Collaboration in a Systems Context
Just before midnight on 1 June 1999, an MD-82 with six crew and 149 passengers ran off the end of the runway at Little Rock, Arkansas (NTSB, 2001). Landing in a thunderstorm, the aircraft failed to stop, collided with structures on the airfield and slipped over an embankment. The captain and 10 passengers were killed. The flight had been delayed leaving Dallas/Fort Worth because of the bad weather. As they approached the airport at Little Rock, the pilots could see the flashes of lightning and their weather radar showed intense activity. The crew' had initially planned to attempt an instrument landing system (ILS) approach to Runway 22L. In the 17 minutes they were in contact with Little Rock АТС, that plan changed three times as the crew tried
FIGURE 7.1 Little Rock initial goal structure.
to respond to the rapidly changing conditions. As the weather front approached the airfield, the shift in wind direction rendered Runway 04 more favourable and so the crew switched to an ILS to 04R. In an attempt to beat the storm and get on the ground quickly, the crew briefly attempted a visual approach but reducing visibility forced them to reposition for the ILS to 04 instead. Early in this sequence the air traffic controller suggested a visual approach to Runway 22L, but the crew declined. The omission of checklist on final approach - an undetected lapse - contributed to the captain failing to arm the ground spoilers. From a systems perspective, the aircraft’s status on landing was not congruent with its required state.
Figure 7.1 represents the crew’s original plan as they approached the airport. The boundary of the system constructed on that day was the point at which sufficient energy had been dissipated for the aircraft to remain on the runway surface under the control of the pilots. It was not their goal simply to land, they had to stop, too. In order to achieve a safe landing, the crew had to manoeuvre the aircraft through a set of intermediate goal states, each with a set of constraints established by the performance characteristics of the aircraft, the design of the approach profile and the prevailing environmental conditions. The crew on board, together with the company Dispatchers and Little Rock АТС, worked as a collaborative team to accomplish this task.
Crew engage in planning to meet the expected applicable constraints, changing plans as events unfold and responding to transient disturbances or momentary perturbations. A disturbance might be the dynamic influence of the environment (turbulence or reduced visibility, for example) or an unexpected change in constraints (such as tailwind limits on landing or runway visual range (RVR) requirements). The environment the crew at Little Rock were operating in comprised the complex air traffic structures codified in terms of the approach procedures to the various runways in operation as well as different classes of airspace. The meteorological situation denied access to parts of the airspace over the airport and also limited what the crew could do in other parts of the available manoeuvring space. The crew were approaching their crew duty time limit and so the legal framework within which the flight was conducted represents a further limiting factor. An array of tangible and intangible factors shaped the freedom of action of the crew but, although these might have inhibited operations, the role of the operator is to shape the task in order to achieve the goal or switch to a different goal.
Collaboration as Shared Decision-Making
Level 2 in the hierarchical model is characterised by collaborative decision-making. Multiple actors agree solutions to achieving the task and then individuals act accordingly. In addition to the two crew members, various air traffic controllers provided information and support during the flight. Air traffic agencies also act to configure the legal airspace through which the aircraft can travel by approving flight plans, authorising amendments to plans and so on. АТС controllers and flight crew collaborated in real time during hazardous parts of the flight and, so can be considered a distributed crew. Other agencies operate at a remove. Dispatch, for example, had an involvement in the work process but of a different order to that of, say, АТС. A key differentiating factor between members of collaborative teams is the span of control: aircrew only have their own aircraft to deal with whereas off-board agencies are typically dealing with multiple aircraft.
The crew' of AA1420 were initially faced with two problems. The first was that their intended airframe was being used elsewhere. To wait for it to arrive would have meant that the crew' would not have sufficient duty time remaining to complete the flight to Little Rock. This was solved by deploying a different aircraft to the task. The second problem was the weather at their destination. By liaising with, initially, the Dispatcher, and in the final stages, w'ith the air traffic controller, a solution was apparently found to the problem of the weather. These elements of the story - airframe switch and weather - represent ‘threats’ or challenges and would have induced, at the least, a sense of frustration in the crew'. We also need to consider the time of day. It w'as close to midnight and, although the first officer (FO) said later that he did not feel that fatigue w'as necessarily an issue, the crew had been aw'ake for 15 hours before they departed on this flight. We saw in Chapter 4 that hours aw'ake before duty is a risk factor.
The original plan was to land on Runway 22L. As the aircraft descended, the crew were in visual contact w'ith the airfield and had radio contact with АТС. The pilots and the controller w'ere discussing the weather. The controller suggested that the aircraft’s weather radar gave a better picture than his. Between them, the three actors (the tw'o pilots and the controller) constructed a view of the weather based on actual measured meteorological parameters from the controller, the visual observations of the crew' and the w'eather radar images. Although the weather front w'as approaching the airfield, the pilots estimate that it was still some 15 miles away and the captain thought that they still had ‘some time’ in w'hich to make the approach.
This statement reflects the role of time in task management. Crew need to establish not just the ‘here and now'’ but also a version of ‘what will be in the future’. Human evolution has equipped us for dealing with change that progresses arithmetically (consider throwing a spear at an animal moving in a straight line. We learn to aim ahead of the animal if w'e w'ant to eat), but dynamically changing situations are a challenge. This example from a line operations safety audit (LOSA) report illustrates the point. The crew had correctly identified the threat posed by a storm cell on the runway extended centreline but reckoned that, because the departure track required a turn soon after take-off, they would keep clear of the weather. Unfortunately, they had not taken the prevailing wind into consideration and the aircraft met the storm cell soon after the turn.
At this stage the crew of A A1420 have a future goal of landing on Runway 22L (Figure 7.2), but we can establish at least two intermediate goal states: the first is to fly the ILS procedure to a point at w'hich landing can be undertaken safely and the second is to configure the aircraft for landing. Of course, these two goals are nested
FIGURE 7.2 Little Rock intermediate goals.
in that configuring for landing can be seen as part of the broader ILS goal. However, both need to be met if the aircraft is to be safely landed. All of this must be achieved before the weather front arrived over the airfield. Although the crew have estimated the approximate distance of the weather from the airfield, they do not appear to have resolved the relative rates of the closure of either the frontal system or their aircraft with the airfield. Furthermore, the severity and implications of the weather front have not been assessed. It seems that the crew were comfortable that the problem could be avoided by landing before the storm arrived. The crews’ construction of the world was approximate in that it included uncertainties that do not appear to have been recognised by the crew. Neither the current status nor the circumstances pertaining to their next desired goal state were clear and so the probability of satisfying the goal constraints was uncertain.
One such constraint was the crosswind on the runway and the crew now discuss the applicable limitations. Initially the FO stated that 30 knots was the limit, revising the figure down to 25 knots to allow for the fact that the runway was wet. The captain said that the correct figure was 20 knots and, when the FO took out his manual to check, the captain gestured to put the manual away as he was confident in his own knowledge. The captain was, indeed, correct but the episode shows how knowledge is not consistent across all team members.
The visibility was also now starting to reduce and the crew could only see the runway with some difficulty. The controller offered the crew a visual approach, presumably in an attempt to expedite the landing, but the crew declined on the grounds that the visibility was marginal and, therefore, a visual approach represented a riskier option than the ILS. Visual range is another goal constraint and the use of the ILS afforded more flexibility. The crew are now involved in trading off one option against another: an expedited, riskier arrival versus a less risky but more time-consuming one. Seconds later the crew received an updated wind report, which suggested that, at best, the crosswind limitation has been exceeded and, at worst, the aircraft would land with a tailwind (another constraint). There was no discussion of the wind situation other than the captain proposing a change to the reciprocal Runway 04R, which would now be, effectively, the into-wind runway. The change in wind direction had rendered the original plan untenable in that the wind constraints could not be satisfied. The change in goal state, rather than being the result of discussion, was based on an implicit understanding shared by both crews (a plausible conjecture, possibly, on my part).
At this point, the controller turned the aircraft away from the airfield in order to make enough space to reposition for the new approach. The FO briefed the captain of the main values associated with the new approach such as the ILS frequency, safety altitude, the minimum descent height and the missed approach procedure. However, the briefing was not complete. We have now shifted from the original intended goal to a new goal. The complete sequence of successive goals is illustrated in Figure 7.2.
During the turn away from the airfield, the crew were unable to monitor the position of the weather front as their radar was now pointing away from the storm cells and they could not see the airport environment. The manoeuvre illustrates the need to consider volumes of space and time associated with goal state accomplishment. As they turn back towards the airfield the captain was having difficulty seeing the runways because of the worsening weather. During the repositioning the weather system had, of course, moved closer to the airfield. The controller updated the crew on the surface wind conditions and the FO suggested that they attempt a visual approach. The captain agreed even though he could not, himself, clearly see the runway: ‘You just point me in the right direction and I’ll start slowing it down here. Give me flaps eleven’. The crew have now', seamlessly, changed to the third goal: a visual approach to Runway 04R. At the same time the captain w'as configuring the aircraft for landing as he called for flap and started to reduce speed. The controller, appreciating the fluctuating visibility associated w'ith the moving frontal system, said, ‘if you lose it (meaning visual contact with the runway), need some help, let me know please’. The three players in this worsening scenario appear to be working on a shared, implicit understanding of the situation.
The crew' then went through a short phase with the FO providing a running commentary to the captain about the location of the runway and the approximate position of the aircraft in a normal visual traffic pattern. For the next 60 seconds the FO acted as an on-board controller providing guidance to the captain until the crew recognised that, once again, the rules applicable to visual range were being violated. This phase show's how fragile formal control structures can be. Instead of formally briefing the new plan to fly a visual approach, the crew w'ere responding to events and creating ad hoc solutions in real time.
Unable to maintain visual contact, the crew were directed aw'ay from the airfield again in order to set up for the ILS approach to Runway 04R, the fourth new goal state. The additional time taken to position for the ILS meant that the storm front was now firmly established over the airfield. T hate droning around visual at night in weather wnthout having some clue w'here I am’ said the captain. ‘Yeah, but the longer we go out here the...’ replied the FO, presumably recognising that the time taken to position for the ILS was simply allowing the situation to worsen. ‘See how we are going right into this crap’, he said. Once the crew' finally intercept the inbound course they receive an update on the weather at the airfield. The controller told them that there was heavy rain, that the visibility was less than a mile and the RVR for 04R was 3000ft. In the next transmission, the controller updated the wind, reporting values right on the crosswind limits. The crew discuss the RVR limit and decided that they were safe to continue. Ordinarily, an approach in bad weather would be flow'n to a specific point and, if the landing was not possible, the aircraft would execute a go-around. The limitations and subsequent actions w'ould be briefed by the crew. On this day, no approach briefings w'ere given as the goals changed. The formal procedural framework unravelled, including checklist activity. Because changes were not anchored to specific points in the standard profile, the correct configuration of the aircraft became secondary to the goal of simply landing successfully. In very difficult conditions the captain managed to land. Unfortunately, because the spoilers w'ere not armed and given the weather conditions, the aircraft did not stop in the distance available. The boundary w'as breached.
In this sequence we can identify a number of discrete desired goal states. Each has a set of constraints in terms of permissible wind vectors and visibility. In addition, the transitions between goals consumed time and space which, in turn, resulted in dramatic shifts in the properties of the environment. The crew had to constantly revise their plan for overcoming the constraints placed upon their intended use of the aircraft. Of course, the crew did not have to attempt to land at Little Rock. Because of the weather, a diversion to an alternate or even a hold while the weather passed through was the option. The crew did have a duty time limitation that they needed to be aware of but these were exceptional circumstances. It is customary to describe a performance such as this as ‘target fixation’, ‘press-on-itis’ or, more formally, ‘plan continuation bias’. The inherent risk aversion in humans predisposes us to avoid losses and to not land at Little Rock represented just such a loss. In Chapter 4, we saw that psychological fatigue can influence risk perception. Late at night, after a long day and in worsening weather, after three previous attempts the crew appeared to be intent on landing. The cognitive load model suggests that switching tasks has a cost. The crew had repositioned the aircraft twice because constraints could not be met and so they were probably feeling the burden. We cannot know what was on the mind of the captain but there are sufficient precursors to explain the performance of the crew at an individual level.
Our main interest here is how the team functioned in its attempt to cope with the demands of the situation. What we see in the sequence is a progressive shift away from a structured team with roles and responsibilities to a situation where one pilot was trying to control the trajectory while the other pilot filled in gaps in their shared understanding. Having established that their initial plan violated crosswind limitations, the crew then lost sight of the fact that their final plan was equally ‘illegal’. The volume of information from the controller about different wind conditions around the airport probably overwhelmed the crew’s ability to comprehend, given the demands of the trajectory control task. As the plans changed, the structured approach briefing process broke down until, for the final two plan changes, no approach briefs were given. The formal checklist process was also degraded until, by the final approach, the checks were limited to calls for essential services: landing gear and flaps.