Wednesday, August 3, 2016

ADS-B Telemetry from EK521

Within minutes of the incident, published a time history of the ADS-B transponder transmissions from EK521.
REVISED with a closer look at the last 32 seconds and adding in position data.
REVISED with a summary plus a look at earlier arrivals wind effects
REVISED with the revelation that EK545, arriving ahead of EK521, had similar speed profile
REVISED with Boeing 777 procedures at end

With this information in hand, along with a short video showing the airplane sliding to a stop, apparently with gear up (or departed) and the right engine departed, the facts are being quickly assessed.
I offer some suggestions as to what might contribute to each observation, but there is hardly enough data to assess what happened. There is no way to assess a stall condition accurately without access to gear, flap, thrust and airspeed data; all of which would be terrific additions to ADS-B routine reporting - or possibly creating a limited ADS-B telemetry message set.

The point of this post is to explore the use of ADS-B data, and other public data.  While curiosity abounds, the official investigation has access to the full data set, to which I readily defer.

In summary

1) The approach was flown with progressively increasing ground speed.  As much as 25 knots was gained in the final minute. The flight immediately prior to EK521 suffered about a 23 knot ground speed increase in the approach.  With two aircraft showing about the same increase it would be more likely an increasing tailwind.  A tailwind landing does not inherently endanger a landing as long as sufficient braking distance and capacity and good tires are available.  The extra vertical speed necessary to track the approach path limits the ability for the airplane to slow down.

2) The landing procedure was done manually, as the autopilot would have a touchdown much sooner than what occurred.

3) Because of the modest altitude balloon coupled with constant loss of ground speed suggests the thrust was set to idle power from the onset of the flare.

4) For whatever reason, it does not appear a go-around was executed unless possibly in the final few seconds of flight.  Reports of the gear up (and visually not-present) might suggest it was raised along with flaps in the last few seconds of flight.  Engine thrust may have been commanded with flap retraction, but the engines may not have spooled up quickly enough to prevent a stall.  Flaps 20 should not be commanded below Vref30 speed.

5) The sharp descent leading to the initial contact at 29 seconds after the flare seems the result of a stall.   It appears to be uncontrolled. More than 50 knots of ground speed were lost from the roundout to initial impact, and notably still decreasing during the descent to impact.

6) Weather appears to be a factor.  It is apparent that the shear was increasing in magnitude. Without airspeed data to compare, it is not possible to be certain what contribution winds had to the accident.   The high temperatures worsened the recovery efforts and added additional ground speed (energy) to the situation.

Final Approach - Wind Shear?

Starting near the end of the approach, it is apparent the airplane ground speed began to increase while the airplane descended. It peaked at the flare point A, and then dropped over 50 knots as the airplane hovered over the runway.

An increasing ground speed could be indicative of an increasing tailwind/decreasing headwind.   Here is the general wind profile at Dubai, with the accident at about 12:40 local time.  Winds were increasing during the hour of the accident.

A very strong downburst has the potential to overcome even a Boeing 777, but there are no reports of convective activity.

Here is the METAR for Dubai, with the accident occurring during the 08:00 observation.

Weather observation for 08:00 decoded.  

Winds were 12 knots from 140 degrees, variable from 100 to 180 degrees, with a wind shear noted at 15 knots from 350 degrees.  The headwind component was about 11 knots steady, but if subjected to the wind shear would shift to about a 10 knot tailwind.  A shear from tailwind to headwind would cause ground speed to decrease and airspeed to increase.  A shear from headwind to tailwind would cause ground speed to increase and airspeed to decrease.

Here are two 777 arrivals about 20 minutes prior to the accident.  The first (EK621) shows a modest 5 knot "bump" in ground speed around 1000 feet above the runway.   The second (EK397) is a more significant 10 knot bump.   The 10 knots are gained in about 13 seconds, which is moderate.  But this magnitude is not significant and would be easily managed.  (Altitude is on the left scale, heading and ground speed on the right scale).

Here are the immediate two arrivals ahead of EK521.  EK409 saw a 10 knot ground speed increase.  EK545, arriving just ahead of EK521, saw a 23 knot ground speed increase, almost exactly as EK521 experienced.  In both cases, the ground speed increase was over about one minute, or less than 0.5 knot/sec, hardly a severe shear to control airspeed through. Notably, EK545 landed without incident.

EK545 shows a couple of kinks in its descent profile at about 1500 feet and again at about 800 feet.

I plotted EK521 descent profile against an idealized 3 degree glide slope.  For this plot, I calculated the distance from the flare point (Point A) and then the appropriate ideal altitude.   

There is a dip starting at about 30,000 feet out.  Otherwise the airplane seems to have followed the glide slope accurately.

An increasing tailwind will increase ground speed.  A hotter day increases ground speed.  Increasing ground speed requires increasing vertical speed to track a fixed, 3 degree glide slope approach. The approach vertical speed steadily increased as the airplane descended to the runway.  The increasing vertical speed can limit the ability to "slow down" the airplane.   

The vertical speed takes an excursion as predicted around 30,000 feet out, resulting in the dip below the glide slope.  There is no explanation for this behavior, but it appears to be an upset.  

What is striking is the vertical speed was decreasing from -1200 fpm to -600 fpm in the final 30 seconds of the approach.

For reference, the ground speed is plotted versus distance from Point A.   

Flare to Impact - Where is the Thrust?

At point A, EK521 leveled off, presumably the landing flare.

Runway 12L at Dubai is nearly 12,000 feet long.  Position A is where the airplane started to flare, Position Q the first contact and position T the end of the active reporting.  The distance is about 8,127 feet from position A to position T.

The airplane came to rest near the end of runway 12L (departure end of 30R).  

Point Q was the initial contact, point T the substantial contact.  The airplane slid the remainder of the way to the end of the runway.

Position Q, the initial impact point, is about 7,500 feet from the point A, and about 29 seconds later, at 08:37:39.   Based on ADS-B data, this is the first time the airplane touched down.

As the plane leveled off at point A, ground speed begins to bleed off.  Ground speed in on the left scale, distance from the flare point on the right scale.  More than 50 knots is lost from point A to point Q.

Vertical speed provides another factor into the event, showing the airplane traveled about 1,500 feet to level off, another 2,000 feet to begin to climb, and a steep descent at about 6,000 feet traveled.  It does not seem likely the pilot would want to continue to land well beyond mid-field, and thus the descent into the runway would appear to be uncontrolled. Vertical speed is on the left scale, distance from the flare point on the right scale.

Integrating vertical speed affords a confirmation of the altitude data.  The two variables match well except at the initiation of the climb and the final descent, which is to be expected given the integrating intervals. Vertical speed is on the left scale, distance from the flare point is on the right scale.

The airplane flared and floated down the runway into about a 100 foot balloon.  Ground speed leveled off and began to reduce rapidly, losing more than 50 knots. Vertical speed reversed into a steep descent and the airplane apparently crashed into the runway.

The Boeing 777 has ample engine thrust for a go-around.  The limited data-set would indicate the airplane did not have the thrust applied necessary to climb - a sustained positive rate of climb was not achieved and there was considerable loss of airspeed.

Upon reaching the flare point, the thrust would be retarded to flight idle. The airplane floated for over 20 seconds, which is adequate time for the engines to spool up to go-around power.  With speed bleeding off considerably, it appears insufficient thrust was added and instead the slight altitude gain was traded for speed.

The airplane could have flown into a stall condition (not revealed directly by the ground speed indications).  A stall condition is fundamentally too-low airspeed.   The Boeing 777 has significant engine thrust to avoid a stall, assuming it is applied in a timely manner.

A stall condition could be the result of mis-configuring the flaps.

Mis-configuring the spoilers can reduce climb performance.

An engine-failure would limit the climb trajectory, but is well within performance expectations to go-around safely on one-engine.

The flight recorder, the voice recorder, the meteorological data, the testimony of all involved, and a close examination of the wreckage will yield answers.

Boeing 777 Procedures

To aid in analysis, I have excerpted some information from another carrier 777 manual.  This information is just for information and should not be construed as directly applicable.


The following are the raw ADS-B data points captured.

Peter Lemme
peter @

Follow me on twitter: @Satcom_Guru

Copyright 2016 All Rights Reserved

Peter Lemme has been a leader in avionics engineering for 35 years.  He offers independent consulting services largely focused on avionics and L, Ku, and Ka band satellite communications to aircraft. Peter chairs the SAE-ITC AEEC Ku/Ka-band satcom subcommittee developing PP848, ARINC 791, and PP792 standards and characteristics. 

Peter was Boeing avionics supervisor for 767 and 747-400 data link recording, data link reporting, and satellite communications.  He was an FAA designated engineering representative (DER) for ACARS, satellite communications, DFDAU, DFDR, ACMS and printers.  Peter was also lead engineer for Thrust Management System (757, 767, 747-400), supervisor for satellite communications for 777, and manager of terminal-area projects.

No comments:

Post a Comment