On 14 November 2016, an ATR72-600 (OY-JZC) was being operated by Jet Time for SAS on a domestic passenger flight from Bergen to Trondheim via Ålesund as SAS 4144. During the first leg of the flight, the crew continued climbing in day IMC after encountering severe icing conditions and temporarily lost control of the aircraft before achieving a recovery and completing the remainder of the flight. No passenger injuries were sustained but there was widespread alarm amongst the passengers during the loss of control.
The event was not notified to the Accident Investigation Board Norway (AIBN) until two days after it had occurred, when the aircraft operator contacted the Board. The Captain had originally submitted a ‘Flight Safety Report’ to Jet Time and the Danish Transport Construction and Housing Authority. The operator later upgraded the report to a Serious Incident. The AIBN was renamed the Norwegian Safety Investigation Authority on 1 July 2020 by which time this Investigation was almost complete so no corresponding name change was made to the text of the subsequently published Investigation Report. Relevant data was obtained from the SSFDR, the QAR, the FDAU (Flight Data Acquisition Unit) as well as from crew interviews, passenger interviews and ATC voice and radar recordings. The Investigation “did not have access to recordings from the CVR” for which relevant data would likely have been overwritten given that the aircraft continued in service.
It was noted that the 60 year-old Captain had a total of 9,423 hours flying experience of which 3,525 hours were on type. He had been employed by Jet Time since 2014, held a Swedish-issued ATPL (Air Transport Pilot Licence) but had not yet completed the operator’s annually-required UPRT (Upset Prevention and Recovery Training). Prior to joining the operator he had previously flown ATR 42/72 aircraft as well as other turboprop types including the ATP and the Jetstream whilst working for “a number of different airlines in Sweden, Norway, Finland and Denmark”. The 51 year-old First Officer had a total of 7,000 hours flying experience of which 5,850 hours were on type, although only 270 of these on-type hours had been on the -600 version of the ATR 72. She had been employed by Jet Time for six months after working for another Danish operator, held a Danish-issued CPL (Commercial Pilot Licence) and had completed the operator’s annually required UPRT about two weeks before the event under investigation occurred. Jet Time had previously acquired a small fleet of new ATR 72-600 aircraft in SAS livery specifically for the SAS contract but this contract was terminated later in the same month the event under investigation had occurred.
Both flight crew had arrived in Bergen the day before and stayed in a hotel overnight. During flight preparation the next day, they noted that moderate icing was forecast for the first flight leg. Whilst taxiing out in rain with a 7°C surface temperature, the electric anti-icing systems for the propellers, windscreens, all flight control horns and key external sensors were switched on and the pneumatic airframe de icing system, which removed ice from the wing, horizontal stabiliser and vertical stabiliser leading edges, was cycled.
After takeoff from runway 17 with the First Officer acting as PF, the aircraft was cleared to climb to FL190 and established on the required northerly track with the AP engaged and a climb speed of 170 KIAS selected. Passing FL100, the aircraft ice detector was activated and the crew responded by switching on airframe de-icing. The crew observed that as the climb continued, ice started to form on the aircraft as it passed FL120 and the rate of climb began to reduce. Passing FL127, FDR data recorded a 765 fpm rate of climb and the Investigation noted that “the data show that during this period the vertical speed dropped by half in 30 seconds”.
Passing FL137, the Aircraft Performance Monitoring (AMP) system displayed the alert ‘DEGRADED PER’. Almost immediately, the First Officer reported having seen “two streaks of water or ice running on her window” and having interpreted this as an indication of freezing rain, she had then informed the Captain. At the same time, both pilots reported having noticed that ice was forming on the (electrically heated) flight deck side windows. Selected airspeed was reduced to 165 KIAS and the power was set to ‘Maximum Continuous’ (MCT) which increased the propeller rpm and would have assisted ice shedding from the propeller blades.
Passing FL160 at 164 KCAS, an ‘INCREASE SPEED’ APM alert appeared and the aircraft was levelled off at that altitude, the selected speed was increased back to the original 170 knots and “after a short discussion, the crew agreed to terminate the climb”. They also decided to reduce altitude slightly in order not to lose any more speed and on obtaining ATC clearance, a 500 fpm descent towards FL150 was started with the AP remaining engaged and the engine power was reduced from ‘Maximum Continuous’ to ‘Cruise’. Once at FL150, the 170 knot airspeed began to decrease as the angle of attack increased. The failure to accelerate or even hold the selected speed was accompanied by “more and more ice forming on the airframe”.
ATC were advised of the problem and a change of track to the west towards the sea was requested in the expectation that this would be likely to reduce the severity of the icing. This was approved and a left turn onto a heading of 330° was selected but the AP then initiated the turn at an abnormally high bank angle for the airspeed and 12 seconds into the turn, it automatically disconnected. The left bank “increased abruptly” and simultaneously the stick shaker activated. The local angle of attack had reached 11.8°.
As PF, the First Officer responded with an attempt to move the control column forwards and to the right whilst also making a right rudder pedal input. However, the PM Captain simultaneously pulled back on his control column and the bank angle increased uncontrollably to a recorded 68.2° left. The First Officer reported finding that “it was unusually difficult to push the control column forward and wondered if ice had formed on the horizontal tail”. It appeared from their statements that neither pilot was aware of the fact that they had made opposing pitch inputs, both stating that they thought “cooperation and communication between them were good”.
During the sharp increase in left bank, the aircraft pitched down to a recorded -3.3° and when the left bank then changed to a right bank, the pitch attitude decreased further to a recorded -8.1°. The angle of attack had increased to 14.5° and the stick shaker and stick pusher then activated for two seconds which recorded data showed had corresponded to the Captain opposing the First Officer’s continued forward control column pressure. The right bank reached 66.2° before reversing to the left, this time reaching only 36°. The First Officer continuously responded to the uncontrolled banking with opposite rudder and aileron inputs. After ceasing for 3 seconds after their first activation, the stick shaker and stick pusher activated again with the local angle of attack at 15.9° and again were found to correlate with opposing pitch input by the Captain. Pitch attitude decreased further to -11.9° and the aircraft quickly lost altitude whilst the forward speed increased to 190 KCAS and the rate of descent to a maximum of 6,448 fpm before the pitch attitude began to return towards the horizontal after losing in excess of 1,500 feet of height. However, since the applicable MSA was 5,800 feet, there had been no risk of collision with terrain.
During recovery from the apparent stall, pitch attitude reached a recorded +12.1° but speed reduced and when increasing power to maximum continuous again was followed 13 seconds into the climb by another APM ‘INCREASE SPEED’ message appearing and the speed had dropped to 150 KCAS, the crew eventually decided to descend. Clearance to descend to FL100 was obtained with approval to operate within the range FL100 - FL140. The track of 330° was continued and on descent through FL127 at a recorded 212 KCAS, “the ice that had formed on the airframe started to disappear” and the AP was re-engaged 3 minutes and 23 seconds after it had disconnected. By FL110, all the airframe ice visible had disappeared and the APM ‘DEGRADED PERF’ alert, which had been continuously active for around 15 minutes ceased.
The Captain made a short PA and the flight subsequently reached Ålesund after a total of 48 minutes airborne without further event. During the 20 minute transit stop there before continuing to Trondheim, the First Officer went into the cabin and informed the cabin crew that the aircraft “had been exposed to icing” but did not provide any further details about what had happened and other than the Captain’s earlier brief PA, the passengers were given no further information. Also, since neither pilot reported having “felt uncomfortable” about continuing with the remainder of their four sector duty that day, “neither considered declaring themselves unfit for further service” and completed it.
The Investigation interviewed two passengers who had been on the flight and were very familiar with such flights as passengers, one of whom was also a PPL (Private Pilot Licence) holder. He had been seated in row 2 on the right and the adjacent seat was unoccupied. From this position, he had been able to see the right engine and propeller as well as most of the wing leading edge outboard of the engine nacelle. He stated that on departure from Bergen, cloud was soon entered and that he had then seen “an increasing amount of white ice forming on the leading edge of the wing as well as on the propeller spinner and the engine nacelle”. He also observed the operation of wing leading edge de-icing boots and heard the sound of what he had presumed was ice shed from the propeller blades hitting the side of the fuselage. At one point, he thought he had detected an increase in propeller rpm and a simultaneous reduction in ice shedding impacts. He also “observed that the de-icing boots did not manage to remove all the ice (and) saw that the ice along the wing leading edge changed character and colour (and) seemed clearer and more translucent”.
As the aircraft stopped climbing and appeared to level out, he noticed that it “started buffeting (and) identified this as the initial stage before a stall” finding it “strange that the crew did not lower the nose of the aircraft to prevent this”. When he saw the wing drop and the nose of the aircraft pitched sharply down, he was “convinced that the aircraft had stalled” and had “feared the worst and that he would not survive”. He also stated that “other passengers on the aircraft were screaming”'. After the recovery, he heard the Captain make a PA announcement '“informing the passengers that the crew had regained control of the aircraft” and having “interpreted this as confirmation of the severity of the situation that had occurred” he had expected to be given more information about what had happened on arrival at Ålesund and “expressed disappointment that this did not take place”. The second passenger interviewed had similar but “less technical” recollections and had similarly formed the opinion from the “severe” wing drops, sharp nose pitch down and the Captain’s subsequent very brief PA after it was over as “confirmation that the incident had been serious”. He added that he had “never experienced anything like it before”.
An animation of the loss of control trajectory (download, 200mb) was created during the Investigation.
The Context for What Happened
The forecast weather for the route valid for the time of departure included a locally produced forecast significant weather chart and an AIRMET which both indicated that moderate icing should be expected en route. The route was expected to be just behind and parallel to an active eastward-moving warm front. Following ATC awareness of the investigated event, an AIREP SPECIAL was issued for severe icing at the event location encountered by the AT76 at FL150.
The Norwegian Meteorological Institute, which was responsible for aviation weather service, subsequently informed the Investigation that whilst they had concluded that moderate icing was the appropriate forecast, especially in areas where the orographic lift was strongest, this icing was expected to be “close to severe”. They added that after the report of the investigated severe ice encounter was received, “new assessments were made according to the procedure, and a ‘SEV ICE’ SIGMET was issued”.
It was noted that five minutes before the ATR 76 took off from Bergen, a rather smaller passenger aircraft, a DHC8-100 operating a Widerøe flight to Kristiansund, had taken off and followed a parallel, but slightly more easterly, route north. The Captain of the Wideroe flight informed the Investigation that after climbing for about 25 minutes, the planned cruise altitude of FL230 had been reached without airframe icing “being of a particularly noticeable nature". The climb profiles of the two aircraft are shown on the illustration below which clearly shows the reduced climb performance of the ATR 72 above FL110 compared to a normal climb, which could have been expected to continue similarly to that of the DHC8 had the icing conditions not affected its climb rate. It was notable that the DHC-8 had passed close to the location of the ATR 72 loss of control three minutes before it occurred but by then it was at FL230.
The vertical profiles of the ATR72 (SK4144) and the DNC8 (WIF564). [Reproduced from the Official Report]
It was noted that following the introduction of the ATR 72 in 1989, airframe icing risk soon became a focus for action after the fatal accident to an ATR 72-300 at Roselawn in the USA in 1994 when the NTSB Investigation found that the roll upset which caused the accident had been due to an aileron hinge moment reversal after ice accretion on the upper wings aft of the leading edge pneumatic de-icing boots during earlier holding in icing conditions which had been - unknown to the crew - outside the icing certification envelope.
Having modified the aircraft design to address this problem, ATR subsequently began a wider review of both the design and the operating procedures relevant to flight in icing conditions. Design modifications were made to improve banking stability in both icing conditions and in the event of stalling and the introduction of the Aircraft Performance Monitoring (APM) function, which was specifically configured to address avoidance of severe icing conditions. This does not use any dedicated sensors or any calculation of atmospheric ice content, it simply compares actual aircraft performance with the expected performance and computes the actual minimum icing and severe icing speeds for the given flight conditions. These data were the source of the ‘DEGRADED PERF’ and ‘INCREASED SPEED’ messages annunciated during the investigated event. Icing-related SOPs were also modified. A new memory item was added to both the stall recovery procedure and the abnormal roll control procedure to require the selection of flaps 15 if not already set and, since it was considered that use of the AP may mask the signs that control is about to be lost, the severe icing procedure was modified to require that it must be disengaged if such conditions were encountered. All changes as a result of this effort were incorporated in the ATR 72-600 which remains in production.
The Investigation did note a potential lack of clarity in respect of the QRH drill for recovery from a wing drop and/or stall in that it was found to make no mention of aileron or rudder use whereas the Flight Crew Operating Manual (FCOM) contained a cautionary note about the use of rudder which stated unequivocally that, unless an aileron jam has occurred, rudder should not be used to induce or counter roll, adding that “aggressive, full or nearly full opposite rudder must not be applied” because such inputs can lead to control loading beyond the limit or possibly the ultimate limit and cause structural damage or failure. Given the First Officer’s use of both aileron and rudder to oppose the rolls experienced when stalled, the Investigation raised the issue and ATR then made it clear that only the ailerons must be used to level the wings and that, in such a situation, the use of rudder should be avoided.
The flight crew’s decision making prior to and during the flight and aircraft handling during it were both considered and some observations, including but not limited to the following, were made (in summary) as follows:
- Both pilots told the Investigation that “they hoped to climb above the cloud and thus above the icing conditions” and seemed to have been surprised that this did not happen when they expected it to. This was considered to be an indication that they had not made a sufficiently thorough pre-flight assessment of the relative severity of icing on the track to be used or how high up icing would occur or that it may (as it appears to have done) increase in intensity above FL100, something to be expected when relatively warm maritime air rises over mountainous terrain.
- Although it was not clear whether there had been any visual signs of ice accretion before the ice detector warning light illuminated, the FCOM procedure requires activation of the airframe de-icing at the first indication of ice accretion without waiting for the ice detection to be activated.
- Over the next few minutes, there were multiple indications that icing was developing into a problem and soon there were two clear indications that severe icing had been encountered and that the icing intensity had surpassed the capacity of the aircraft ice protection systems. These indications were the decreased rate of climb and the decreasing airspeed which had activated the ‘DEGRADED PERF’ APM alert and the appearance of streaks of water/ice along the window, both mentioned under ‘Detection’ in the Severe Icing Checklist. The required procedure was to disconnect the AP and immediately exit icing neither of which was done.
- Having continued to attempt to climb for a further one thousand feet in circumstances where it was completely contrary to procedures to do so, on levelling off and having to accept that further climb was impossible, the ‘INCREASE SPEED’ APM alert was activated. It became necessary to use the ‘SEVERE ICING’, ‘DEGRADED PERF’, AND ‘INCREASE SPEED’ Checklists one after another. It was noted that one item in all of these checklists was a requirement to ensure that the indicated airspeed remained at least 10 knots above the preset ‘Icing Bug’ speed which on this flight was 156 knots, but this was not done.
- Having descended back to FL150, using the ‘Cruise’ power setting, power was not returned to ‘Maximum Continuous’ despite still being in severe icing conditions which required this.
- On levelling at FL150, the altitude hold mode of the AP was selected despite still experiencing icing. It was noted that as the AP gradually and imperceptibly increased the angle of attack in order to compensate for the loss of lift and increased drag as ice built up, this, in turn, would have exposed the aircraft to the further icing that eventually led to loss of control.
- Once at FL150 and having decided to change course seawards, there was a delay talking to ATC to arrange this which “resulted in unnecessary loss of time in an already tight time frame”. Since this delay and the ice accretion which occurred was enough to create the conditions for a stall, it was concluded that following the old, but still current, pilot task prioritisation mantra of "Aviate-Navigate-Communicate" had been applicable.
- When the right turn was commenced, it was made using the HDG function of the AP which, as the angle of attack continued to increase resulted in an abnormally high bank angle of around 30°.
- As the airspeed dropped in the turn, the angle of attack eventually reached the stick shaker threshold which automatically disconnected the AP and immediately released the aileron deflection required for the turn. It was not clear whether this or the wings being stalled (or close to stalling) had caused the sudden left wing drop, although the latter was considered as an explanation for the subsequent sharp roll to the right.
- It was not possible to determine whether the First Officer’s prompt response to the wing drop had improved the situation. FDR data showed no indications of aileron hinge moment reversal having occurred and it was concluded that it was not a factor.
- The angle of attack may have been increased during both stick shaker activations by the Captain opposing the First Officer’s pitch down attempt and may have contributed to the first sharp roll to the left when it was not reduced quickly enough. The Investigation was “not able to determine whether the aircraft actually did experience an aerodynamic stall, or if the nose pitched down due to elevator inputs”.
- Neither pilot seemed to have noticed that they were applying opposite forces on the control column nor the two stick pusher activations. Two important memory items from the ‘Stall or Abnormal Roll Recovery’ Checklist, Flaps 15° and Power to Maximum Continuous were omitted. It was noted that it is unusual for extending flaps to be part of a stall/abnormal roll recovery procedure but considered that this may have a stabilising effect on aircraft that lack sufficient lateral stability near the critical angle of attack.
- It was concluded that the Captain’s inappropriate initial response may have been a result of startle effect.
- It was noted that there seemed to have been no callouts when the loss of control occurred. Relevant calls would have been ‘STALL’ when the stick shaker activated and if the Captain had wanted to take over control he should have called ‘MY CONTROLS’ and received the acknowledgement ‘YOUR CONTROLS’. Timely callouts may also have prevented ‘startle effect’. It was found that Jet Time had standards for callouts in the OM but none referred to these two circumstances.
- It was, however, recognised that the crew had “managed to recollect themselves and gain control of the situation, thus averting an accident”.
The Investigation considered that it was “hardly a coincidence that the First Officer, who had recently completed simulator training as part of her UPRT requirement, was the one to respond correctly when she tried to push the control wheel forward when the stick shaker activated”'. It was concluded that “considering the occasionally challenging flying conditions along the Norwegian coast during the icing season, it would probably have been better had the Captain not been assigned these routes until he had completed this annual training requirement”.
Finally, the Investigation noted that it had previously been concerned at the effectiveness of regulatory oversight of ACMI (Aircraft, Crews, Maintenance, Insurance) operations like the one in this case where an operator from another jurisdiction undertakes flights for an operator’s flights, in this case with aircraft bearing an external appearance to the lessor’s own aircraft. Reference was made to this issue in the investigation of both a 2008 Sikorsky S61 event near Bodø in 2008 and a BAe 146-200 event at Stord in 2006.
The concern in both the current investigation and these two earlier ones was the risk of inadequate operational safety oversight when fully crewed aircraft are operated on behalf of a Norwegian AOC holder by aircraft approved for operations under the AOC of a foreign operator. The operational risks foreseen previously and reviewed again here were:
- The unfamiliarity of foreign operators with Norwegian winter conditions.
- A lack of information transfer between the two aviation authorities with an interest in such an operation.
- The possibility of the object of inspection, in this case Danish operator Jet Time operating aircraft based long term in Norway “falling between two stools”.
In respect of the current Investigation, an explicit concern was that “icing sensitivity did not appear to have been given special attention", either by Jet Time, SAS or by the Scandinavian regulatory authorities and it was not a subject included in the Danish regulatory oversight regime for Jet Time prior to the investigated event. It was therefore stated that the AIBN would like to call attention to the following relating to operations during the icing season:
- The Norwegian icing season must be characterised as challenging, particularly along the coast.
- The icing certification of the majority of existing aircraft types give little or no guarantee of the aircraft's resistance and performance should it encounter icing conditions involving SLD.
- This and two other serious icing incidents involving the ATR 72 500/600 (see below) illustrate the importance of monitoring airspeed and climb in order to operate this aircraft within its defined performance limitations.
Recent Similar Occurrences
It was noted that similar airframe icing loss of control events have occurred to two other ATR72 aircraft, an ATR72-500 in the UK in 2016 and an ATR 72-500 in Spain in 2017.
In addition to the fact that the commander of the aircraft in the Spanish event also initially reacted by pulling the control wheel back when the stick pusher activated, a number of significant similarities were observed between both these events and the one under investigation:
- Prior to losing control of the aircraft, the crew had tried to climb out of the icing conditions.
- The aircraft autopilot was engaged in severe icing conditions.
- The airspeed dropped to below Vmin-ops + 10 knots.
- The aircraft went into a combination of large uncontrolled banking excursions and pitching oscillations.
It was also noted that both of these aircraft were equipped with the same upgraded icing protection system as the aircraft which was the subject of this investigation.
The main Conclusions of the Investigation included the following:
- The loss of control was the result of a combination of insufficient planning and inappropriate decisions en route, particularly the crew's attempt to climb above the icing conditions despite degraded aircraft performance and the continued use of the autopilot when it should be been disconnected.
- Recovery of control may have been impeded by the Captain’s initial response of pulling back on the control column as the stick shaker activated. It is likely that he became startled when the stick shaker activated and the autopilot automatically disconnected whilst at the same time, the aircraft suddenly banked sharply and simultaneously pitched nose down. He may consequently have pulled the control wheel back due to the so-called startle effect.
- The non-optimal recovery of control was affected by the omission of two memory items on the checklist for stalling/abnormal roll control, the deployment of flaps and the increase in engine power.
- In-flight icing should be a priority item in risk analyses for airlines when planning to operate in Norway during the icing season, and that it is important to take the characteristics of the aircraft type into account. Such analysis should conjointly consider the routes and flight levels flown, expected icing conditions and mitigation of the consequences of adverse en route weather conditions including icing, taking account of the aircraft type involved and its specific performance capability.
Safety Action taken by ATR following this event was noted as having included the following:
- The issue of the revision 3 of the ‘Aeroplane Upset Prevention and Recovery Training Aid - AUPRTA’ in cooperation with ICAO, Airbus, Boeing, Bombardier and Embraer in February 2017 and the adoption of the AUPRTA Manual procedures with a view to identifying objective criteria for providing better decision support for crews.
- The speed margin for high bank angle protection now takes into account the APM degraded performance alert so that if this is active, the high bank (30°) speed threshold will increase to the applicable manoeuvring speed in icing conditions plus 20 knots instead of plus 10 knots.
The Final Report was published on 9 September 2020. No Safety Recommendations were made.
- AT73, en-route, Roselawn IN USA, 1994
- S61, vicinity Bødo Norway, 2008
- B462, Stord Norway, 2006
- AT75, vicinity Manchester UK, 2016
- AT75, en-route, near Almansa Spain, 2017