On 11 March 2017, an Airbus 340-300 (F-GLZU) being operated by Air France on a scheduled international passenger flight from Bogota to Paris CDG as AF423 with an augmented crew and making a night VMC departure only became airborne from the 3,800 metre-long runway 13R just before its end despite using full thrust. The crew estimated being only 4 feet agl as the end of the runway was crossed and obstacles in the initial climb were overflown with little clearance. The remainder of the flight was without further event but on arrival at destination, a detailed safety report was submitted.
After receiving an Air Safety Report from the crew that the aircraft had only become airborne just before the end of the runway, Air France began a detailed internal investigation but were unable to find a cause and on 24 March, thirteen days after the occurrence, they reported it to the French Civil Aviation Accident Investigation Agency, the BEA. The BEA in turn informed the Colombian Grupo de Investigación de Accidentes e Incidentes de Aviación (GRIAA) and requested that the Investigation should be delegated to the BEA. This request was accepted by the GRIAA on 5 April 2017 and the BEA commenced its Investigation. By the time the BEA was notified, relevant data on both the FDR and CVR had been overwritten but Air France supplied a QAR data file for (only) the event flight on 28 March, four days after first reporting it and this allowed the Investigation to proceed once the GRIAA had accepted the request for delegation.
It was noted that the 47 year-old Training Captain in command of the flight had been PF for the departure and had a total of 14,050 hours flying experience including 815 hours on type and 1,608 hours on the Airbus A330 and had made just one previous rotation to Bogota. At the time of the investigated event, he was a current A340 Type Rating Instructor and was supervising the First Officer who was undergoing line training on type. He had joined Air France on qualifying as a pilot. The 50 year-old ex-military First Officer had a total 11,312 hours flying experience including 117 hours on type and 441 hours on the Airbus A330 and was on his first rotation to Bogota. The 34 year-old Relief First Officer, who was occupying one of the supernumerary seats on the flight deck for the takeoff, had a total of 6,494 flying hours including 1,021 hours on type and 1,035 hours on the Airbus A330. He had joined Air France on completion of pilot training in 2006.
It was noted that reference altitude above sea level of Bogota airport is 8,360 feet. It was established that the takeoff performance calculations had assumed a wet runway and had resulted in a V1 of 128 knots, a VR of 142 knots and a V2 of 149 knots with the performance limited MTOW being 237 tonnes for an actual takeoff weight of 236.3 tonnes. They noted that the takeoff performance was limited by the one engine inoperative takeoff run and that the margin over the ASDA was 57 m. The operator’s recommended practice in situations like this was to set 50 % TOGA thrust on brakes with the air conditioning packs fed from the APU. The aircraft was also lined up carefully at the runway threshold so as to maximise the distance ahead and was 45 metres from the runway 13R threshold at the point of brake release.
Takeoff clearance was given with a spot wind of 300° at 4 knots which corresponded to the wind velocity entered into the EFB used for performance calculation. One minute after brake release, the PF initiated the rotation at the calculated VR at which point the aircraft was 2,760 metres from the 13R threshold. Two seconds after initiation of rotation, sidestick deflection was recorded at 9° nose up (57% of the maximum). The nose-up input was then recorded as having “oscillated between 5° and 11° for the duration of the rotation". Eleven seconds after the beginning of rotation, the main landing gear was recorded as decompressed when the aeroplane was just 140 metres from the opposite runway threshold with the aircraft pitch attitude at 8.5°. One second later, the end of the runway was crossed at 6 feet agl at 155 KCAS - V2+6 knots after which the end of the clearway was crossed at 20 feet agl at 158 KCAS - V2+9 knots. The first obstacle on the climb out, the 20 feet high ILS LOC antenna, was cleared by 12 feet at 159 KCAS - V2+10 knots. At this point, the pitch attitude was 12.3° and the vertical speed “close to zero”. The elapsed time between the first nose-up input at VR and the aircraft reaching its target pitch attitude of 12.5° was 15 seconds and 35 feet agl (i.e. the TODA) was reached some 4,350 metres from the beginning of runway 13R. As the takeoff continued, the aircraft overflew obstacles in the second segment “with margins greater than the regulatory minimums”.
The crew reported that they had heard the ‘PITCH PITCH’ audio warning begin when the pitch attitude was at 10° and the aeroplane was still on the ground and had been “aware of the erosion of the takeoff margins” which it indicated. The relief pilot then asked the Captain if it was possible that the aircraft had struck the ILS antennae but on checking the ‘WHEEL’ page, no anomaly was displayed. The rest of the flight was completed without further event and an ASR was completed advising “the long takeoff and flight over the opposite threshold at an estimated radio altimeter height of 4 feet”. This performance was outside regulatory compliance and contrary to the crews expectations based on their pre flight takeoff performance calculation.
An annotated view of the takeoff. [Reproduced from the Official Report]
Establishing an explanation for the event
It was noted that the performance calculation model as correctly applied to the investigated takeoff gave an all engines operative takeoff run 2,989 metres which meant a lift off at 789 metres before the end of the runway compared to the 140 metres actually achieved. It also gave the TODA aircraft at 35 feet agl with all engines operating as 3,318 metres (as opposed to the 4,350 metres it took during the investigated takeoff). There was no evidence to indicate that the loading or airworthiness of the aircraft, the calculation and input of the takeoff performance or relevant environmental conditions other than those used in that calculation had any bearing on the abnormal takeoff.
Readout and analysis of the QAR data file, already carried out by Air France and Airbus, was carried out by the BEA. It was considered that differences in reconstruction of the takeoff phase by the three organisations “could be explained by both the choice of data to be read out and their sampling frequency” with different biases thereby introduced. It was accepted that this had made the accurate determination from the data of main wheel lift off, variations in rotation rate and tailstrike margin difficult and noted that such difficulties “had already been observed by the EASA OFDM Working Group". However, it was noted that in March 2014, the Group had produced a good practice guide which showed that relatively ‘simple’ indicators could be set up to detect unusual takeoffs in terms of takeoff distance or rotation time such as the distance between wheel liftoff and end of runway and the time between initial nose-up input and wheel liftoff.
The Investigation reviewed the type certification process as applied to the Airbus A340, noting the initial certification was conducted in accordance with JAR 25 which required that “the takeoffs performed to determine aircraft performance must not require the use of exceptional piloting techniques or vigilance”. It was noted that the equivalent EASA requirements in AC 25-7 state that their intent “is to establish takeoff performance representative of that which can reasonably be expected to be achieved in operational service” and (in summary) specify that:
- It must be possible for the procedures used during such flight tests to be consistently executed in service by crews of average skill.
- Such procedures must use methods or devices that are safe and reliable and include allowances for any time delays in their execution that may reasonably be expected to occur in service.
- Under no circumstances must there be any use of exceptional piloting techniques, such as use of higher control forces or higher pitch rates than would occur in operational service, to generate unrealistic takeoff distances.
It was noted that when asked about the existence of definitions for a conventional rotation technique and exceptional piloting technique, the EASA had responded by stating that “there (are) no specific criteria to define a conventional rotation technique in the regulations and that the JAR 25 paragraph quoted [...] means that the pilots must be able to repeat without difficulty, the procedures drawn up by the manufacturer”.
It was noted that in addition to Bogota, five of the other airports served by the Air France Airbus A340 fleet were “considered limiting in terms of takeoff performance” - Antananarivo, Montreal, Niamey and Saint-Martin. It was also noted from a study of Air France A340 pilot initial and recurrent training documentation that no “specific attention” was given to rotation technique and that “likewise, the rotation technique is not the subject of specific attention in flight proficiency checks”. Further, an examination of the training records of the crew involved “did not reveal any comments about the take-off rotation technique of these pilots during checks and training”.
Air France advised that “the take-off limitations (at Bogota) were not the subject of specific attention during the analysis of the route” but noted that pilots could access a video to familiarise themselves with the airport which advised that takeoffs there are performed “with limits” in respect of tailwind component, the maximum permitted EGT, the possible impact of tolerances accepted for the outbound leg from Paris, and the need to request a particular departure runway in advance to assure its availability. It was found that this video “makes no mention of the influence of the rotation rate on take-off performance or of the difficulties in arbitrating between the rotation rate and the risk of tailstrike”, the latter having been a particular focus of attention by Airbus in respect of the A330/A340.
It was noted that since 2007, no previous Air France ASRs had related to any concerns about takeoff performance or a long takeoff at Bogota but also, more generally, that prior to the investigated event the prevalence of long takeoffs was unknown since such events were not detected by the company OFDM programme, although this had been recommended by the EASA OFDM Working Group in 2014. However, it was stated that at the time of the investigated event, work had recently commenced to develop a long takeoff indicator event on “various Fleets operated by Air France” and this detected another very similar event which occurred on 6 April 2017. This event was reported to the BEA and is currently also the subject of a BEA Investigation.
Once available, the new OFDM event was retrospectively applied to initially all previous A340-300 Bogota takeoffs for which data was still available and then generalised to all A340-300 takeoffs. This showed that “a considerable proportion” of previous A340-300 takeoffs had recorded a rotation rate lower than that assumed in the (Airbus-generated) EFB takeoff performance software which used the AFM performance model. The illustration below was created from the data and shows that the average rotation rate (blue curve) reached during all these A340-300 takeoffs was less than 2° per second whereas the AFM performance calculation model used 1° per second for 1.7 seconds and then 3.1° per second (green curve) with the investigated takeoff (red curve) showing a much lower rotation rate than both of these.
The Air France A340-300 average rate of rotation (blue), the rate assumed by the AFM (green) and that achieved during the investigated takeoff (red). [Reproduced from the Official Report]
A comparison of average rotation rates achieved during Bogota takeoffs by Air France and Lufthansa showed that these were very similar - around 1.8° per second in both operators. In respect of the investigated takeoff, it was noted that an initial input of 2/3 sidestick pitch up then held “would not have permitted the takeoff performance expected by the AFM to be reached”. The Investigation found that Airbus had been aware of this difference between A330 sidestick deflections being made during takeoff in normal operations and those which would be needed to reach the certified performance of the aircraft and had “reiterated that [...] a continuous rotation rate of 3° per second is best adapted to takeoff performance while avoiding tailstrike risks”.
In respect of the Flight Crew Training Manual (FCTM) guidance to apply an initial sidestick pitch up input of two thirds of the maximum defection at VR, Airbus advised that this was “not representative of the inputs applied during the type certification process" and that “in the 1990s, the rotation technique during takeoff performance certification tests was not specifically formalised for aeroplanes under direct law on takeoff”. Airbus also stated that “it was not in a position to find the origin of these practices” and added that for more recent type certifications, “the rotation techniques used by the test pilots during the certification process had become more consistent with the FCTM operational recommendation”.
Airbus calculated from the available Air France data that the average pitch profile achieved in their A340-300 operations was likely to lead to an additional 200 metres being added to the takeoff ground run and to an additional 300 metres being needed to reach 35 feet agl.
The Cause of this Serious Incident was recorded as “inadequate nose-up inputs from the PF which, in the operational conditions of the day, increased the takeoff distance by 424 metres above the certified theoretical takeoff distance plus regulatory safety margins”.
It was further stated (in summary) that:
- the consequence of this increased takeoff distance was a significant increase in “the risk of a runway overrun or collision with obstacles”.
- the nose-up input from the PF during the rotation of the aircraft had been situated at the lower end of the range of values observed in several company’s A340-300 operations and was lower7 than the nose-up input recommended in the FCTM in force at the time of the event.
- In the conditions of the investigated Serious Incident, the application and continuation of an initial nose-up sidestick input of 2/3 of the possible maximum as recommended in the Airbus FCTM does not permit reaching the rotation rate of 3°/second mentioned in the same document and contained in the type certification performance model.
- Prior to the investigated event, the OFDM process at Air France did not detect long takeoffs and no notification of an insufficient takeoff performance had been made by European A340-300 crews at the time of the event.
Safety Action taken as a consequence of this event and its investigation was noted to have included but not been limited to the following:
- Air France introduced an artificial reduction of the runway length into the takeoff performance calculation process for Bogota. This was initially set at 200 metres but then increased to 380 metres. It was noted that during the process to determine an appropriate reduction in runway length, Air France had contacted Airbus, the French DGAC and EASA. Airbus responded by validating the calculations but “did not wish to give an opinion on the need for these operational measures based on the most critical case” and the DGAC and EASA responded that “the validation of the precautionary measures was not part of their prerogatives”.
- Air France set up specific pilot training on the rotation technique to be used on the A340. This had been completed by all pilots by October 2017. It was subsequently found that although fleet average takeoff distances had reduced, they were still greater than those assumed in the AFM performance model on which the EFB calculation of takeoff performance data relies.
- Air France has improved its OFDM programme so that it can now detect long takeoffs and their relative severity using around ten specific parameters derived from available QAR data.
- EASA has published a Safety Information Bulletin ‘Slow Rotation Take-off’.
- Airbus updated the A340-300 FCTM content on ‘rotational technique’ to replace the previous guidance on the initial sidestick input to be applied with guidance based solely on the continuous rotation rate and rotation time. However, no information was provided on changes to the risk of a tailstrike if operators concentrate on standardising initial sidestick inputs to reach the rotation rate of 3° per second.
Seven Safety Recommendations were made as a result of the Investigation as follows:
- that EASA in coordination with Airbus re-examine the validity of the initial certification hypotheses of the A340-300 take-off performance. [FRAN 2019-020]
- that EASA in coordination with Airbus take the necessary measures to re-establish consistency between the takeoff performance in operations and that established during certification on the Airbus A340-300. [FRAN 2019-021]
- that EASA in coordination with Airbus examine with the other primary certification authorities whether other CS-25 type aircraft are affected by this type of difference in performance and take the corrective measures that may be necessary. [FRAN 2019-022]
- that EASA in coordination with the national oversight authorities require operators of the A340-300 to set up safety measures to reduce the observed variability in the pilots’ rotation technique. [FRAN 2019-023]
- that EASA in coordination with the national oversight authorities require operators of the A340-300 to set up safety measures to restore sufficient take-off distance margins by comparing the possible difference between the take-off performance reached in operations and that established during certification. [FRAN 2019-024]
- that EASA in coordination with the national oversight authorities ensure that European operators introduce the indicators required to monitor take-off performance and at the very least, long take-offs, into their flight analysis programmes. [FRAN 2019-025]
- that EASA in coordination with the national oversight authorities collects and analyses the results of the monitoring referred to in Safety Recommendation 2019-025 in order to produce a report on the actual situation in operations. [FRAN 2018-026]
The Final Report was published in English translation on 22 July 2019 along with the initial and definitive publication in French.