Again the world’s commercial air transport industry has smashed all records for keeping passengers safe – in 2017 there were no fatal accidents involving a passenger jet airliner, making for a perfect score in two of the last three years and bringing within reach the expectation of zero fatalities.

Meanwhile, out of sight beneath the low fatal accident statistics, growing quantities of “big data” from accident and incident reports around the globe reveal how frequently flights come perilously close to disaster, but since no one actually dies the event is less visible, especially to the news media. Sometimes these mishaps start with a technical problem, but more often they are the result of inadequate crew knowledge, poor procedural discipline or simple human carelessness.

Another growing development requiring industry review, passenger awareness and regulator oversight is associated more with changes in traveller lifestyle choices than actual danger. As the global population’s disposable income grows, so do the choices of exotic holidays or bespoke holiday add-ons that entail flights in small aircraft or helicopters. While these are generally safe excursions in absolute terms, the risk is statistically far greater than the almost risk-free travel that travellers enjoy today on mainline carriers. The 2017 accident list (see P??) records three fatal crashes involving chartered single-engined aircraft associated with holiday trips, two of which occurred on the last day of the year. As this industry sector grows, perhaps regulators need to review its performance.

International Air Transport Association (IATA) director-general Alexandre de Juniac has voiced his growing faith in the use of “big data” gathered in from the global industry to ensure continuing safety. "The aim is to build the greatest possible collection of data that will enable us to identify and eliminate potential issues before they arise," he says. When luck is the only factor that stands between a non-fatal accident and a catastrophe, clearly there is still safety work to do, which is what de Juniac is talking about. The accident list and the section of this review summarising recently published final accident and incident reports are testimony to how many near-disasters happen but are forgotten because, whatever trauma the passengers were put through, they survived.

Download the full accidents and incidents list

In 2017 there were 12 commercial air transport fatal accidents that resulted in a total of 56 deaths. That is a global figure that no other mass public transport mode could possibly match. The best figures previously were in 2015 when there were nine fatal accidents and 176 deaths. Just to illustrate the fact that things can temporarily get worse, in the following year – 2016 – the figures were respectively 13 and 306; but that was still a good year sustaining the longer-term trend.

The simple accident numbers for 2017 are small (see graph showing annual figures since 2008), but the long term trend in terms of accident rates bears witness to a steady but accelerating improvement (see accident rate graphs) in the number of flights per fatal accident, with the five-year moving average line for all aircraft categories showing a steeper improvement since 2011. A glance at the fatal accident rate for western-built jets tells the same story, with zero fatal accidents in 2015 and 2017. The rates for western-built turboprops are less good than for jets, but also improving, with a big leap in the figure for 2017.

West Wind ATR 42 crash

Amazingly, ATR 42 crash at Fond-du-lac, Canada, killed just one of the 25 people onboard

Transportation Safety Board of Canada

The main reason for the particularly low number of fatalities in 2017 is that no mainline passenger jets crashed. Also, the fatal passenger accidents that did occur involved small commuter types, and the remaining fatal accidents involved freight or other non-passenger operations. In fact one of the freighters was indeed a big jet – the MyCargo Airlines Boeing 747-400F that crashed at Bishkek, Kyrgyzstan, as a result of a tortuously badly managed descent and approach (see accidents listing P??).

That badly managed descent ended in tragedy, but in the section summarising the findings of recently published final reports from previous years (P??) there are several examples of poor aircraft management that nevertheless avoided actual disaster while coming very close to it. On 20 July 2014 an EasyJet Switzerland crew mismanaged the autopilot flight modes during the descent and were slow to respond as the aircraft’s speed began to nudge the top end of the flight envelope. The crew finally resorted to manual flying to regain proper control but, in doing so, injured the cabin crew. The report comments on the “lack of diligence” in monitoring flight parameters.

On 15 May 2015 in Australia a Skytraders Airbus A319 also flew a mismanaged descent but got away with it. There are several other examples of poor descent and approach management in the same section, the worst involving a Shaheen Air Boeing 737-400 approaching Lahore, Pakistan in November 2015. Almost every mistake that could be made during the descent was made, but in the end the six crew and 142 passengers survived a landing in which both main undercarriage units were sheared off.

As a scan of the accident list for 2017 (starts on P??) will quickly confirm, the most common kind of airline accident is still a runway excursion during landing. It is usually not fatal, but almost always results in damage to the aircraft – sometimes serious damage. This has always been a weakness, but until the last decade it was masked as a potentially preventable accident category by the fact that it usually was not fatal, and by a focus on the accident categories that posed a greater risk to life at a time when accident rates were higher.

The Flight Safety Foundation (FSF) began carrying out studies in 2007 about how the risk of excursions and overruns might be reduced, and the main proposal was for airlines to brief their crews to ensure that they achieved a “stable approach” by a certain “gateway”: for example passing 500ft on final approach. The basic criteria for a stable approach were these: at the gateway height the aircraft should be configured for landing; at or very close to the correct airspeed; and close to the correct glideslope and extended runway centreline. Weather conditions should have been confirmed to be within minima.

The hope was that the discipline imposed by having this defined objective would encourage pilots to work hard to achieve stabilised approaches early, rather than allowing the pilot the discretion to decide to continue an unstable approach if a safe landing was deemed feasible. The FSF advice was that, on the rare occasion in this slightly idealised new world that the approach was still unstable at the gateway, the crew should abandon the approach and go around.

It has since become apparent from further FSF studies that runway excursions on landing remain commonplace, and that the guidelines for go-arounds are largely ignored. This is partly because go-arounds themselves – previously considered an uncomplicated manoeuvre – were found to involve potential hazards that needed to be balanced against the risk of continuing an unstabilised approach to land. This created confusion about what best to do.

So in March 2017 the FSF published updated guidance in its go-around decision-making and execution project, which considers how best to achieve this balance between two sets of risks. It proposes that getting this decision right represents the best single chance that commercial air transport has to reduce accidents further (see below). While not abandoning the intention to establish a stable approach by 500ft, the most tangible recommendation of the new report is that the actual decision to land or go-around can be taken at 300ft above runway level.

Although 2017 did not include any fatal accidents resulting from loss of control in flight (LOC-I), it remains the biggest killer accident category of the past decade. For that reason the European Aviation Safety Agency has published initial changes to simulation requirements to improve pilot training for stall and in-flight upset scenarios. These changes are part of a rulemaking task designed to update flight simulation training device (FSTD) capabilities and specifications over the next two years.

The first package, WP1, is intended to exploit FSTD technological advances, and support authorities and training organisations by providing a "competencies framework" to guide inspectors. EASA adds that the package will support approach-to-stall training as well as upset and recovery requirements, and increase the fidelity of simulated airframe and engine icing effects. It also aims to approve full-flight simulators for training in the post-stall flight envelope, which had not previously been replicated by FSTD manufacturers because it is so unpredictable in the real aircraft, being subject to so many potential atmospheric variables.

EASA says 19 accidents during commercial air transport operations, over the five years from 2012 to 2016, were classified as LOC-I – of which 17 were fatal. Two of these fatal accidents involved companies with an EASA air operator's certificate. "Analysis of accidents and serious incidents shows that, in many cases, flight crew are caught by surprise in the event of an upset, or have limitations and difficulties in detecting the upset and the approach to stall," EASA says, adding: "In certain cases, the flight crew does not realise that the [aircraft] is in an actual stall." The agency says an analysis of 58 serious events over the 2012-2016 period indicated that in 10 cases – including four fatal accidents – training the crew in a simulator that has these advanced capabilities would have improved their ability to recognise and handle the approach to a stall, as well as recover from a full stall.

Meanwhile, with all those near-catastrophes in 2017 and the years before it that did not actually kill any passengers, it seems apparent that luck – an unfashionable concept these days – is playing an uncomfortable part in making the airlines’ safety performance look better than it actually is.

Go-arounds remain critical safety challenge

In March 2017 the Flight Safety Foundation’s Runway Safety team published the latest version of its report “Reducing the risk of runway excursions”. Runway excursions continue to be the most common accidents involving airlines today.

The report states confidently that the implementation of its recommended procedures for flight discipline during final approach is the key to eliminating – or at least reducing – this most common accident type. But although airlines have been advised for more than 10 years to require their crews to abandon an approach if it is not stabilised – and to go around instead – crews have continued to ignore this standard operating procedure (SOP). The result has been a continued high rate of runway excursions during landing.

The FSF explains: “The problem of go-around policy noncompliance is real and is arguably the largest threat to flight safety today. The potential impact of improvement in compliance is significant. No other single decision can have such an impact in the reduction of aviation accidents as the decision to go around.”

The foundation’s headline change in March to its earlier guidelines is to declare that the 500ft go-around decision gate should be reduced to 300ft above airfield level, but it adds a qualification: “It should be understood that the 300ft AGL value is not intended to be absolute; it can be approximated to take advantage of aircraft automatic callout systems. For example, consider an ILS minimum set for 200ft AGL. Some manufacturer automatic callout systems provide an alert 80ft above minimums, so in such cases, 280ft AGL could be established as the go-around gate value and utilised in the auto callout in the active call procedures.”

In providing this qualification to its advice, the FSF shows that it is fully aware of the complexities of decision-making for pilots at this most intense of all flight phases. Indeed it spells this out: “Analyses indicate that flight crews who continue an unstable descent below 300ft do not recognize the need for increased concern – or the need for a go-around.” So 300ft – or close to it – has been nominated as a kind of psychological tipping point beyond which the pilot should know that risks will increase – either the risk of continuing an unstable approach or of delaying a go-around decision further.

If the report has a fault it is that it tries to describe every single consideration, and there are hundreds. So while an airline operations policy team would do well to read the whole screed, they must still ensure that their SOPs are simple and clear.

The FSF identifies where the industry needs to start: “The first and foremost change required is that the industry must improve its awareness of the problem; to achieve this, a shift in focus and cultural norms is required. It is believed that significant improvement is attainable; however, the cultural shift will be much easier if the industry shifts collectively, as opposed to individual companies making changes on their own.”

A part of the preparation for adopting the go-around philosophy is for airlines and crews to know what risks are associated with an all-engines-operating go-around. The most dramatic recent example of the risks, especially at night or in IMC, is the 2016 FlyDubai Boeing 737-800 crash at Rostov-on-Don, Russia. A full-power go-around creates rapid linear acceleration, and the acceleration can create somatogravic illusion in the pilots. The latter is a powerful signal from the balance organs that the aircraft has pitched up dramatically even when it has not, and the pilot reaction can be – as at Rostov – to push the nose down.

When aircraft had a lower power/weight ratio, go-arounds were gentler. With today’s jets, plentiful power means things happen fast – so pilots need to be ready for it. That means there is a need for recurrent training in all-engines go-arounds.

The FSF has found that “go-arounds occur at an average rate of one to three per 1,000 approaches, but there is a large variation of go-around rates among different aircraft operators and operational environments.” On average, the foundation observes, a short-haul pilot will conduct one or two in a year, a long-haul pilot one every two or three years.

Finally, the FSF says: “A just culture must prevail if problems in go-around safety are to be sufficiently understood and addressed.” Also, airlines must ensure their pilots do not perceive that they are under pressure to make a first-time landing come what may.

Download the full accidents and incidents list