Use of ADS-B and other Mode S data in accident investigation

[DaveReidUK]

(mods may wish to move to Accidents and Close Calls or even – dare I suggest – make it a sticky there)

Always a controversial topic on PPRuNe is the question of whether ADS-B and Mode S data can add to the understanding of an accident or incident - viz the recent Maui, Austin and JFK events.

A recently-published accident investigation report from NSIA, the Norwegian AIB, discusses in detail the contribution (and caveats/limitations) of using such data and is well worth a read. It concerns an accident to a GA twin (Diamond DA42), so not equipped with either FDR or CVR, and uses data from both the ANSP (Avinor) and FlightRadar24.

Here’s an extract describing the philosophy:

As the wreckage, crash site and witnesses provided little information, the NSIA decided to investigate whether a better understanding of the flight could be gained by compiling Automatic Dependent Surveillance-Broadcast (ADS-B) and Enhanced Surveillance (EHS) data. Data were obtained both from Avinor and from Flightradar24. The data were validated and obvious errors excluded. Data from different sources were corrected to enable comparison. The data were then collated and presented graphically. Based on the data, it has been possible to gain a better understanding of the flight up until the time when it was 417 ft above the terrain. The purpose was primarily to understand the airplane’s movements when the crew lost control and it subsequently fell to the ground. A further description of the methods used is given in Appendix A.

Link:

Air accident 8 km west-northwest of Larvik in Vestfold og Telemark County, Norway on 23 November 2021 involving a Diamond DA 42 NG, LN-PFM, operated by Pilot Flight Academy (https://www.nsia.no/Aviation/Published-reports/2023-02?pid=SHT-Report-ReportFile&attach=1)

[First_Principal]

This a worthwhile and pertinent discussion to have.

To answer your question simply; yes such data may add to understanding of incidents and accidents in some scenarios.

However I think some caution should be exercised, particularly for those using the the interpreted graphical displays often shown up here on PPRuNe.

Why? - a while ago Pilot DAR enquired about an accident in which some ADSBExchange data was presented showing, as I recall, a very significant rate of descent. At least with ADSBExchange the underlaying code (tar1090 IIRC) is open-source and available for inspection. Having had a, admittedly brief, look at that code it suggested to me that the raw data may be interpreted to some extent before it's displayed to the end user. While I have no doubt the intent would be honourable and designed to improve utility, when it comes to important accident investigation I suggest that only the most base raw data should be used. It may be that some interpretation would still be necessary, but at least that would (should) be known and under the control of the investigators. In such cases it's my further view that were any data interpretation carried out that detail should included in the report so it is available for scrutiny by all.

In the matter of FR24 and other similar systems I claim no real knowledge, although I'm aware of the name and the purpose of the site. If, however, it's a proprietary system and the methodology/code is not not available for inspection then I would be even more cautious about relying upon information from these sources. It could still be useful to assist in understanding, but without full knowledge of the path the data takes from the transmitter to your screen I'd have to consider it not necessarily reliable per se.

I'd like to think that investigators would be aware of such possible issues, and be able to obtain raw unadulterated data, but at the present that would seem nothing more than conjecture? Thus I reiterate full disclosure a la the open-source philosophy would be advisable, and allow readers of results to be assured (and be able to check if they felt so inclined) how such results come to pass.

FP.

[AerocatS2A]

I think ADSB data is excellent provided its limitations are understood and there is no better source available. In the case of a GA aircraft with no FDR or CVR it makes a lot of sense to use data that is available. From what I’ve seen over the years in this forum DaveReidUK, and others, have a good understanding of the strengths and weaknesses of data from FR24 and similar sites and are capable of using it responsibly. Caution just needs to be taken that we don’t read too much into single data points etc. in fact the same caution needs to be taken with FDR data, the difference is that we get FDR data after experts in the field have studied it and so we’re not just handed the raw data to make of what we want, it tends to come with an expert narrative attached.

[Asturias56]

ADSB data processed by professional investigators - yes, of course

data processed by amateurs - no

[ehwatezedoing]

https://cimg5.ibsrv.net/gimg/pprune.org-vbulletin/750x1334/img_6301_45c114c51b11b36cca92da4f27346a628248524f.jpg

567KTS for Caravan at 12000'
Although it's pretty obvious in this case, here is why you should never bank on those "data" at 100%

[DaveReidUK]

This a worthwhile and pertinent discussion to have.

To answer your question simply; yes such data may add to understanding of incidents and accidents in some scenarios.

However I think some caution should be exercised, particularly for those using the the interpreted graphical displays often shown up here on PPRuNe.

Why? - a while ago Pilot DAR enquired about an accident in which some ADSBExchange data was presented showing, as I recall, a very significant rate of descent. At least with ADSBExchange the underlaying code (tar1090 IIRC) is open-source and available for inspection. Having had a, admittedly brief, look at that code it suggested to me that the raw data may be interpreted to some extent before it's displayed to the end user. While I have no doubt the intent would be honourable and designed to improve utility, when it comes to important accident investigation I suggest that only the most base raw data should be used. It may be that some interpretation would still be necessary, but at least that would (should) be known and under the control of the investigators. In such cases it's my further view that were any data interpretation carried out that detail should included in the report so it is available for scrutiny by all.

In the matter of FR24 and other similar systems I claim no real knowledge, although I'm aware of the name and the purpose of the site. If, however, it's a proprietary system and the methodology/code is not not available for inspection then I would be even more cautious about relying upon information from these sources. It could still be useful to assist in understanding, but without full knowledge of the path the data takes from the transmitter to your screen I'd have to consider it not necessarily reliable per se.

I'd like to think that investigators would be aware of such possible issues, and be able to obtain raw unadulterated data, but at the present that would seem nothing more than conjecture? Thus I reiterate full disclosure a la the open-source philosophy would be advisable, and allow readers of results to be assured (and be able to check if they felt so inclined) how such results come to pass.

Some very good points made.

The (professional) investigation report that I linked to makes a distinction between the data captured by the ANSP and that provided by FlightRadar24 (the latter presumably similar to the data it publishes for public users).

To be fair to FR24, they make no attempt to run sanity checks on the data that they host - doing that is down to the end user.

One fundamental limitation of FR24 ADS-B data is that they fabricate transmissions that include both position and velocity data - that isn't how ADS-B works, and the inevitable consequence is that you can't be sure whether a supplied timestamp relates to the position or the velocity transmission from the aircraft. The Norwegian reports states that "Many [velocity and track] datapoints [in the FR24 data] were deleted because of great deviations".

A further limitation (regardless of the data source) is the Vertical Rate (i.e. RoC/RoD) ADS-B data item. Clearly it can't be an instantaneous value as it's normally baro-derived, so it's often at odds with the altitude timeseries in the data.

Lastly, the Norwegian report references Mode S EHS (Enhanced Surveillance) data. FR24 can't provide that data, so the ANSP was the only source for that, which include lots of useful parameters including IAS, TAS, roll angle (particularly relevant for the investigation in question).

So, in summary - is FR24 data useful in understanding what might have happened in an accident/incident ? Answer: Possibly, but only up to a point, and only if one appreciates the caveats and limitations associated with it.

[ATC Watcher]

It' been a few years since I did (ATC) incidents investigations but FR24 was not used to investigate , we have better tools , real ADS-B and enhanced Mode S ( downlinking of aircraft parameters) . which is available via RADNET on almost the whole of the core area of European continent ,and I suspect the same is true for mainland US ( exculding Alaska I think ) . For out of "standard ground based " ADS coverage , there are other options , Iceland , parts of Greenland and Canada use Space-based ADS-B . There are however still gaps, on the standard terrestrial system notably at low levels and in mountains areas
As you correctly point out , DaveReid UK, FR 24 data is often incomplete and "voids" are calculated. It is also not cerified. It gives a good indications . but not the exact values , Howver , like in Norway or in the Alps, (and probably the US rocky mountains too) some valleys where there is no "official " ADS-B coverage but where a local farmer had puchased a FR24 or OGN antenna on his roof and sends the data, there, yes FR24 can provide good additional info . Especially for SAR .

Note (*) the nterm ADS-B is generically used to mean many things. What FR24 does is not really the ADS-B we used for separation in ATC. Real ADS-B is more than just downloading transponders data. That said , it is fantastic generic tool and , above all , unlike real ADS-B , it is free :E

[DIBO]

... where there is no "official " ADS-B coverageand in the case of the Austin incident, do you expect much "official" ADS-B coverage below, say 300 - 200ft, to be available? Wouldn't be surprised if the NTSB will include the "internet ADS-B" into its investigation...

567KTS for Caravan at 12000'. Although it's pretty obvious in this case, here is why you should never bank on those "data" at 100%It has been discussed many times before, that FR24's MLAT for non-ADS-B equipped aircraft, has only limited value and gives only a rough indication what the trajectory is. Nevertheless it's a pity that FR24 after all this time, still has not yet implemented a few simple filtering and smoothing algorithms so that these plainly stupid calculated values, can be discarded.

[Pilot DAR]

I am a neophyte with ADS-B and FR24, so I guard myself against taking it too seriously. That said, ADS-B provided to me by an authoritative source with respect to a recent accident, though not truly conclusive, certainly triggered a worthwhile discussion about what the airplane was doing just before an inflight breakup. And, knowing [roughly] where and when for the in flight breakup (because the output stopped) was useful. That said, two more recent test flights I made were picked up by someone on FR24 as having been two minutes duration each. That is not actuate. So my trust in that information is not entirely certain.

But, with accident investigation, you consider everything you have, and then how credible it is, and how it fits in with the rest of what you have....

[RatherBeFlying]

FR24 is pretty much exclusively Earth Frame of Reference. Wind Frame is not given. Some ADS-B sequences show Vehicle Frame data such as TAS. Wind values can be computed by taking the differences between Earth and Vehicle Frame data, provided sufficient data is available.

[DaveReidUK]

FR24 is pretty much exclusively Earth Frame of Reference. Wind Frame is not given. Some ADS-B sequences show Vehicle Frame data such as TAS. Wind values can be computed by taking the differences between Earth and Vehicle Frame data, provided sufficient data is available.

The ADS-B spec allows airspeed as an alternative to groundspeed, but it's very rare to see that used.

Mode S EHS, on the other hand (not part of ADS-B) provides for airspeed (both TAS and IAS), as well as heading, and as you rightly say it's straightforward to calculate wind velocity from those.

For example I'm just looking at an Iberia A320 on the Heathrow ILS: heading 273°, track 270°, groundspeed 126 kts, TAS & IAS 146 kts