.  This report shows how a good safety investigation looks at the context of an occurrence and how actions made sense at the time.

An ATR72 (Credit: ATR via AAIB)

An ATR72 (Credit: ATR via AAIB)

Subsequent Icing Occurrence

An further icing related Serious Incident involving ATR 72 G-COBO occurred on 21 December 2016 near Guernsey according to the BEA, who describe a:

Brief loss of control due to ice accretion on the aircraft while in the climb.

This Serious Incident is being investigated by the UK Air Accidents Investigation Branch (AAIB).

UPDATE 4 October 2017: Swiftair ATR 72-500 EC-KKQ, was involved in a stall incident on 9 September 2017 when it was passing through an area with icing conditions while climbing to FL170 en route from Alicante to Madrid.

The crew recovered from the stall and declared an emergency. The autopilot was switched off and the aircraft experienced a sharp drop in altitude and oscillations. The crew declared MAYDAY to the air traffic control services and was given priority for landing in Madrid. During the approach, there was a blockage problem in the vertical stabilizer which was later linked to a TLU (travel limitation unit) operating problem.

This has been classified as a serious incident.

UPDATE 14 December 2017:  The AAIB issue their report on the 21 December 2016 G-COBO Serious Incident:

The aircraft was on a scheduled flight from Guernsey to Manchester. While climbing to a cruising level of FL170 the aircraft began to accrue airframe icing. The crew were presented with a DEGRADED PERF and an INCREASE SPEED caution. The appropriate checklists were not fully actioned and the correct climb speed was not maintained because the crew focused on climbing the aircraft clear of the cloud and icing conditions. The aircraft experienced an in-flight upset whilst levelling at FL130 (as requested by the pilots) and commencing a turn instructed by ATC. The aircraft was subsequently recovered to controlled flight, after which the crew elected to return to Guernsey. There were no injuries.

The loss of control resulted from airframe icing accrued during the climb and incomplete use of the appropriate checklists, leading to selection of an unsuitable speed followed by the use of the LNAV mode of the flight director to initiate a turn.

The operator and manufacturer took several safety actions, including an amendment to the aircraft’s checklist and operating manuals.

UPDATE 4 February 2020: The CIAIAC issued their final report into the 9 September 2017 EC-KKQ.  They determined that the probable cause of the loss of control in icing conditions was a deficient flight management by the crew and an inappropriate use of automation.

Ice Buildup Swift Air ATR-72-212A EC-KKQ (Credit: CIAIAC)

Investigators were critical of the climb management as the climb target speed of 170 KIAS was not maintained.  Instead 180 KIAS and later 176 KIAS were used. The crew also used the PITCH autopilot mode rather than the IAS autopilot mode, even though none of the conditions for PITCH mode existed.

“These are clear indications, and are described as such in the procedure, that the aircraft’s performance was degrading,” the inquiry says, pointing out that the crew had already been issued cautions of ice build-up and increasing drag.  “The crew only considered the visual indications [of icing] and did not take into account the rest of the information.”

Despite the evidence for a loss of performance, the crew commanded a series of actions to force the aircraft to climb beyond its capabilities…“According to the crew’s statement, they were about to clear the cloud layer and thought that climbing – and not descending – was the best option to escape the icing,” says the inquiry, adding that the crew was “completely focused on this objective” and “ignored the cautions” from the aircraft.

Other Safety Resources

We have previously published other articles on icing including:

Icing conditions (ground and in flight) was the topic for a European Aviation Safety Agency (EASAconference in 2013.

UPDATE 4 October 2020: Investigators Suggest Cultural Indifference to Checklist Use a Factor in TAROM ATR42 Runway Excursion

Aerossurance is pleased to sponsor the 2017 European Society of Air Safety Investigators (ESASI) 8th Regional Seminar in Ljubljana, Slovenia on 19 and 20 April 2017.  Registration is just €100 per delegate. To register for the seminar please follow this link.  ESASI is the European chapter of the International Society of Air Safety Investigators (ISASI).

Aerossurance has extensive air safety, operations, design, airworthiness, human factors, and safety analysis experience.  For practical safety advice you can trust, contact us at: