Fly on the right – A mid air mitigation

On 29 September 2006, A GOL B737-800 and a private Embraer Legacy business jet collided at FL360 some 200 miles north of Brasilia, over the Amazon Jungle. The Embraer’s left winglet hit the 737’s left wing, and the 737 crashed killing all on board. The Embraer was luckier, and made a successful emergency landing at the Cachimbo air base.
Both aircraft were equipped with transponders and TCAS, so how could it have happened? But the technicalities divert attention from a vastly more important point: Was it simply bad luck that the aircraft were so close together horizontally that they collided?

The answer, as everyone flying today knows full well, is ‘No’. Thanks to GPS, horizontal accuracy worldwide is now officially less than a typical wingspan. There is no protective scatter, no fat left in the solution (published accuracies understate the problem: put two GPS receivers together in the same place and, because they have the same ionospheric errors, they will agree their position to only a few metres). If you make a mistake in the vertical plane on a two-way airway, you can expect to find someone coming right for you at Mach 1.6.
We have been here before. In June 1999, a BA 747-400 and a Korean Airlines (KAL) 747 came within an estimated wingspan of a mid air collision over China, while both of them were responding to TCAS alerts (the estimate came from the co-pilot’s remark that half the KAL 747 ‘filled P1’s DV window’, see fig 1). Extraordinary detective work by BA engineer, Andrew Rose, showed that a single wiring fault would cause the failure of the altitude comparison function. Via email and fax, Andrew led  KAL engineering to a single bent connector pin on the rear of the Transponder.  A subsequent undetermined error in KAL’s altimeter encoding was therefore not detected, resulting in the intermittent erroneous transmission of own altitude data to KAL’s TCAS equipment, and that of other aircraft. The result was that the two TCAS units issued Resolution Advisories which brought the two aircraft together instead of separating them.

After such a close shave, BA naturally carried out an investigation that resulted in a number of recommendations. In addition to several TCAS-specific issues, the final recommendation was that the issue of offset tracks should be pursued as a potential mitigation (the two aircraft involved in this incident being on reciprocal tracks). Between 1999 and 2005, BA was aware of 8 TCAS incidents where incorrect altimetry appears to be involved, five of which concerned aircraft pairs on reciprocal tracks. It is likely that this is merely the tip of the iceberg, as many other operators may not be able to identify the problem.
It is obvious, however, that TCAS failures  are not the only possible source of altitude conflicts, though they do have the unique effect of both causing the error and preventing its resolution by TCAS. Other potential errors include human errors in TCAS interpretation (e.g. Uberlingen), and ‘straightforward’ level busts. These can be due to simple errors or misunderstandings, autopilot anomalies and turbulence. An example of an incident involving both of the last two was the Turkish Airlines A340 incident in October 2000, which suddenly left its flight Level on a NAT track, narrowly missing an A330 on the same track 1000 ft above. The AAIB investigation (Bulletin 6/2001) recommended:

• 2000-68 – That the CAA take forward a recommendation to the appropriate international bodies that they reconsider the need for commanders to inform ATC of all lateral offset manoeuvres of less than 2 nm in Oceanic airspace, irrespective of the reason for the manoeuvre.
• 2000-70 – That the CAA take forward a recommendation to the appropriate International bodies to consider standardising lateral track offset procedures which are independent of wind direction.

In November 2000 ICAO issued State Letter AN13 11-6.00 96 which recognised the problem and introduced guidelines for the application of track offsets. These were confined, however, to ‘Remote or oceanic’ airspace and to aircraft equipped with GNSS navigation. But it is obvious that risks are higher over land due to greater traffic density and proportion of non-cruise traffic. Over land, DME/DME gives highly repeatable results so that two FMS equipped aircraft can be expected to overlap, as we all know; both the BA and the KAL 747 were DME/DME FMS equipped. Serious thought needs to be given widening the application of offset procedures. The AAIB’s recommendations apply to all RVSM airspace.
‘Flying on the right’ is good, but it isn’t enough to protect against both opposite and same direction traffic, we need random offsets too. But we cannot apply random offsets of more than a mile without blundering into nearby routes…. One simple modification to the most common FMS solves the problem of integrating offsets in congested airspace: the ability to ‘micro’ offset in steps of 0.1 nm. Most current FMS limits offsets to whole mile values. The change would increase the number of random choices available by 10 without eroding separation standards, see fig 2. Such a change would not be difficult; it should be added to future systems and be available to present units through upgrades. In fact the B737NG can already do this, and there are lots of them around.

After the Uberlingen mid air, there was immediate action in ICAO (mandating adherence to RAs). The silence after the Brazilian accident was therefore deafening and puzzling.
I must admit now to the deepest naiveté. I thought that once I had published micro offsets, an idea that falls in the category of ‘blindingly obvious’ if ever there was one, in various official places in the UK and elsewhere, things would begin to happen of their own accord. They didn’t. There has been progress, but each step was only as the result of a personal shove. Micro offsets are now on the agenda of ICAO’s Separation Assurance Panel (SASP), and are awaiting the pronouncements of their Maths sub group. Acceptance at an international level is of course vital. Equally important is that FMS and Aircraft manufacturers offer the capability as standard. Responsibility for the FMS functionality is divided between the two groups, and there is a natural tendency for each to claim that it is the other’s job to make the (small) changes needed here. I am pleased to say that I have received personal assurances (though sadly no official press statement as yet) that both Airbus and Boeing will include micro offsets on their products in the near future. Finally it will be up to individual pilots to use the capability sensibly. IFALPA are in favour, so I am hopeful this last step will happen.
I hope this blog will be read by ATC as well as aviators. I think I detect a natural reluctance on the part of ATC  to accept micro offsets. All sorts of objection are raised (see below) but in the background I think I detect a feeling that this is some sort of criticism of ATC in general, It isn’t, It is simply a recognition of the extraordinary accuracy achieved by modern Nav systems. True, 1 mile right is a de facto rule in areas of the world where, shall we say, ATC is not as good as it should be, but that isn’t what micro offsets are all about. They just put back the scatter that was once there naturally.

Offsets – Frequently Raised Objections
This will introduce new errors	Offsets are current technology and well protected (check your flight manual for how the FMS prevents you from applying an offset on the approach for instance)
Europe doesn’t have many two way routes	No, but it has some, and anyway aircraft fly world wide.
Random routes offer better protection	True, but the world doesn’t yet have a totally random route structure
Retrofit of micro capable FMS  will be expensive	Maybe, but who said retrofit is necessary? – there are aircraft (737NG) that are capable of micro offsets now, to the benefit of all. As more aircraft are able to fly micro offsets, the protection just increases
This will increase controller workload	Why? It will be invisible to the controllers; the pilot makes the random choice of offset, not ATC
It would be better to eliminate two way airways and/or to code the  routes to fly right of the nominal track.	Good ideas, not incompatible with pilot selected offsets, but they don’t protect same direction traffic.
Offsets are incompatible with low-RNP routes	We should be so lucky as to have such things, until then, fly offsets.
The risk isn’t worth the effort	I wish it were true. In recent years there have been 3 well known head-on midairs (Canada, India and Brazil), and the risk can only increase (see main text)