DH8A, vicinity Svolvær Norway, 2010
DH8A, vicinity Svolvær Norway, 2010
On 2 December 2010, a DHC8-100 crew briefly lost control of their aircraft after encountering a microburst and came very close to both the sea surface and a stall when turning onto night visual final at Svolvær during an otherwise uneventful circling approach. After recovery from 83 feet agl, involving an unplanned change of control, an uneventful diversion to an alternate followed. Commencement of an investigation was delayed by failure to report the event at all initially, or fully. It was found that during loss of control, airspeed had dropped to 72 knots and rate of descent had exceeded 2,200 fpm.
On 2 December 2010, a De Havilland Canada DHC8-100 (LN-WIU) being operated by Widerøe on a scheduled domestic flight from Bodø to Svolvær as WIF814 encountered a sudden and severe downdraft and rapidly lost airspeed and altitude as it completed the turn onto visual final following a circling approach at night. Recovery to controlled flight was only achieved after almost stalling. None of the 38 occupants were injured and the attempted approach was then discontinued and a diversion to Leknes made without further event. The crew subsequently continued with a flight to Bodø.
EDITORS NOTE: An animation of the investigated upset was prepared during the Investigation:
The Accident Investigation Board Norway (AIBN) was only made aware of the event in December 2012. On the basis of the information then available, the Board concluded in June 2013 that the event had not been a Serious Incident and did not therefore require an independent investigation. Then, in February 2015, the event "was subject to significant attention" and the earlier "decision not to investigate was changed". In mid-March 2015 it was decided that a reassessment of the occurrence should be carried out and after this initial work, the event was re-classified as a Serious Incident and an Investigation in accordance with ICAO Annex 13 principles followed.
Relevant data on the 30 minute CVR had been overwritten before a download was considered. The SSFDR installed was different to the fleet standard at Widerøe at the time and the Company had arranged for an external agency to carry out the download. The data recovered was central to the Company investigation at the time and was then archived by them and available when the event was reassessed by the AIBN in March 2015. At that point, it was sent to Bombardier who cooperated with AIBN to verify the integrity of the retained data and analyse them. The FDR did not record the geographical position of the aircraft, the Angle of Attack (AOA), activation of the stall warning (stick shaker), or whether flight control inputs were made from the left or right hand pilot positions
It was found that the 43 year-old Captain, who had been PF for the accident flight, had joined Widerøe as a First Officer in 1999 after initial pilot training in the USA in the "early 1990s" and subsequent experience as a pilot in the USA, Africa and Sweden. He had worked as a First Officer based first at Bodø (for three years) and then at Bergen (for six years) before obtaining his command in 2008, after which he was again based at Bodø. Prior to beginning his duty at Bodø on the day of the investigated event, he "got up at 5 a.m. and travelled as a passenger on the Bergen-Oslo-Bodø route before starting his flying duties in the afternoon". He reported that later he had felt tired and had spent a short break prior to the investigated flight - his third sector - resting. The First Officer and the Cabin Attendant both stated at an interview that they had noticed that the Captain had appeared to be tired but the Investigation concluded that on the balance of probability, any such tiredness was unlikely to have had any effect on what happened. The 42 year-old First Officer had joined Widerøe based in Bodø in 2003 and after obtaining a DHC-8 type rating on arrival had accumulated 3,090 hours on type out of a total of 3,940 hours. With the exception of the Captain's tiredness neither pilot made any "negative remarks concerning their own health condition (or) how they were feeling on the day".
It was noted that the general weather situation in the area between Bodø and Svolvær was a showery south-westerly near gale force airflow with CB clouds, variable surface wind direction and gale force gusts. The METAR for Svolvær around the time of arrival gave a surface wind from 240° at 30-44 knots, visibility 8km in light rain and hail showers with the lowest cloud at 800 feet agl, scattered CB at 1,200 feet agl and broken cloud at 1,400 feet agl and a temperature of 5°C.
It was established that after about 20 minutes in flight and with the prevailing south-westerly surface wind now within company limits for landing which it had not been earlier, the approach to Svolvær had been commenced with an offset LOC approach to runway 01 (MDA 580 feet) followed by a break to the right into the "Precision Circling Approach" (see the illustration below) for a landing on runway 19 - LDA 776 metres and slightly uphill. Use of this 'Precision Circling Approach' is a special regulatory concession granted to Widerøe which allows circling approaches within, among other things, a smaller obstacle-free area compared with the ICAO standard for the DH8-100 and DHC8-300 operations and this facilitates a lower minimum altitude for circling.
Dark night conditions prevailed for circling but the in-flight visibility below cloud was good with isolated lights in the area and the red obstruction lights close to the 19 final approach (see the illustration below) readily visible and visual reference with the runway and its surroundings obtained well before reaching the MDA of the LOC approach. Circling was made downwind at 600 feet with the AP remaining engaged and with a target airspeed of 110 KIAS - the applicable Vref was 100 KIAS. The landing gear was lowered and the flaps were selected to 15. There was no "particularly strong turbulence, neither en route nor while in the circling pattern" and there had been no discussion about whether the airspeed should be increased as is sometimes done "in order to increase margins in strong winds and with unstable wind conditions". The AP was disconnected before turning base i.e. from a northerly to a westerly track, so that the only remaining configuration change was to select flaps 35 on final approach. The altitude was kept at 600 feet, the Captain subsequently explaining that it had been his intention to maintain this until the turn onto final had been made and the position achieved had been confirmed as suitable for a landing. This was compatible with Company requirements that specified that "base leg shall be adjusted so that, after the turn to the final approach, one has level wings no later than at 300 feet above the runway height, in position for normal glide path with flap setting 15 or 35 degrees”.
Both pilots agreed that the upset had occurred as the turn onto final was almost complete and this recollection was supported by FDR data and other sources of evidence. FDR data showed that the single upset event had occurred suddenly and without warning whilst the aircraft was making a gentle turn left and descending at an average of 550 fpm at an airspeed at or just below 110 knots with the engines set to 42% torque. Then, over a 10 second period, airspeed had quickly dropped to 4½ knots below estimated stall speed for the aircraft as loaded and configured and that the aircraft had come a lot closer to sea impact than either pilot had realised - 83 feet radio height when estimated to have been 0.8 nm north northeast of the runway threshold which has an elevation of 14 feet amsl. The severity of the downdraft experienced and a rate of descent in excess of 2,200 fpm appeared to have exceeded the initial corrective response of the crew. The successful recovery, which eventually used all available engine power "was abrupt and the aircraft was subject to high g-forces" slightly above the limits requiring a structural inspection of the aircraft. At some point after the stick shaker had activated, which could not be exactly determined with any certainty, the First Officer had taken control until the aircraft had been fully recovered to normal flight tracking more or less along the extended runway centreline and overhead the runway.
At this point, it was agreed that a diversion to Leknes should be made and the Captain had again taken control. The diversion was without further event and the aircraft landed at Leknes approximately 23 minutes after the upset had occurred. It was noted that after completion of the flight, despite both pilots subsequently acknowledging that they had been "shaken" after the incident, they had then operated back to Bodø. Then "upon arrival at Bodø, the three crew members went to the crew room together and found a vacant room where they could sit and talk together". The First Officer was then rostered off duty anyway, the Captain "skipped" his next rostered two sectors but had then completed the rest of his flying duty and the Cabin Attendant, who had 10 years of experience in that role with Widerøe, "opted to drop out of the flight program for the rest of the evening". She had been familiar with "downdrafts" but noted that in this case, the expected recovery had not accompanied the usual increase in engine noise and "through the window on the right side of the aircraft she could see that they passed a red light, and she remembered that one of the passengers in the forward row commented "that light is not supposed to be there". Given the estimated position of the upset, the red light seen was likely to have been the lower of the two obstruction lights visible in the view of the runway on the photograph above and marked as being at 115 feet amsl on the Precision Circling Approach chart also shown above.
It was noted that no EGPWS activation appeared to have occurred but also that warnings are blocked if the stall protection system is activated. FDR data showed that a number of exceedances of AFM limitations for flaps and g loading had occurred. There was no evidence that AMM requirements for "inspections after severe turbulence or buffeting" had been actioned and the aircraft involved had remained in service.
Taking account of all sources of evidence available to the Investigation, a 'likely scenario' was established and summarised as follows:
- The aircraft was exposed to severe wind shear.
- The commander executed "recovery" with full power, but the aircraft continued to lose both altitude and speed.
- The stick shaker (stall warning) activated.
- The commander pushed the control column forward resolutely, before pulling it backward again.
- The aircraft accelerated while its nose fell to 14 degrees below the horizon.
- The First Officer considered that the commander's corrections were not sufficient, and that there was an acute danger of the aircraft crashing into the sea. At one point, he took over the flight controls and increased the engine power to the maximum available.
- The result of the combined control movements made under the prevailing wind conditions was that airspeed increased, the loss of altitude was stopped in time and the aircraft climbed rapidly.
A large number of observations on various aspects of the findings from the Investigation as developed into the above scenario were made. These included the following:
- Whilst flying in darkness with sparse visual references, the conditions were right for the Captain to have experienced a somatogravic illusion. If this was so, the Captain and the First Officer, who did not experience a somatogravic illusion, would have perceived parts of the critical phase differently.
- It was apparent that the Widerøe Precision Circling Procedure was based on thorough considerations, but the Investigation considered that on a dark night, the absence of a visual horizon and little chance of making out the terrain profile below seemed not to have been emphasised. In addition to problems associated with sensory illusions, two other aspects relevant to circling in darkness were identified as:
- Monitoring of aircraft altitude in the base turn is problematic.
- Weather Radar, although it was available in this case, does not provide a practical means of detecting convective cloud bases and strong squalls at low level in darkness and the difficulty of such detection remains.
- The possibility of further reducing the risk associated with a combination of darkness and unstable atmospheric conditions should be recognised as different to that in good weather and in daylight with a visible horizon. In particular, missed approaches in darkness may add to the general risk if as in this case they start off with visually-based manoeuvring while climbing in turns near the terrain which can be challenging even in daylight.
- Studies have shown that in a fright/threat situation, time is perceived to pass more slowly. There is great variation as regards how much a person perceives time to slow down. One or more crew members may have experienced a distorted time perception due to the distressing event. Both the pilots and the cabin attendant conveyed an impression that the incident lasted significantly longer than the 10 seconds established by the flight recorder.
- It is likely that the upset occurred when the aircraft flew in under a CB cell which was not apparent to the crew in the dark and was exposed to severe wind shear - a microburst. The aircraft was probably exposed to both downdraft and horizontal wind shear. The tendency for increased mechanical turbulence when strong south-westerly winds prevail in the area may have made the situation worse and possibly also extended the duration of the adverse external affect. It may be that the crew did not sufficiently consider the additional threat of localised windshear below heavy shower clouds during the approach and were surprised by the powerful wind shear it triggered.
- The windshear experienced was localised and not directly related to either actual conditions over the runway or whether the surface wind at the aerodrome was inside or outside the Operator's limitations when the approach formally started. The full flight simulator used by Widerøe for pilot training could not realistically simulate wind shear during a circling approach in mountainous terrain as Svolvær.
- Although an AoA of up to 30 degrees is indicated based on the recorded airspeed, vertical speed and the position of the aircraft nose above the horizon, it is impossible to perform meaningful aerodynamic calculations of lift and the wing angle of attack for the few critical seconds when the aircraft was exposed to strong variable external forces, increased slipstream and rapid pitch variations. Since at that time, the aircraft was affected by a strong downward vertical wind component, the AoA was probably lower. Likewise, the parts of the wing that were inside the propeller slipstreams would have had a significantly lower angle of attack.
- Given that in an aerodynamic stall, a significant reduction in G-load would be expected and the FDR data does not show this, it is possible that external forces camouflaged the expected G-load pattern. Stick shaker activation certainly supports the aircraft having been on the verge of stalling and the possibility of there having been a momentarily stalled condition cannot be ruled out on the evidence available.
- During the critical phase, it was not surprising that normal dialogue and coordination between the two pilots ceased completely. In such circumstances, a takeover of control may prevent the situation from becoming an accident. It was considered that when the aircraft's pitch was very low and it was just above the sea, the First Officer's impression that the Captain's intervention was insufficient indicates that it was correct to intervene. This action was in accordance with good CRM and also in line with the Widerøe approach to CRM training. The view was taken by the Investigation that in this case, it was sensible of the Captain not to oppose takeover of the flight controls by the First Officer. However, it was acknowledged that takeover of flight controls is a complex issue about which it is impossible to give general advice that applied to all circumstances - a fact that becomes evident when taking into account the risk of sensory illusions during visual manoeuvring in a dark night environment.
The formally stated Conclusion of the Investigation was as follows:
The aircraft was exposed to severe wind shear from a cumulonimbus cloud (microburst). This resulted in the aircraft losing speed, forcing it down towards the terrain as it entered the turn to final approach during visual circling in darkness. The wind shear could perhaps have been predicted, but it would have been difficult to detect with the equipment that was available.
The commander reacted quickly. First he (responded) by increasing the engine power, but the aircraft was still on the verge of stalling at low altitude. The commander then responded by resolutely pushing the control column forward. It was correct and necessary to lower the nose of the aircraft to prevent, or recover from, stalling.
The manoeuvring to regain control of the aircraft took place with sparse visual references and with no visible horizon. During a few critical seconds in the recovery phase, the commander may have been exposed to somatogravic illusion. However, the Accident Investigation Board has not found any evidence which warrants a conclusion that sensory illusion had an impact on the way the wind shear was handled.
At some point, the First Officer intervened and took over the flight controls. Based on the available facts, it has not been possible to determine whether this affected the outcome.
The Accident Investigation Board is of the opinion that the combined actions of the crew averted an accident. A marginally longer response time and/or less resolute application of engine power would probably have resulted in collision with the sea.
In the six year period since the event occurred, a considerable number of relevant risk mitigations have been introduced. These have included:
- The installation at Svolvær of improved guidance lights for the circling base leg and new lead-in lights to Runway 19.
- The installation of a Double Pulse Light Approach Slope Indicator (PLASI) for runway 19 at Svolvær.
- The extension by Widerøe of their mandatory DHC8-400 OFDM programme to the smaller DHC8-100 and -300 variants.
- The introduction by Widerøe of a new wider circling track which makes better use of the available over-water area for the downwind leg and the turn onto final.
The Final Report was completed on 22 November 2016 and published initially in Norwegian and subsequently in English translation. No Safety Recommendations were made but it was noted that "this investigation concerns Svolvær in particular, but a learning organisation will, as part of its safety work, use experiences from 'local' findings to assess whether there is a need for measures at other destinations or operations in general". The Operator concerned in this case, Widerøe, is encouraged "to also apply the lessons learned in this incident in a broader perspective (with) particular focus […] on the risk of sensory illusions in combination with complex and demanding missed approach procedures".
- Loss of Control
- Stall Warning Systems
- Aerodynamic Stall Awareness and Avoidance
- Recovery from Unusual Aircraft Attitudes
- Low Level Wind Shear
- Squall Line
- Circling Approach
- Sensory Illusions (SKYclip)
- Somatogravic and Somatogyral Illusions
- Angle of Attack (AOA)
- Radio Altimeter
- Crew Resource Management