act700, I understand and agree with you up to a point. However, GA's are not all that frequent. So one may ponder whether said Captain had, in those 800 hours on the B744, ever experienced a GA and (again specifically on the B744) what sort of training for such an event had the airline given him?

If we take in account FR24 height data (including last recovered points) it seems that they aim in the point ~900m right after the further end of the rwy, which is marked 0 on the graph

https://cdn.aviaforum.ru/images/2017...e3a2866c34.png

A shift in altitude and position is normal in FR24. A horizontal shift by 4 km would be a lot, but not unheared of.
All this data tells us, is that the approach was stable until the very end of the transmitted data.

Do we know by now whether this has been an ILS approach or a GPS approach?
All reports say go-around, at least one engine is missing all its fan blades, looks like a crash at full power (of at least one engine, to be exact). Post impact fire shows plenty of fuel was left. The rest is gueswork from here. Maybe locals know more.

FR24 shows quite reliable horizontal position of route and last known point in this case.

https://blog.flightradar24.com/blog/...ata-TK6491.jpg

(Edit: Kulverstukas beat me to it).

The rwy has only an ILS ( cat II) or a non-precision VOR approach. No Rnav approaches.
Besides, it were LVP's.

An ILS is an ILS. It guides you towards the touchdown point on a given runway.
Not 4km. Behind it.

An LNAV/VNAV approach could guide you to the wrong spot obviously ( map shift, although very uncommon on B744) but I can not even start to imagine why the crew would be doing this.
I would not trust FR24 data.

It would be extremely strange for ADS-B data to be shifted precisely in the runway direction, precisely along the extended centerline, and precisely ending at the actual crash location.

Absent other data, for now it seems more probable that the FR24 data is correct and the aircraft was not on the CAT II ILS DME 26 glideslope.

I also read in local forums that besides Bishkek airport was renovated, VOR/DME equipment is obsolete and strongly needs repair

But what could cause this when you follow an ILS cat2 for rwy26?!
I see no possibility to be that much offset in GS if you follow an ILS. The AFDS reverts to raw data when the ILS is identified and captured and follows it until 200ft. After which it will follow stored points and data up to touchdown, crosschecked with raw data to avoid ground interference.

Anybody have an explanation?

@Kulverstukas I also ran the figures and came to the same inexplicable conclusion.

I can say from my observations of the ADS data.....

- The approach was established Vref +5 prior to 1000' AAL so a stable approach.
- The descent rate was consistent with a 3deg glideslope, a false G/s capture would have been steeper but descending to the same correct touchdown point.
- The ROD was consistently accurate as one would expect with an ILS approach coupled to the autopilot.
- The QDM tracking is also very accurate for the same reason.
- The data at the end appears to show a GA from just below 100' AGL

......Which is all very inconsistent with the aircraft then crashing. Can this ADS-B data be considered accurate?

Those discussing the rising terrain upwind of the runway... Remember the crew is basing their GA decision on the Radio Altimeter at ~100' RA so even if they were well upwind of the correct touchdown point, the Rad Alt will still give you accurate ground separation for the GA decision.

If the crew were trying to grub-in below the weather.... there is no evidence of anything other than a continuous 3 deg descent followed by a GA.

I notice the ILS frequencies at both ends are identical. I wonder if some human error on the ground might present the correct localiser signal for 26 but a glideslope backbeam for the touchdown point of runway 08?

That would be more consistent with the trajectory of the accident aircraft.

The last data points at ~600 ft altitude are not credible IMO. It does not look like LOC during GA. If there was an attempt at GA, it had no effect, because at that point they were in contact with the ground well beyond the end of the runway.

Where there any aircraft landing before the ACT, if so, how much before?

Looks odd this, wrong ILS G/S trapped from the backcourse?

Didn't I just say that ?????

But then I would expect the GPWS to start shouting during the approach, resulting in an immediate G/A.
Also, your map on the ND would not coincide with your presumed position according to the glideslope.

I have never heard about such an occurence.

I'd think the path from a false LOC BC GS signal would not look like the steady descent profile we see from the ADS-B data.

Btw I compared FR24 data with Flightaware data. They seem to be from different sets of ADS-B receivers (possibly overlapping), but the data over the runway match within +/- 25ft altitude and < 100m position along the centerline. I.e., both show the aircraft well above where it should have been, well past the runway threshold.

Someone else asked if there were aircraft approaching that runway prior to this one. If so, how did the FR24/FlightAware vertical profile compare? Is someone willing to dig that data up?

Another crosscheck on the positional accuracy of the ADS-B data can be made by looking at the departure taxi from the cargo ramp in HKG to runway 07L. The turns are in the right place and the plane stays right on the centerline. This was hours earlier and things can change or drift if IRS position or faulty radio nav is somehow put into the mix I suppose. This plane was a former SQ freighter, anybody here know how they were spec'd for initial ADS-B installations?

Looking at the vertical approach profile in the FR24 .kml file it looks to me like possibly the glide slope was captured from above. Or, was it chased with vertical speed and paralleled a dot high with the throttles back to try to get on profile and never captured before minimums?

On some aircraft if you arm the approach while above the glideslope you get a roller coaster pitch over to capture so the 'technique' was to get right on the path with vertical speed and LOC and then arm the G/S when it was centered. As things got stable it was easy to forget to arm the glideslope. Or, so I'm told. Nowadays if things aren't right inside the marker you break it off and sort it out before another try. At least, that's what we are supposed to do.

Was this really a stable approach? Take a look at the groundspeeds in the FR24 and FlightAware data. It appears that the plane was still decelerating significantly all the way down the glideslope. This may have been due to a decreasing tailwind in the descent or maybe late configuration for the approach, either way with light runway winds looks like the power was back to lose speed as they approached the runway.

On FR24 the previous landing appears to be Aeroflot 1882, an A321 from SVO at about 0010Z, a little over an hour earlier. They appear to touch down in the landing zone and take the mid-field turnoff to the ramp.

The two sources are consistent, but that doesn't mean they are accurate.

We have already seen in a previous post (#19) that, based on the QNH in the METAR, the aircraft would have been reporting heights roughly 300' lower than actual.

RoD will, of course, be the same as implied by the ADS-B data.

Yes but that just makes it worse, doesn't it? I mean, the aircraft would be even higher (above the GS position) than where it was supposed to be.

Just the fact that the data shows a steady descent (with no leveling off), if the ADS-B horizontal positions are correct (as they appear to be) then must conclude that the aircraft was well above where it should have been, well past the runway threshold.