Swiftair MD83 over Mali on Jul 24th 2014, aircraft lost altitude

Last Update: April 22, 2016 / 14:50:38 GMT/Zulu time

Bookmark this article
Incident Facts

Date of incident
Jul 24, 2014



Aircraft Registration

ICAO Type Designator

Mali's ANAC released their final report via the French BEA concluding the probable causes of the crash were:

The aeroplane speed, piloted by the autothrottle, decreased due to the obstruction of the pressure sensors located on the engine nose cones, probably caused by ice crystals. The autopilot then gradually increased the angle of attack to maintain altitude until the aeroplane stalled. The stall was not recovered. The aeroplane retained a pitch-down attitude and left bank angle down to the ground, while the control surfaces remained mainly deflected pitch up and in the direction of a bank to the right. The aeroplane hit the ground at high speed.

The accident was the result of a combination of the following events:

- the non-activation of the engine anti-icing systems

- the obstruction of the Pt2 pressure sensors, probably by ice crystals, generating erroneous EPR values that caused the autothrottle to limit the thrust produced by the engines to a level below that required to maintain the aeroplane at FL310.

- the crew's late reaction to the decrease in speed and to the erroneous EPR values, possibly linked to the work load associated with avoiding the convective zone and communication difficulties with air traffic control.

- the crew's lack of reaction to the appearance of buffet, the stickshaker and the stall warning.

- the lack of appropriate inputs on the flight controls to recover from a stall situation.

These events could be explained by a combination of the following factors:

- The FCOM procedure relating to the activation of the anti-icing systems that was not adapted to Pt2 pressure sensor obstruction by ice crystals

- Insufficient information for operators on the consequences of a blockage of the Pt2 pressure sensor by icing

- The stickshaker and the stall warning triggering logic that led these devices to be triggered belatedly in relation to the aeroplane stall in cruise;

- the autopilot logic that enables it to continue to give pitch-up commands beyond the stall angle, thereby aggravating the stall situation and increasing the crew's difficulties in recovery.

The absence of a usable CVR recording limited the possibility of analysing the crew’s behaviour during the flight.

Specifically, it was not possible to study CRM aspects or to evaluate the possible contribution of the employment context and the experience of the crew members.

The ANAC stated in their findings:


- When the aeroplane speed reached 203 kt, vibrations attributed to buffet appeared.

- Four seconds later, when the aeroplane’s speed was 200 kt, the stickshaker activated, followed three seconds later by the stall warning. At that time, the aeroplane reached the 12 degrees stall angle of attack.

- Both engines suffered a surge due to the aeroplane’s high angle of attack.

- The autopilot disengaged about 22 seconds after the triggering of the stall warning. The aeroplane angle of attack had then reached approximately 25°. There was no apparent crew action between the stall warning activation and the autopilot disconnection.

- During the aeroplane’s fall, its pitch attitude and bank changed significantly. The aeroplane continued to pitch down with a left bank angle down to the ground. During this flight phase, the control surfaces remained mainly deflected pitch up and in the direction of a bank to the right. When the elevators were commanded close to the neutral position, around 27 seconds before the impact, the stickshaker and stall warning system intermittently deactivated and the engines resumed normal operations.

- No problems were mentioned by the crew during their contacts with the Ouagadougou and Niamey air traffic controllers.

- No distress message was received by the control centres.

- The last recorded values were a 58 degrees pitch-down attitude, a 10 degrees bank to the left and a calibrated airspeed of 384 kt.


- The crew had not received any training relating to approach to stall and recovery from it since joining Swiftair. The investigation could not determine how far back their last training on these points took place.

- During simulator training, the approach to stall is performed manually by the crew after disengagement of the autopilot.

- The stickshaker and stall warning initiation speeds on the airline’s simulator are values lower than those calculated for the accident flight.

- The FCOM procedures relating to engine or airframe icing do not mention the characteristics of icing by ice crystals, and are thus not adapted to the specifics of this type of icing.

- The phenomenon of obstruction of the Pt2 pressure sensors due to icing is only described in the FCOM in the climb phase

- The stall warning devices triggered at high altitude with a speed margin below the 7% mentioned in the certification regulations. These specific features were not brought to the attention of operators.

- The autopilot remained engaged beyond the triggering of the stall warning devices and the stall. This specific feature, and its consequences on the detection of and recovery from a stall, is not explicitly detailed in the manufacturer’s FOB, the only documentation supplied to operators on this point.

The ANAC analysed that the crew initially had planned to climb to FL330, however, changed their mind prior to departure and requested FL310, which was the maximum flight level possible with all anti-ice systems activated.

The ANAC analysed: "Thirteen minutes after take-off, as they climbed and passed FL215, the crew performed the first course alteration to the left to avoid a storm area located on their route and reported this to the Ouagadougou ACC to which they had just been transferred. The engine anti-icing system had not been activated by the crew at that time. The total temperature was then greater than the one below which the anti-icing systems should be activated according to the FCOM. Even if the FCOM provides that they should be activated when icing conditions are expected, it is possible that the crew may have postponed the actual operation. In the absence of any usable CVR recording, the investigation was not able to determine if subsequent non-activation of anti-icing was due to an oversight or to a decision by the crew considering that the environmental situation did not require it. There were no subsequent signs of activation of the anti-icing systems during the flight. Thereafter, the changes in the aeroplane’s heading made it pass along the western edge of the convective system in an area where ice crystals were likely present. The presence of supercooled water was however less likely, and as a result it is unlikely that the airframe was affected by icing. The probable absence of signs of icing on the airframe, in particular on the windshield wipers, the possible lack of clear signs of ice crystals, which may be difficult to visually detect, especially at night, and which are usually not detectable on the weather radar, and the absence of significant turbulence probably did not encourage the crew to activate the engine anti-icing systems."

The ANAC analysed with respect to the sequence of events leading to the loss of altitude:

At 1 h 37 min 28, the aeroplane levelled off at FL 310 at Mach 0.740. The autopilot then held the altitude and heading of the aeroplane, while its speed was controlled by the autothrottle. At the same time, the aeroplane was transferred to the Niamey ACC.

In this phase of flight, priority tasks of the crew would be to manage the flightpath to avoid weather hazards and to establish contact with ATC control centre while performing monitoring of primary flight parameters.

Within two minutes of levelling off, the speed of the aeroplane increased and the crew selected the cruise thrust regime on the TRP. The EPR values of the right engine became incorrect, probably due to the obstruction of the Pt2 pressure sensor of this engine by ice crystals. The autothrottle then adjusted the thrust to prevent the erroneous values from exceeding the EPR LIMIT in cruise setting. The thrust delivered by the engines was then lower than the thrust required for level flight, and the speed of the aeroplane continued to decrease.

For about one minute, the gap between the EPR values of the left and right engines gradually increased and then stabilized between 0.2 and 0.3 and the autothrottle switched to MACH ATL mode three times.

Fifty-five seconds after that of the right engine, the left engine’s EPR also became erroneous and started to increase. Five seconds later, and for four seconds, this increase was interrupted by a decrease in both engines’ RPM. This decrease could have resulted from the crew reducing the Mach target, or from a manual decrease in engine RPM by over-riding the autothrottle. That may therefore correspond to the crew becoming aware of an anomaly.

However, the information indicating the occurrence of an anomaly at that time was the gap between the left and right EPR values, and the intermittent appearance of the MACH ATL mode on the FMA, the speed of the aeroplane still being close to a nominal value in cruise. At that moment, these non-prominent elements,, whose interpretation is not straightforward, may not have led the crew to suspect a problem of insufficient thrust due to blockage of the Pt2 pressure sensors on both engines.

The engine RPM then increased again until the erroneous left EPR values reached EPR LIMIT. The thrust delivered by the engines remained lower than the thrust required in this phase of flight and the aeroplane continued to decelerate. About one minute later, that’s to say two minutes after the appearance of the measurement error on the first engine, the aeroplane passed behind the power curve in thrust-drag ratio. The gap between the thrust required in level flight and the thrust actually delivered by the engines accentuated and the decrease in speed became more marked.

For about four minutes, the autothrottle was in MACH ATL mode, but the gap between the left and right EPR had decreased and their values were close to those expected in cruise. The N1 values were slightly lower than the typical cruise values (77% instead of 80 to 85%). The inconsistency between the EPR values and N1 values was therefore hardly noticeable to the crew, more so since the documentation they had did not have a table of correspondence between EPR and N1 and they had not been trained to monitor the correct correspondence between these two parameters. In addition, the crew were still avoiding the convective weather and in addition were busy trying to establish contact with the Niamey ACC. In fact, during this time, the crew tried to contact ATC eight times. Of these eight occasions, only two messages were received by the Niamey control centre.

During this period the aeroplane speed decreased from 278 kt to 213 kt. The Fast/Slow indicator on the PFD probably reached the lower stop and the attitude of the aeroplane increased by 4°, accompanied by aural warnings indicating movements of the THS. Although these elements were more perceptible and are not encountered during cruise flight at normal speed, they do not seem to have resulted in any reaction by the crew.

The representation of speed on the Mach-airspeed indicator means that up to 250kt, which corresponds to the first two minutes of deceleration, the loss of aeroplane speed caused very small movement of the needle. Even if the Mach number is indicated by three digits on the instrument, the general image of the dial corresponded to that expected by the crew in this phase of the flight.

It is likely that the crew workload involved in managing the circumvention of the convective weather and various attempts at establishing contact with the Niamey area control centre contributed to the lack of timely response to the decrease in speed, despite the visual and auditory information aimed at warning the crew.

The review of previous events showed that other MD80 crews had suffered significant losses of speed at altitude without detecting them.

The autopilot in altitude hold mode remained engaged and compensated for the decrease in altitude due to the speed decay by increasing the nose-up position of the trimmable horizontal stabilizer.


When the speed reached 210 kt, the position of the needle on the Mach-airspeed indicator was close to vertical. At that moment a quick reading of the instrument could have made it possible to detect inadequate speed. A prompt reaction consisting of putting the aeroplane in descent and increasing thrust to regain speed would be expected from a crew in this case.

Two variations in engine RPM caused by crew inputs on the throttle levers were observed. It is therefore possible that the crew at this moment suspected an EPR problem. In fact, these actions are consistent with the items in the "EPR erratic or fixed" procedure which recommends moving the throttle levers and observing the indications of the engine parameters. However this action alone does not correspond to the reaction expected of a crew in an approach to stall.

The autothrottle was disengaged between the first and second variations in the RPM. The speed was then 203 kt, the conditions of the "SPD LOW" display were met and it might have been displayed on the two FMAs. In this situation, this indication is expected to trigger a quick check of the flight parameters, the attitude and speed in particular.

The investigation could not explain the lack of reaction to the « SPD LOW » display. However, when this indication appeared, the crew was handling the radio messages with Niamey ATC control centre and making inputs on the throttles. The AP was still engaged.

The buffet phenomenon likely also appeared at the same time. Without usable CVR recordings, and taking into account available data, it was not possible to determine if the crew noticed and identified it, or whether they assimilated it as turbulence associated with avoiding the convective system. When the speed reached 200 kt, the stickshaker triggered, followed three seconds later by the triggering of the stall warning.

From this time onwards, the captain’s side loudspeaker only broadcast the “STALL” warning, while that on the co-pilot’s side alternated the “STALL” warning with the other warnings that were active (altitude and THS movement, where applicable).

The triggering of the STALL warning would have called for the following corrective actions as provided for in the airline’s procedures in relation to stall:
- disconnection of the autopilot and autothrottle

- application of maximum thrust
- decrease in the angle of attack until the buffet stops.

The crew did not disconnect the autopilot and did not execute the stall recovery procedure.

Presumably they did not identify this critical situation.

In order to maintain altitude, the autopilot then commanded a continuous nose-up movement of the trimmable horizontal stabilizer and the elevators. This resulted in an increase in the angle of attack of up to 24°, or 13° above the stall angle of attack in the event conditions, as well as the broadcast of several “STABILIZER MOTION” warnings.

Both engines suffered a surge probably due to the aeroplane’s high angle of attack. Both engines rpm then decreased to values close to idle. This surge may have been noticed by the crew.

There was no sign of a reaction by the crew, other than the throttles movements, until the disconnection of the autopilot51 which occurred 25 seconds after the triggering of the stickshaker. The speed was then 162 kt, the altitude had decreased by about 1,150 ft. The aeroplane was banking to the left and its pitch was decreasing. The crew applied input mainly to roll to the right to bring the wings level. At the same time, they applied mainly noseup inputs, contrary to the inputs required to recover the stall and continued to do so until the ground.

In the current case, the environmental conditions, over the desert at night, deprived the crew of visual references to help them recover from this situation.

The ANAC analysed with respect to anti ice protection activation:

The FCOM procedures for protecting the engines and airframe against icing ("Airfoil Ice Protection Operation" and "Engine Anti-Ice on Ground and in Flight") indicate that icing conditions can exist when the total temperature (TAT) is less than 6°C and visible moisture is present or if ice build-up occurs on the windshield wipers or edges of the windshield.

The latter description, as well as performance considerations, may well encourage some flight crews to consider that the criterion of ice accretion on the windshield wipers or edges of the windshield can be used instead of the visible moisture criterion. They can thus be led to assess the risk of icing reactively.

Reports from two major airlines with wide MD80 experience show that some crews did not activate anti-icing systems in visible moisture conditions, in the absence of visible signs of icing. Other reports indicate that Pt2 icing may occur outside of visible moisture.

The wording “when visible moisture is present” can be ambiguous and subject to different interpretations, such as of the presence of the aircraft in clouds or not, the visual detection of clouds in the vicinity or the detection of echoes on the weather radar.

In the case of icing by ice crystals, experience shows that ice crystals do not adhere to external parts of the aeroplane, that they are difficult to detect with current on-board weather radars and not necessarily visible to the naked eye especially at night or at low concentrations. Feedback from some crews that experienced this type of in-flight icing is in line with this. The criterion of ice accretion on the windshield wipers is thus inadequate for an ice crystals environment. The visible moisture criterion, if it is interpreted as meaning “flight inside clouds” can also be inadequate for this situation.

In addition, the possibility of icing of the Pt2 pressure sensors is only described in the FCOM in the case of the EPR LIM engaged mode and only describes the consequences of Pt2 icing in such cases in the climb phase.

These procedures also state that the warmer the air mass, the higher its water content and the more severe the icing conditions. They state that below -20°C, icing should be less severe. In the present case, the temperatures were below - 30°C, under which the water in the atmosphere is essentially present in the form of ice crystals, without the water content necessarily being very high. The procedure does inform flight crews that heavy icing has on occasion been reported at temperatures as low as - 60°C, but does not inform of the risk of Pt2 icing due to ice crystals, which may not be clearly visible at low concentrations but may nevertheless alter Pt2 measurement when anti icing systems are not activated.

These are indications that the FCOM procedures for the protection of the engines and airframe against icing were developed from classic icing phenomena (the formation ice resulting from the impact of supercooled water droplets), the only type of icing taken into consideration in the context of the certification of the aeroplane. These procedures are not adapted to icing by ice crystals to which Pt2 pressure sensors are sensitive.

As a result, under the current FCOM procedures, crews, although aware of the classic risks of icing, may not be aware of the warning signs (or lack thereof) associated with icing by ice crystals, and may ignore the possibility of the potential obstruction of the Pt2 pressure sensors and the associated consequences in cruise.

With respect to the approach to stall the ANAC analysed:

During the initial phase of falling speed, the autopilot has no protection to prevent the speed from dropping below a speed that would guarantee adequate margin in relation to a stall. Correcting this initial speed decrease thus depends on attentive monitoring of this parameter by the crew and taking into account the visual and aural indications provided: air speed indicators, Fast/Slow indicator on the PFD, MACH ATL display on the FMA and PHR aural warning.

The study of similar approach to stall events on MD80 type aeroplanes shows that these safety barriers can be crossed. In fact several of these crews were not aware of the degraded situation of the flight before the triggering of the stall warnings.

On MD80 type aeroplanes, if the speed continues to decrease, the buffet and stick-shaker alert the crew to the approach to stall and represent the last safety barriers before the stall, the stall warning being designed as an additional device to assist in stall recognition.

Data from flight tests at high altitude show that the buffet and stick-shaker appear with small margins in relation to the stall, lower than the 7% criteria called for in the certification requirements. No explanation has been found on the rationale for accepting this lower margin.

Recovering this situation thus depends on a prompt and appropriate reaction by the crew to their activation, in particular with the disengagement of the autopilot and the auto-throttle.

Without any disengagement by the autopilot, the latter tries to keep its target. If it is in altitude hold mode, as is the case in cruise, the autopilot gives nose-up orders that are contrary to those required to recover the stall. After the triggering of the stall warnings such orders, by maintaining the aeroplane's pitch and by increasing the angle of attack and the nose up position of the horizontal stabilizer, rapidly make the situation worse, as well as the difficulty of recovery for the crew. In fact, these nose-up orders mask or delay the occurrence of a nose down movement which for the crew would constitute the ultimate indication of entering the stall, and make it necessary for the crew to apply a stronger and sustained nose-down input to recover from the stall. The FOB published by Boeing indicates that the aeroplane can slow down until the stall warning before the autopilot disengages. It does not however state that the autopilot may continue to give nose-up orders after the stall warning and does not indicate the consequences of this behaviour in terms of detection of and recovery from the stall.

The crew was not trained to immediately take back manual control and perform a recovery procedure, including under the startle effect. On the contrary, the operator’s pilots were trained for this type of exercise by causing an approach to stall and a stall by disengagement of the autopilot, with reduction in thrust and manual input on the THS control. Furthermore, the specific qualities of the MD80 in relation to the late triggering of the stall warnings in cruise and the absence of automatic disengagement by the autopilot are not sufficiently brought to the operators' attention and are not integrated in the crew training programmes.

However, in cruise, the following factors might explain the absence of immediate manual autopilot disengagement as a reaction to the triggering of the stall warnings:

- in case of startle effect, the desire to understand the situation before undertaking the recovery procedure, considering that the high altitude of the aeroplane and the engagement of the autopilot would ensure some safety;

- hesitating to take back an unusual and little practised manual procedure in this phase of flight.

The engaged autopilot may have contributed to the crew's lack of reaction to the triggering of the stall warnings during the accident.

Training is crucial in relation to stalls because a flight crew, during their career, is rarely exposed to an approach to stall, even less so to a stall during an actual flight. The date of the crew's last training on approach to stall and recovery there from could not be determined. It was in any case prior to their arrival at Swiftair and was thus not recent.

With respect to information dissemination the ANAC analysed:

The accident to flight AH5017 shows that the dissemination of information published by the manufacturer, the civil aviation authorities and the investigating authorities did not result in sufficient assimilation by operators and crews of the specific features of the MD-80 in case of icing of the PT2 sensor by ice crystals and an approach to stall at high altitude. The fact that the operator had no criteria in its flight analysis at the time of the event for detecting the decrease in cruise speed at high altitude confirms this shortcoming.

The risk associated with PT2 sensor icing was known to the manufacturer and the oversight authority. Certain operators knew of this risk and were collecting icing-related cruise speed reduction events. One of them had taken the initiative to build a table of correspondence between the EPR and N1 values that is easily accessible in the cockpit.
Incident Facts

Date of incident
Jul 24, 2014



Aircraft Registration

ICAO Type Designator

This article is published under license from Avherald.com. © of text by Avherald.com.
Article source

You can read 4 more free articles without a subscription.

Subscribe now and continue reading without any limits!

Are you a subscriber?

Read unlimited articles and receive our daily update briefing. Gain better insights into what is happening in commercial aviation safety.

Free newsletter

Want to know more and stay ahead? Get our free weekly newsletter and join 4845 existing subscribers.

By subscribing, you accept our terms and conditions and confirm that you've read our privacy policy.

Send tip

Support AeroInside by sending a small tip amount.

Related articles

Newest articles

Subscribe today

Are you researching aviation incidents? Get access to AeroInside Insights, unlimited read access and receive the daily newsletter.

Pick your plan and subscribe


Blockaviation logo

A new way to document and demonstrate airworthiness compliance and aircraft value. Find out more.

Virtual Speech logo

Train yourself online in VR with the special course for aviation: "Crisis Communications: Airlines". Find out more.

Get updates

Never miss an article from AeroInside. Subscribe to our free weekly newsletter and join 4845 existing subscribers.

By subscribing, you accept our terms and conditions and that you've read our privacy policy.

AeroInside Blog
Popular aircraft
Airbus A320
Boeing 737-800
Boeing 737-800 MAX
Popular airlines
American Airlines
Air Canada
British Airways