Monday, January 23, 2017

TK295, URSS Sochi Runway 02, Sep 27 2016 TC-JLP A319 ATC Intervention in Go-Around

There is a difference in the reported actions (turn left) and the data (turn right) after commencing a particular missed approach at Sochi in Sep, 2016.  The FR24 data offers a climbing right turn, whereas avherald depicts a left turn direct.  

The lengthy discussion reflected in this blog is a tool to understand the relevance and benefit of ADS-B data as a tool for safety investigation. This is not a safety investigation of TK295, there was no accident, no injury reported, no damage; and an official report does not appear forthcoming.

Please take note that other explanations or factors not known or misunderstood herein may correctly determine different conclusions - in other words, the conclusions herein may be wrong.

These observations stem from sparse ADS-B reports plus other references not known to be accurate combined with best-intentioned judgements and helpful advice from online commentators.

(25 Jan)
NOTE: A mistake in my spreadsheet (calculating the timer) rippled into this discussion.  The conclusions remain unchanged; the supporting data is better aligned. Sorry for any confusion!
(25 Jan)
Added takeoff plots to validate methodology.  Appears takeoff weight was around 60,000 kg; landing weight about 57,000 kg.
(27 Jan)
Much cleanup on the charts, corrected a few minor errors. 
(30 Jan)
Added a chart to compare both go-around and the takeoff profile together.  Required building out the avherald flight path with waypoints, speeds, altitudes, times.

A THY Turkish Airlines Airbus A319-100, registration TC-JLP performing flight TK-295 (dep Sep 27th) from Istanbul (Turkey) to Sochi (Russia), was on final approach to Sochi's runway 02 at about 00:54L (21:54Z Sep 27th) in low visbility when the crew went around from low height (below 400 feet MSL) due to windshear. 
The aircraft turned left in conflict with the missed approach procedure requiring a turn to the right and was climbing towards terrain when ATC intervened instructing the aircraft to immediately climb to FL060 which was read back and complied with by the crew, then tower instructed to immediately turn right, the crew read back that instructions but continued to climb straight ahead towards terrain. 
ATC therefore instructed the aircraft to climb to FL150, which took the aircraft above terrain and into safety. The aircraft was subsequently vectored for another ILS approach to runway 06 and landed safely on runway 06 about 30 minutes after the go-around.
The highlighted section is the part that is in question.

I could find no evidence of windshear in the FR24 approach data.  The airplane ground speed increased after the airplane had begun to climb and turn, presumably accelerating and raising flaps and gear.   I have to assume a significant windshear if connecting the avherald waypoints directly, a headwind to drive down the ground speed.   The magnitude of the assumed headwind exceeds any wind measurement I have.


URS 272200Z 02006MPS 2300 R02/3000D -SHRA BKN018 OVC040CB 12/11 Q1021 R02/220348 R06/220348 NOSIG RMK R06/05003MPS MT OBSC QFE764 

URSS 272130Z 04005MPS 9999 -SHRA BKN050CB OVC083 13/11 Q1021 R02/290050 R06/290050 NOSIG RMK R06/35002MPS QFE764

Location: URSS
Day of month: 27
Time: 21:30 UTC (00:30 Local)
Wind: True direction = 040 degrees, Speed: 5 m/s
Visibility: 10 km or more
Weather: Light Showers of Rain
Clouds: Broken sky , at 5000 feet above aerodrome level, cumulonimbus
Clouds: Overcast sky , at 8300 feet above aerodrome level
Temperature: 13 degrees Celsius
Dewpoint: 11 degrees Celsius
QNH (Sea-level pressure): 1021 hPa 30.15” Hg
No significant changes expected in the near future

Altitude correction is +230’ to pressure altitude. The minimum altitude based on the ADS-B data is 530 feet MSL.

Runway 02 Missed Approach

Missed approach instruction is to climb ahead to 200m before turning right.  

avherald reports that TK295 initially turned left in the missed approach.

FR24 data shows TK-295 started to turn right early, at about 160m.  This put TK-295 over a small hill in the right turn.

From the ADS-B reports, the airplane continued to turn right in a 360 climbing turn, and headed north towards the mountains, rather than proceeding as the procedure calls along the 210 radial.

The ATC command to climb to 6,000 feet appears to align with a pitch up maneuver, perhaps the flight crew were anxious to climb based on the urgency of the ATC command.

ILS02 procedure

ILS02 Missed Approach

TK295 Missed Approach - Turn left? published the following flight path. flight path

The avherald flight path matches FR24 except for what happened initially after the missed approach was initiated. flight path flight segment in question
avherald connects these two points.  They are 2 minutes, about 6400 feet in altitude and about 3 nm apart.

The same flight path is shown on

For the avherald flightpath: Speed at the beginning waypoint was 183 KTAS and it finished at 226 KTAS based on data.

For two minutes, at 180 KTAS = 6 nm.  
At 240 KTAS = 8 nm. 
For an average speed, 7 nm.

Distance traveled would include the altitude gained.  A right triangle is formed by distance translated on the x-axis, altitude gained on the y-axis, and the distance traveled along the hypotinuse.  Solving for distance translated with 7 nm traveled and 6,400 feet gained is 6.91 nm.

Normal climb performance (6400 feet) would cover about 9 nm.

FR24 path is about 8.3 nm.

avherald shows 3 nm translated, 3.1 nm traveled to include the 6,400 foot altitude gain.

There was a reported 10 KTAS wind at the runway 02 surface from 040 deg.  That equates to a 9 KTAS headwind.

TK295 reported windshear in the missed approach.  

The data shows a steady ground speed of about 118 KTAS, add 9 KTAS headwind equals 127 KTAS.  This closely matches the landing speed for 57,000 kg. airplane.  

Stall speed for 57,000 kg, sea level, full flaps/gear down is about 99 KCAS.  1.23 times 99 = 122 KCAS.  Add 5 KCAS for coupled approach = 127 KCAS = 127 KTAS at sea level, standard day

If the approach was flown at 1.23 Vs1g, then Vs1g = 103 KTAS (KCAS)...or about 61,000 kg.

If there was a windshear, with constant airspeed, then the ground speed carries the full extent of a wind shear.  In this case, there was no notable change to ground speed until the go-around commenced.

TK295 climbed and returned at about 11,000 feet.  Plotting the data in that flight segment offers an estimate of winds about 25 KTAS from 270 deg.  


For the two minutes and over 6,000 feet of altitude gain, if flown direct, would suggest a ground speed average of 90 KTAS.  This is without regard to the speeds of 183 and 226 noted in the flight data.

Disregarding the flighradar24 speed data, did TK295 climb over 6,000 at an airspeed of less than 120 KCAS?  

With winds of 25 KTAS from the West, the highest airspeed would be about 115 KTAS (or 107 KCAS at 5,000 feet).

Setting aside any estimate of winds, I created a direct route showing a left-turn then climbing direct to the 21:56:15 waypoint, and then proceeding as both AvHerald and show.

The waypoints were spaced equally along the AvHerald route.  The time between each point and the speed between each point was adjusted to minimize the error between the line and Air distance.  The altitude gain was adjusted to correlate with the low speed event.  The heading was adjusted to match the track (for roll angle calculations). 

The following table shows the inserted data (highlighted in yellow) to match the AvHerald direct flight path.,

Taking data from later in this discussion, the AvHerald and Flightradar24 climb out data are shown in the following graph.  

If TK295 did not fly directly between these two points as avherald depicted, what other route did it take to make up the lost distance?

The FR24 data shows a climbing right 360 turn.  

the climbing 360 turn on TK295 go-around. The airplane gained about 1000m in the climb.

Flightaware did not have any useful data for the period of interest.

For this analysis, a timer is started at 21:53:56 marking the beginning of the go-around missed approach procedure.  This is about 24 seconds earlier than the waypoint position that matches the avherald data point. starting data point is at 21:54:15, 19 seconds into the go-around.  

the point that TK295 began its go-around: FR24 shows a right turn. tracks to the 21:54:15 UTC point.

TK295 climbing turn as shown by flight

TK295 climbing right turn shown with missing flight segments

Radius of initial climbing turn

Radius at completion of climbing turn

Radius of Curvature in the climb out.

TK295 climbed in a right turn, presumably raising gear and flaps and accelerating initially to 250 KCAS.

Distance traveled is not necessarily computed by connecting the ADS-B reports, as they granularity overlooks a more circuitous routing.

It appears TK295 initially started with a radius of curvature of about 0.86 nm, and finished the circuit at about 1.04 nm radius.

Radius of a turn in level flight is based on bank angle and speed.  For level flight, it is easily calculated using the following formula:

Using the turn rate equation offers a level roll angle solution for unaccelerated flight (stable climb or level flight). 

I have adjusted the baro altitude up by 230 feet to account for the high pressure.  

Note that there are some periods where the time interval is extended (lost reports).  These areas are highlighted in yellow under the time column.

Flightradar24 and google earth plots connect these points without regard (line distance).  

Using integrated speed (air distance) affords a better estimate of the distance traveled than a straight line measurement. 

Air distance, being based on ground speed, includes wind effects.  

Three occurrences where air distance significantly increased more than line distance are highlighted in yellow in the Air and Line distance columns.  

Note, I have corrected Air distance to equate to Line distance (translated distance) by subtracting the contribution from any altitude gained.

There is speed excursion in the climb from 290 to 204 and back to a proper 250 (noting this is ground speed).  This is highlighted in the Gnd Spd column.

Air Distance/Time column is a speed estimate from the go-around point. This column shows the collective "average speed", simply total distance traveled by time traveled. 

A turn radius of 0.86 nm was observed graphically in the initial (accelerating) part of the climb.  

As speed increased towards the end of the right 360 to 250 KTAS, the turn radius was observed graphically to be about 1.04 nm.  

By using heading rate, roll angle can be calculated. As mentioned, the roll angle calculated is worst case, for climbing flight it would be less than shown.

It can be seen that a climbing turn starting at less than 30 deg roll, but increasing to about a 45 degree bank (peak 54 degree) to complete the turn.  

The airplane was beyond 30 deg roll for about 40 seconds. This is highlighted in yellow in the roll angle column.

The distance traveled in the modeled route, to the completion of the climbing turn (wings level)  is about 7.1 nm, over a gain of about 4450 feet in about 106 seconds (or about 2500 fpm).  

Note that during the completion of the turn, the airplane was pitched up excessively and was trading speed for altitude gain (descelarating).

A guess for a gross landing weight is 58,000 kg.

Candidate Weight estimate

Unauthorized A319 FCOM excerpt

Brake release-climb performance (book)
Time to climb to   1,500 feet is about    90 seconds       3 nm
Time to climb to   5,000 feet is about  150 seconds       7 nm
Time to clime to 10,000 feet is about   255 seconds     13 nm

Brake release-climb performance (TK295 takeoff)
Time to climb to   1,500 feet is about   85 seconds       2.8 nm
Time to climb to   5,000 feet is about  190 seconds      9.2 nm
Time to clime to 10,000 feet is about   271 seconds     15.1 nm

Brake release times include about 30-45 seconds extra in comparison to TK295 missed approach, due to takeoff ground roll from brake release.

To compare TK295 takeoff numbers, I subtracted 36 seconds and 0.7 nm so that they both start at about the same energy position.

climb performance (TK295 takeoff)
Time to climb to   1,500 feet is about   49 seconds       2.1 nm
Time to climb to   5,000 feet is about  154 seconds      8.5 nm
Time to clime to 10,000 feet is about   235 seconds     14.4 nm

climb performance (TK295 missed approach)
Time to climb   1,500 feet is about    60 seconds     3.5 nm (Air)      2.7 nm (graphic)
Time to climb   5,000 feet is about   112 seconds    7.5 nm (Air)      8 nm    (graphic)
Time to climb 10,000 feet is about   209 seconds  13.9 nm (Air)    15 nm    (graphic)

The graphical distances are within reasonable tolerance of the Air distances for such rudimentary analysis.

The climb rates of 2500-3000 fpm (average) are to be expected in a light-weight climb.  

The bank angle of 45 degrees would degrade the climb performance.  The use of go-around power would offer additional performance compared to normal climb power.

Plotting the data.  A timer is started at the point of the go-around, where altitude begins to climb.

The airplane accelerates (clean up flaps and gear) for the first 30 seconds (presumed by procedure and observed speed increase).

Comparing the Air and Line distance revealed three points where the Air distance reflected a longer path than the Line distance.  Each increasing circumstance is aligned when there is an extended gap in data points.

The Line distance connects these points with a straight line, but the path may have been curved.  That appears to be the case with the first two cases, circled in yellow.

The last occurance, at about 133 seconds, was during the speed upset.  In this case, the average speed may reflect higher than what was encountered (suggesting the underspeed upset may have been more severe than pictured).

The climb was not conducted in a stable manner.

Between 2,000 and 7,000 feet, the airplane appears to have gone through a bit of a roller-coaster.

At around 100 seconds the airplane is pitched up sharply and vertical speed spikes.

Initially the airplane nosed over and sped up to about 290 KTAS, and then it appears to have been nosed up considerably, causing the airplane to slow to 204 KTAS, from where it appears to nose down again and capture a stable 250 KTAS climb.

At about 60 seconds, the airplane pitch appears to level and roll angle increase to a peak of about 54 degrees.  This appears to be a moment where the airplane was maneuvered outside of its normal climb envelope (overspeed).

Headings plotted greater than 360 are shown for visual benefit.
450 is 90 deg (E)
540 is 180 deg (S)

The time to climb and altitude gained are plotted against distance traveled.

The climb profile through the first 5 nm is flatter, presumably due to acceleration and turning.

The airplane climbs steeply between 5-9 nm, where speed appears to be traded for altitude.

The climb seems stable past 10 nm.

Takeoff Comparison

TK295 takeoff data was put into the same spreadsheet as used to evaluate the missed approach.

Takeoff roll is over 45 seconds.

Vertical speed is limited in the takeoff clean-up and acceleration.

Heading change in the climb out - note the overshoot.

Estimating Takeoff Gross Weight

Normal takeoff power climb is at V2+10 KCAS up until thrust reduction altitude.

The initial TK295 climb speed is about 154-155 KTAS (KCAS).  

SA 27/09/2016 20:50-> METAR COR LTBA 272050Z 02005KT 360V060 9999 FEW035 SCT080 18/10 Q1025 NOSIG=

SA 27/09/2016 20:20-> METAR LTBA 272020Z 01006KT 340V050 CAVOK 18/10 Q1025 NOSIG=

The takeoff was about 20:37Z.  Winds were 010 - 020 5-6 kt.  Climb track was 359.  Headwind was 5-6 kts.

That puts V2 at 148-150 KCAS.

IST LTBA runway 36 3000m standard day, below sea level pressure alt., 
2% downward slope = +56m
6 kt headwind = +66m
Runway effective length is 3122m

A predicted gross weight of around 60,000 kg on a 3122m runway at below sea level pressure altitude  could be accommodated at the greatest thrust derate in configuration 1+F.  There would be no need to go to configuration 2.  

In the case of the greater derate, configuration 1+F, allowed to a gross weight of about 65,000 kg,
V2 148-150 KCAS.

The V2 speed does not confirm gross weight, rather it would suggest it was less than 65,000 kg.  If configuration 2 was selected, gross weight would be less than 76,000 kg.

Using the following climb charts from the FCOM to estimate takeoff gross weight.

In my original publication, I overlooked about an 8 second discrepancy just after liftoff.  The difference between Air and Line jumped about 0.3 nm in that one step.  At that moment, I had hand-calculated the time difference between each waypoint, and I had screwed up putting in 24 seconds for what should have been 17 seconds.  Thus, the checks worked!  I have updated the charts and tables, and adjusted any quoted numbers that were changed.  Note that about one-half of the offset came in one long update interval of 23 seconds between 46-69 seconds after brake release. 

The difference in Air-Line distance is due to integrating a one KTAS bias in ground speed.  I reduced the ground speed by 1 KTAS in the following chart to show nearly zero long-term error.  This may be due to granularity in the ground speed (one KTAS), or a small calibration error.

Time stamps from brake release, liftoff through 20,000 feet.

Plugging in the observed performance to reverse-lookup takeoff gross weight.

1500 feet:    80 sec     2.5 nm     (66-66 kkg)
5000 feet:   190 sec    9 nm         (72 kkg)
10000 feet: 265 sec  15 nm          (66 kkg)
12000 feet: 301 sec  18 nm          (58 kkg)
14000 feet: 367 sec   23 nm        (60 kkg)
16000 feet: 415 sec   28 nm       (60-62 kkg)
18000 feet: 466 sec   33 nm       (62-64 kkg)
20000 feet: 500 sec   36 nm       (56-60 kkg)

from 10,000 to 20,000 took 4 min (60-62 kkg) and 21 nm (50-54 kkg).

It would appear that the takeoff weight was approximately 60,000 kg.

Estimating fuel consumed

The stage length was about 500 nm, with cruise at 37,000 feet.

For a takeoff weight of 60 kkg, the climb should take 21 minutes and travel 132 nm.  This would leave 370 nm to the destination.   In this case it took about 23 minutes and traveled 158 nm, leaving 350 nm to destination.

It would have burned 1566 kg fuel in the climb.

For initial takeoff weight of 60,000 kg, without wind, and a distance of 350 nm would take 1497 kg and about 56 minutes.

Total would be 3063 kg and about 1:20 minutes.  Actual time was 1:18 minutes.

Landing weight would be about 57,000 kg (modeled at 58,000 kg).

Comparing Takeoff to Go-Around

The following charts show the takeoff and two go-arounds for comparison:

The missed approach gained less altitude than the takeoff traveling over the first 6 nm.

The missed approach was flown at a faster speed for the first 7 nm than during the takeoff.

The missed approach was flown at a lighter weight and would have used G/A power for the beginning part of the procedure.  

The extra energy went into accelerating the airplane to a higher speed, and to offset the loss of lift due to the high-bank turn.

The missed approach was flown at a slower speed from 7-14 nm than during the takeoff.

The missed approach and the takeoff arrived at the same speed and altitude at about 17 nm.  

The missed approach had extra drag due to the turn during the initial climb out that included a segment between 40-54 deg roll angle. 

It appears the extra performance was traded off by the extra drag in the turn (missed approach compared to takeoff).

Accuracy of Reported Position

Navigation based purely on Inertial Reference System is subject to drift.  
Looking at two arrivals (Istanbul, Soshi) TC-JLP ADS-B reported position appears to be based on GNSS, with high precision.

TC-JLP on TK1201 arrival IST on 27 Sep, earlier in the day

TC-JLP, TK295 arrival URSS (Sochi) after the incident

Stay tuned!

Peter Lemme
peter @

Follow me on twitter: @Satcom_Guru

Copyright 2017     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 lead engineer for Thrust Management System (757, 767, 747-400), also supervisor for satellite communications for 777, and was manager of terminal-area projects (GLS, MLS, enhanced vision).

An instrument-rated private pilot, single engine land and sea, Peter has enjoyed perspectives from both operating and designing airplanes.  Hundreds of hours of flight test analysis and thousands of hours in simulators have given him an appreciation for the many aspects that drive aviation; whether tandem complexity, policy, human, or technical; and the difficulties and challenges to achieving success.


  1. "The distance traveled in the modeled route, to complete a 360, is about 8 nm, over a gain of about 3200 feet in about 80 seconds (or about 2500 fpm)."

    This means about 360 kt, right.....?

    1. Hi Heinz,
      Revised Jan27-Upon close examination of the better charts. It is amazing how much more incite can be gained seeing the data plotted. I had judged the turn complete based on "roll" angle", but should have paid more attention to track heading.

      Firstly, note that I had a mistake in my spreadsheet causing the timer to be calculated wrong, and this rippled into the discussion and charts. I have revised ***again*** the post to correct ***errors***, and to expand on the discussions. The climbing ***320 degree*** turn was completed in ***84*** seconds over a distance translated of about ***5.4*** nm, an average ground speed of ***212*** KTAS.

  2. Hi Peter,

    sorry to disagree again. I have personally flown various Airbus types, so the A319. IMHO an average ground speed during a 360 deg go around manouver of 241 kt including a clean up from F FULL to F UP isn't realistic.

    Kind regards

    1. Hi Heinz,

      We agreed the first time I was wrong - which I corrected.

      I understand your skepticism. It was relayed to me quite vigorously by I am just trying to figure out where the airplane went. The FR24 data represents a departure from expected operation, including what apparently was a very steep turn, and what appears to be a point where the airplane is maneuvered very aggressively.

      In comparison to the preceding takeoff, 3,000 kg heavier and at normal climb power, the takeoff climbed to 3000 feet at 226 KTAS in 107 seconds; and to 10,000 feet in another 127 seconds.

      The missed approach got to 3,000 feet in about 93 seconds and at 287 KTAS, and while turning; and another 110 seconds to arrive at 10,000 feet. The missed approach flight path was noticeably "flatter" than the takeoff, initially falling back more than 2 nm in the initial climb out, but catching up by 4,000 feet in the climb.

      I am appealing to anyone that has access to a credible A319 simulator to try the maneuver as provided and report back the results. I do not have any data to suggest how long G/A power was maintained.

      If this is not the flight profile, then (1) how did FR24 record this data and (2) where did TK295 go then?

      FR24 data is faithfully captured as received, and a local timestamp applied. The source of the data is the airplane transponder; the timestamp in the ground receiver. TK295 position data has been shown to be accurate. How did FR24 record data that does not come from TK295, that does not represent its flight path; yet other than this missed approach seems to match perfectly?

      If not the FR24 climbing turn, avherald suggests the airplane never accelerated and climbed the 6,000 feet at Vref, flaps/gear down. This is the only explanation for taking two minutes to travel about 3 nm (and into a larger headwind than I can find evidence of).

      I don't have any skin in this game. I have faithfully plotted out the FR24 data to understand the projected flight path. The data is continuous. I hope a simulation will offer some substantiation, one way or the other.

      Thank you for your interest and comments. I hope you will remain open-minded.