Flydubai crash shows why CRM is so important
Updated: Nov 30, 2019
The release of the final Report into the Flydubai 737 crash at Rostov airport on March 19th 2016 offers strong support for the guiding principles behind Crew Resource Management (CRM). This is worth examining in more detail.
The crash took place in the early hours of the morning, following two failed attempts to land that had to be aborted due to wind shear. After the first, the flight had remained in a holding pattern for two hours. Night time lack of visibility was certainly a factor, contributing to the disorientation experienced by the pilot-in-command (PIC), weather conditions also played a significant role in adding to his cognitive load and injecting a high level of stress. Fatigue was undoubtedly present too, in the wake of the crash the operator came under severe criticism for its flying schedules. It appears the captain had resigned from the company as a result of this and was serving out his notice at the time of the accident.
A full account of what happened is available here. In summary, it appears that the PIC became confused as to the position of his trim wheel as he attempted to level off following a steep climb as part of a go around to attempt another landing or divert. This led to a nosedive at full power and impact with the ground at some 600km an hour.
As no technical malfunctions took place, this is a good example of an accident where the human element played the decisive role. It is clear from the Report that the decision making capacity of the PIC was critically compromised during the final go around attempt. Unfortunately, it was precisely at this crucial moment that information flows between the pilot-in-command and first officer also broke down, meaning that the first officer was unable to act as a 'resource' in the way CRM is designed to achieve. This arose with the initiation of the go-around, where the PIC failed to state out loud whether this was a standard go-around or a Windshear Escape Manoeuvre (WEM). It seems that his intent was the latter but the F/O interpreted his actions as part of the former, further reinforced when the PIC accepted his suggestion to retract flaps to 15' and raise the landing gear. The PIC, in spite of this, still functioned under the assumption that the actions required on his part were as they had been in the first go around, a WEM. The two pilots were now out of step with one another.
At this stage, the position deteriorated into a "non-standard (non-trained) situation" (p. 149). It is in such situations that the conditions for 'High Consequence Decision Making' (HCD) (see the book, 'Shoot, Don't Shoot' for a full description, and our forthcoming training courses) apply with full force. One aspect of this kind of situation is that technology is of little use and is more likely to complicate things further rather than help, precisely because conditions are unanticipated by technical systems designers and software programmers. SOP's and checklists are often inappropriate here too, as it is judgment rather than a procedure that is most required. It is the point where not only does the human element dominate at an individual level, but also in the quality of interactions among crew members, and between teams, in this case air traffic control. It is also where organisational inputs such as the operator's policy on rostering, on diversions, on overnighting, on documentation and so on, all come into play, as they did on this occasion, with mixed results.
The first sign of trouble showed itself in the difficulty found by the PIC in maintaining a correct pitch angle during the climb out, finding himself forced to continually correct the aircraft due to its trim settings. Unlike the first go around, where his main problem was the nose pitching up too high, this time it stubbornly refused to lift. This effort absorbed his attention to the full, ('a tunnel effect' p. 152) causing him to miss the increase in airspeed. The first officer, however, was also caught up in this effort, both were taken by surprise when the maximum allowable speed for their flap setting (15') was reached. Ideally, according to the principles of CRM, the first officer would have stood back from the action and in doing so put himself in a better position to grasp where the problem lay, in the aircraft being out of trim (Situation Understanding in the HCD Framework). He was hampered in doing so by the lack of clarity over just what manoeuvre the pilot was seeking to carry out, a standard go around or one with a WEM. The PIC, by this time, was suffering from severe cognitive overload with the failure of his efforts to control the aircraft's pitch. In the words of the Report,
"Most likely, by this moment of time the flight mental mode at the PIC had been completely disrupted, whereas his emotional condition may be characterised as 'distress/a very strong stress'. It is confirmed by the fact that from 00:41:33 and up to the end of the flight quite dynamic pedal inputs are recorded, with that they had not been anyhow justified by the flight situation." (p. 154)
It is clearly the case that from this point on, it was only the first officer who could save the day. According to this report in Flight Global (see here), "The captain's psychological incapacitation and disorientation, says the inquiry, prevented his responding to prompts from the first officer – who, in turn, did not recognise the signs of the captain's deteriorating mental state in time to take decisive actions." This may well be unfair, as the final Report states,
"The F/O... realized that the hazardous situation was emerging. Desperately, with the increasing anxiety in voice, he tried “to return” the PIC to a control loop and rectify the situation. From the point of time of 00:41:34 the F/O prompted the appropriate actions to the PF (the control column pull) to prevent the situation transition to the accident. Nevertheless, most likely, the PIC had no longer heard the F/O and had not reacted to his words – he had fully lost the control of the situation, as well as the capability to control the aircraft." (p. 254)
The drama of the aircraft's final moments and the desperate attempts of the F/O to rectify the situation are captured in this transcript from the flight recorder,
The F/O went as far as to grab the control himself and attempt to pull the aircraft up. By now, it had entered a steep roll to the left and impact with the ground was inevitable.
If we look at this from a CRM perspective, then we find that the first officer DID have a good appreciation of both the aircraft's status and the lack of cognitive capacity on the part of the PIC to cope with the demands of the situation. On that basis he DID intervene forcefully, both verbally and even physically, BUT, his efforts were in vain.
The question then shifts as to WHY the F/O was unsuccessful in his interventions.
If we examine the issue from this angle, then the first thing we have to do is discount everything that happens from 00.41.33, as the position was already irretrievably lost by then. The F/O acted correctly, but it made no difference to the outcome. Instead, the critical period shifts to the moment the go around was initiated - 00.40.50 - until the point where the die had been cast, a time frame of 45 seconds. It was during this window that CRM failed to function as intended, that the first officer was unable to act as a proper 'resource' to make up for a captain sinking under pressure. From our account so far, we can describe this failure as -
poor information flow leading to a loss of a shared situational understanding over the go around and WEM
tunnel vision taking over for both pilots, compromising their situational awareness and ability to control the aircraft
An indicator of the deteriorating position inside the cockpit was provided when the F/O failed to report in to the radar controller after being instructed to do so by ATC. This omission was a clear sign that both pilots were overloaded and that their decision making ability was compromised as a result. A sharp observer would have been able to pick this up, especially as the F/O had displayed no previous difficulty in meeting his communication demands with ATC, even seconds earlier on approach. It is unrealistic to expect the air traffic controller to have done so, in fact ATC came under criticism in the preliminary report for their poor English skills which had come into play during earlier radio traffic when the PIC was considering whether to divert or attempt another landing. Nevertheless, we mention this because bringing a THIRD PARTY perspective into the discussion is IMMENSELY VALUABLE.
The benefit of shifting to the point of view of a third party, listening in on crew interactions, is that it injects a REFLECTIVE mode into our consideration of CRM/Non Technical Skills (NTS) in real time. The point is, ANYONE can do this, including the first and second parties involved, anyone can step back and take a moment to evaluate the quality of these interactions, and whether everything is proceeding as it should, or is going off track. The advantage this gives is that it allows warnings and indicators that things are not going as they should to be detected EARLY, while time remains to fix the problem.
From this perspective, the critical moment becomes the decision to go around, which the PIC did not verbalise, and the F/O did not query for clarification. This was a departure from what happened during the first go around, where the position was also made clearer by the wind shear warning that triggered the PIC's decision to abort the landing. In the second attempt, there was no warning, only a rapid increase in airspeed which the PIC interpreted, most likely correctly, as due to wind shear. Furthermore, during the long period between the two approaches, the pilots had made good use of the time available to talk through the options and prepare for the contingency of another go around. This informed the F/O's understanding that they were now engaged in a standard manoeuvre rather than a WEM.
During our CRM/NTS training sessions, we often 'stop the clock' at points such as this, and survey our training participants as to what they think the situation is, and more importantly, what they think both the PIC and F/O believe it to be, whether they are on the same page or not, ie have a shared situation understanding. This is obviously a luxury that often does not exist in real world situations due to time pressure, but then again it is precisely the advantage that scenario-based training of this nature has, and is why it is so useful. The idea is to develop the ability, hopefully to the extent that it becomes instinctive, to sense that the quality of team interactions has degraded and needs to be fixed, BEFORE its too late.
If either the PIC, or the F/O had acted to clarify the position in terms of what manoeuvre they were carrying out, there is a much greater likelihood the F/O could have detected the incorrect trim settings before the fatal nosedive. Likewise, if either had recognised the significance of the F/O's failure to report in to radar control, and had broken off from what they were doing to query this, then this would have severely reduced the 'tunnel vision' both were suffering at that point. We normally think of distractions as a source of error, but in cases such as this they can be a blessing, 'breaking the spell' and allowing for a mental reset.
CRM/NTS training from a third party perspective can help build such a capability. If we consider that good relations between crew members have by their very nature to be built in person, live and face to face, then it is also obvious that training can not be a substitute for such efforts, its has to serve a DIFFERENT function. During the flight in question, the transcript shows that the two pilots made the most of the time they had available to build a good rapport with one another, prepare for contingencies, and discuss through their options. The PIC also had a conversation by phone with a company representative, where agreement was reached over whether, when, and where, to divert. All of this helped, but unfortunately it was not enough. If training is to contribute something to a situation such as this, then it has to offer something else, something that can not be created in the cockpit, in the moment. This is what a third party perspective can do.
The Flydubai crash is instructive. It shows how the application of CRM principles would undoubtedly have been of benefit, possibly decisively so. It also indicates the contribution sound CRM/NTS training can make by developing a reflective capacity among aircrew, allowing them to evaluate in real time the quality of cockpit, and air to ground, interactions, in time to intervene effectively and avert disaster.