ANALYSIS: Go-around is supposed to be an escape, not a trap

This story is sourced from Flight International
Subscribe today »

Go-arounds are supposed to be the safe route out of an approach that is not going as smoothly as it should. Accident statistics, however, reveal this ostensibly simple manoeuvre has often gone catastrophically wrong, whereas an unstable approach continued to landing frequently causes damage to the aircraft, but no fatalities.

When - some five years ago - the Flight Safety Foundation initiated an industry campaign to reduce the causes of runway excursions and overruns, one of the facts established was that they frequently followed an unstable approach - usually too high or too fast, or both. An inescapable conclusion was that, if crews took note of the advice regarding unstable approaches, one of the potential effects would be more go-arounds. European airport statistics show that one or two go-arounds occur for every 1,000 approaches, which works out at about one go-around per year for short-haul pilots and one every five to 10 years for long-range crews.

Until recently, it had always been assumed that go-arounds were easy, and since a regular recurrent training exercise in the simulator has always been - and still is - an approach with a single-engine failure followed by a late decision to go around, it was also assumed that such a practice manoeuvre should ensure individual pilot competency in the event of a go-around. However, the evidence shows it is not as simple as that. First of all, crews in the simulator actually anticipate a late go-around because it is a statutory exercise, whereas in the real event the need for a late go-around will be a surprise or, at best, a late decision. The ultimate in late go-arounds is carried out just above the runway, or can even follow a bounced landing. Such late go-around decisions, Flightglobal's Ascend Online accident/incident database reveals, have resulted in multiple severe tail-scrapes in the past few years.

Besides which, an all-engines go-around generates its own problems, such as the sudden high rate of climb and forward acceleration combined with a powerful nose-up pitch moment generated by the below-wing thrust line, and - especially at night or in instrument meteorological conditions - that giddying sensation known as a somatogravic illusion, which can lead to disorientation and loss of control.

There are many examples of the failure of crews to cope with go-arounds (see below). Some of these were catastrophic, while others came close to catastrophe but the crews recovered control in time to save the aircraft. One of the common factors in all go-arounds, especially those following a decision made on short final approach, is the lack of height - and therefore time - in which to recognise a destabilising mistake and correct it.

There is a dichotomy here: runway excursions and overruns are aviation's most common accident type by far, but most are not fatal; whereas the risk associated with a go-around may be relatively low but mishandling could lead to catastrophe. A recently published report by Dutch research agency NLR confirms that, despite all recent efforts to increase awareness of the risk, global numbers of runway excursions are not reducing, and that close to 90% of them are associated with landing.

As Capt Bertrand de Courville, of Air France's corporate safety department, observes, a successful campaign to encourage go-arounds from imperfect approaches tantalisingly promises a potential reduction of about 25% in overrun/excursion landing accidents. "No other single defence could have that impact," he says. But, on the other hand, a badly executed go-around could lead to a catastrophe. So what does an airline tell its crews to do?

De Courville describes this shifting of risk from one manoeuvre to an alternative as being a systemic issue, whereas crews operate in the immediate operational environment.

He explains: "We are not giving guidance to our pilots on the basis of go-around-related risks. Go-around decision-making is already difficult. We should keep it as simple as possible and be very cautious before adding more complexity."

"The best strategy at this stage is to encourage and train - in real time - a TEM [threat and error management] approach to the go-around. We still have good potential to make the go-around more robust by doing this."

But what training is appropriate? What skills does a go-around manoeuvre demand to ensure it can be accurately and safely flown under all circumstances? One way to find out is to examine the strategy adopted by Thomson Airways - then Thomsonfly - after one of its crews almost lost control of a Boeing 737-300 while attempting a go-around from an approach to Bournemouth airport on England's south coast

The aircraft was on an autopilot/autothrottle-linked instrument landing system approach, when the autothrottle silently disengaged and the airspeed started to drop until the stall warning sounded.

By that time the autopilot had trimmed the horizontal stabiliser to a nose-high setting commensurate with the low airspeed, so when the crew applied full power to go around, the aircraft pitched up dramatically to a dangerously nose-high attitude. Control was ultimately regained and the crew returned to land safely.

Following an investigation, Thomson arranged for pilot eye-tracking to be carried out in its training simulators and discovered that many pilots had a disorganised instrument scan which frequently left out vital displays - such as the airspeed indicator - for critically long periods. Such a fundamental failure in the exercise of a skill which, it has always been assumed, was basic to all licensed professional pilots was a shock, raising the question as to how widespread this failure is in the industry as a whole.

As a result, since 2009 Thomson has added eye-scanning technology to its training repertoire, as Capt Colin Budenberg, manager of training standards, explains: "We expect it to be used for the retraining of pilots who have been identified with performance issues."

It will also be useful in pilot selection and recruitment, he hints, adding: "I expect the most significant outcome will be to develop the skills of the pilot not flying - the pilot monitoring [PM]. Currently, we only know when someone isn't good at this [the PM role] when the pilot flying makes mistakes that are not spotted."

Airbus, NASA, the UK Civil Aviation Authority, French accident investigator BEA and Air France have also, in the past two years, been using eye-tracking to assess PM activity and produce a best-practice guide for the PM role.

De Courville has presented on the subject of go-around risks and flying technique at numerous aviation safety seminars in the past few years. He maintains a disciplined primary flight display (PFD) instrument scan is crucial to establishing a safe go-around trajectory - particularly in instrument meteorological conditions (IMC) or at night.

On a modern, single-screen PFD the technique for an effective instrument scan entails using the same traditional "T-shaped" eye-movement pattern associated with classic round-dial instruments, with the artificial horizon (AH) at the intersection of the T-shaped pattern. For go-around, the pilot flying's eye scan pattern centres on the AH where they initially select the go-around attitude, then radiates outward in turn to the air speed indicator, back to the AH, then right to the altimeter, back to the AH, then down to the compass strip for heading, and so on repetitively, with an occasional glance at the power setting to ensure it is what the pilot flying intended.

The pilot flying's scan must contain some of the same, but naturally has to range a little more widely to take in system performance, flightpath monitoring and any necessary intercepts of height or heading dictated by the missed approach procedure. There is a lot to take in, and if ATC has set a low level-out height, things happen quickly. This is the same whether the pilot is flying the aircraft manually or monitoring what the autopilot/autothrottle is delivering, with the additional complication in the latter case of a possible need to alter the flight-management system mode or its preset parameters.

On a night or IMC approach just before the go-around decision, if the pilot flying was beginning to divide his attention between instruments and emerging external visual cues, once go-around has begun there has to be a transfer of attention completely to the instruments, because when the nose rises to a go-around attitude, any surface lighting that was becoming visible just before go-around can suddenly become partially or completely obscured, robbing the crew of external visual cues or leaving them with only a few fatally misleading ones.

De Courville cites two unidentified - but non-fatal - events as examples of what can happen. The first involved a Boeing 757 crew in 2002. After initiating a go around in IMC, the pilot flying, when reaching the 2,500ft (760m) altitude intercept, applied and held a prolonged pitch-down input, resulting in a dive until the aircraft was in an extreme negative attitude (minus 40˚) from which recovery was made. The pilot reported: "When we suddenly got the altitude capture commands from our flight director, when both of us were in the mindset for a go-around, we became confused. And then, with the unbelievable nose-down pitch attitude, we became even more confused." Fortunately, they recognised the situation in time. The second event he cites involved an Airbus A330 in 2007. After initiating a go around at night over the sea, the altitude capture mode activated, the pilot flying pitched down to level off. The IAS increased towards VFE (flap exceedance speed) with the red strip becoming visible on the speed tape. Instead of maintaining a level flightpath at altitude capture, the pilot flying again maintained a prolonged pitch down input. Pitch attitude reached minus 9˚, vertical speed 4000ft/min (20m/s). The GPWS activated and the climb was resumed. The minimum altitude was 600ft over the sea, the total duration about 15s.

De Courville comments: "Degraded instrument scanning leaves an open door to somatogravic illusion and spatial disorientation. The effect of somatogravic illusion should not be considered the initial cause of this type of LOC [loss of control] incident/accident, but the robustness of the pilot eye-scan pattern in a dynamic phase is where the safety efforts should be placed."

He quotes two other training captains who have studied the causes of disorientation and loss of situational awareness: "Considering how critical an effective scan is, it is surprising that the development of a good set of scan patterns is not given high priority during training; especially since one of the most commonly cited forms of visual problems associated with mishaps is the breakdown in cockpit scan."

The US Federal Aviation Administration has this observation to make about the importance of the PM role in dynamic situations such as go-around: "Studies of crew performance, accident data, and pilots' own experiences all point to the vital role of the non-flying pilot as a monitor. Hence, the term 'pilot monitoring' is now widely viewed as a better term to describe that pilot."


The airspeed started to drop until the captain, who was pilot monitoring, noticed it passing 125kt (230km/h) and took control. By that time, the autopilot had trimmed the horizontal stabiliser to a nose-high setting commensurate with the low airspeed, so when the captain applied full power to go around, the aircraft pitched up dramatically to a 44˚ nose-high attitude despite full nose-down elevator input by the captain, and the airspeed bottomed at 82kt. Control was ultimately regained and the crew returned to land safely with no injuries.

  • 3 May 2006: An Armavia Airbus A320 was approaching runway 06 at Sochi, Russia at about 02:00 local time over the sea in limited visibility. Deteriorating weather caused approach control to instruct the crew to abandon the approach with about 3.5nm (6.5km) to go, which entailed carrying out a climbing turn to the right to avoid terrain. During the turn, the captain made nose-down inputs on the sidestick when nose-up would have been appropriate, and the aircraft hit the sea. The report does not mention the possibility of somatogravic illusion but does cite the time of night, possible fatigue and poor monitoring by the co-pilot, as factors. All eight crew and 105 passengers were killed.
  • 23 August 2000: A Gulf Air Airbus A320 was carrying out effectively a non-standard night-visual approach to runway 12 at Bahrain when the captain made the decision to abandon the approach because the aircraft was too fast and too high. ATC cleared the aircraft to climb to 2,500ft (760m) and turn left on to a downwind heading of 300˚ to reposition for approach. During the intended climbing turn the captain, the pilot flying, made nose-down inputs on the sidestick and the aircraft hit the sea with a nose-down attitude of 6.5˚ and an airspeed of 280kt. All eight crew and 135 passengers were killed. The report's verdict was that there were multiple causes starting with incomplete training and a poorly executed approach, but that somatogravic illusion was the reason for the pilot's nose-down stick inputs.
  • 16 February 1998: A China Airlines Airbus A300-600 was on approach to runway 05L at Taipei, Taiwan, when the crew abandoned the approach because the aircraft was too high. The crew applied go-around power but for a long time failed to counteract the nose-up pitch moment generated by the increase in power. Pitch-up reached 35˚ nose high. The aircraft stalled and did not recover. All 14 crew and 182 passengers were killed.