The pilot was on a visual approach to 28R. 28L was closed at the time and it’s own lights were turned off. It is certainly possible to imagine how the pilot, expecting to see two illuminated parallel runways and land on the rightmost one, mistakenly lined up on the taxiway to the right of 28R.
I am surprised that the rules for airport lighting permit the lighting for parallel runways to be independently switched. Certainly the pilot should have realized the taxiway was not actually lit like a runway, but it could definitely have been a very confusing picture to him, especially if he had never been to the airport before (or only during the day).
Came in to say this. I landed here two weeks ago [KBBG Branson, MO]. Note there is no taxiway to runway 32, just a loop for the larger planes to turn around at the end. This is a real “gotcha” for departing pilots who are unfamiliar. After the readback to Clearance Delivery, we were instructed to “Runway 32, via Charlie”. I almost zipped out onto the runway to back taxi, before stopping myself. (We were coming from the GA area, the easternmost ramp area on the chart]
For nonpilots: In days of yore, clearance to “Rwy X” meant clearance to cross everything in between, including runways. This is no longer true. Airports like this, without taxiways to a specific runway can be confusing.
That does seem like the most likely explanation, although it requires multiple safeguards to be overlooked (ILS, co-pilot, etc.).
Some are saying that that Air Canada pilot would have seen the planes when he got closer, and pulled up. But considering that he cleared the first plane by only 30 meters, and that they were closing at 140MPH-ish, I would guess that he had 3 seconds or less to act. He might not have collided head-on, but he wouldn’t have cleared the first plane, either, and probably sheared off the top like what happened at Tenerife.
The problem with night time and lights is that sometimes a bunch of lights just looks like a bunch of lights.
If the 28L lights were off and the crew were expecting them to be on then it’s a pretty understandable mistake to make initially. The brief might have included something like, “threats/hazards: It’s got parallel runways and we are expecting 28R, so the right set of lights then.” Ideally they would tune the appropriate ILS, but even then, once presented with a visual picture that matches their expectations and an instrument picture that doesn’t, they may assume the instruments are in error or they’ve accidentally tuned the wrong ILS. If you’ve got a mental picture of what things should look like, and they appear to look the way you expect, it can take a proverbial whack over the head to shake you out of the misconceptions you’ve built up.
I have heard of kind of the opposite problem at Sydney, Australia. We regularly fly into SYD which has parallel runways like SFO but in the wee hours of the morning the right runway is closed and the left runway normally has a partial closure with a displaced threshold and work vehicles all over the closed part of the runway. I have personally seen it like this where the runway that was fully closed had all of its lights on bright and the partially closed runway, the one we’re supposed to use, was just a sea of work vehicle lights. Not surprisingly some pilots have tried to line up on the fully lit runway. Familiarity is essential in these circumstances.
So having both runways lit can be just as bad as only having one runway lit, for different reasons. The key to the mess is knowing whether both runways are lit or not. If the ATIS said “RW28L approach and runway lights are off” then the crew would be less likely to make the error. In the Sydney case, if the ATIS said, “34R runway lights are ON”, we would know what to expect once we get visual.
As far as parallel taxiways goes, the space is limited and they have to make the most of what they’ve got. In most cases this means parallel taxiways. The lights on the taxiways are useful for distinguishing between a taxiway and runway on the ground, but not so much from the air. Blue lights, orange lights, green lights, red lights… What’s is the effect of a lack of ambient light on your vision? Poor colour detection.
It’s not just taxiways that have been mixed up in the past. There have been cases of jets lining up with highways.
There is an ATC recording of all this happening. The tower told them to go around, but we don’t know if they were already in the process of going around anyway. If that was the case then they had already fixed the problem. If not, then it could’ve been pretty bad.
Here is a report of a similar event from 2009 that happened to Delta Airlines at Atlanta. A difference being that no other aircraft were on the taxiway and they landed without harm other than to their careers.
There’s lots of discussion on that messageboard by guys who mostly know what they’re talking about. It’s a darn good bet the pilots, both of them, had flown in and out of Atlanta nearly every day of their careers there. Which proves that even tremendous familiarity can be a two-edged sword.
For reference, here’s the airport diagram: http://flightaware.com/resources/airport/ATL/APD/AIRPORT+DIAGRAM/pdf. Nothing material has changed in airport layout since that event. The key point is that while on approach you’re looking at quite a mass of lights. From left to right there’s runway 28, taxiway SG, a big gap, taxiway R, runway 27L, taxiway N, runway 27R, taxiway M, taxiway L, then the terminal complex. Beyond the terminal continuing to the north are 4 to 6 more taxiways interspersed with two runways.
Just within the narrow span of taxiways R through L encompassing the 27L & R runways there are 4 columns of runway edge lights, 2 columns of runway centerline lights, 6 columns of taxiway edge lights and IIRC 2 columns of taxiway centerline lights. Some of the black spaces between columns of blue lights are taxiways. Other black spaces are grass or culverts. Some of the blank spaces between white lights are runways or halves of runways. Others are grass and culverts.
As you’ll read about the Delta report, they let themselves be set up by accepting ATC offers to change the intended runway a couple of times when there wasn’t time available to fully realign 2 minds, 2 radios, and two computer systems. I’m not trying to shift blame. The point is that a screw-up is almost always a team effort to set the stage for failure then some specific individual fumbles the ball and now we’ve got an incident or accident.
This stuff gets done right hundreds of times every hour of day or night. And once in a great while it isn’t.
The eventual story of Air Canada at SFO will be known. It may be a rather different scenario than Delta had. I offer it only for comparison and the fact that the fully digested incident details are available for review now. Which won’t be true for Air Canada SFO for a year or so.
Visual ops work fine 24/7 if the ceiling & visibility are good enough.
Believe it or not, ILS is usually not used at any time of day or night on that runway at SFO. See below for more.
Llama, I know you know what follows, but you intro’ed the situation perfectly so I’ll pick up from your lead-in …
At SFO the two primary landing runways are the 28s. See http://155.178.201.160/d-tpp/1707/00375AD.PDF. SFO was built before WWII and the runways are waaay too close together by modern safety standards to permit airplanes to land side by side using both runways at once. But that level of throughput is an absolute necessity to handle the traffic demand at one of the most dynamic business cities in the US that’s also a gateway to the Far East.
They square that circle by operating an ILS directly aligned with the runway 28L for one stream of traffic. To drive a separate stream of traffic to 28R they use either an LDA Localizer type directional aid - Wikipedia such as http://155.178.201.160/d-tpp/1707/00375LDAPRM28R.PDF or an RNAV GPS http://155.178.201.160/d-tpp/1707/00375RPRMX28R.PDF approach leading sideways at an angle generally towards, but not to, runway 28R. The latter approaches drop you off about 4 miles out well east of the extended centerline of runway 28R. From there if you just keep going straight, you’ll fly by the right side of the airport by 1/4 mile with the airport off to your left.
Once the approach dumps you off with that 1/4+ mile offset at 4 miles to go, your job is to have already seen the runway visually, perform an S-turn to the left to settle in laterally, while also maintaining proper descent angle and speed to arrive at the runway threshold. And also keeping the other airplane landing just ahead on the other runway in sight. And make damn sure you don’t overshoot the runway you’re going towards by even a smidgen. That sets off alarm bells when you encroach on the other runway & airplane and you get to go around and try again.
The punchline to all this jargon being that from 4 miles out there is no radio-based guidance to the end of the runway. Depending on the specifics of the nav setup in the Airbus for that particular airport (I’ve never flown busses) there may not be lateral and/or vertical guidance from the onboard GPS nav system either.
Add a little fog or haze, and you’re looking for a dimly lit runway alongside some dimly lit taxiways most of whose lights are obscured by taxiing aircraft.
To further add to the fun, at newer airports with more adequate facilities ATC doesn’t try to orchestrate take offs and landings on the same runway. That calls for extra precision timing which increases pilot workload and detracts from a well-stabilized approach.
It also arrays taxiing aircraft right next to where landings are happening. Which raises a risk aptly highlighted by this incident. If a landing airplane leaves the runway at any point in the approach, touchdown, or rollout, the odds of a chain-reaction event are much higher than if distantly separated runways are used.
More typically SFO takes off on the cross runways. Which has its own timing challenges, lest we do something resembling the Figure 8 racing Figure 8 racing - Wikipedia common in the 1960s.
But late at night when SFO is launching lots of very heavily fueled 777 & 747 & A380 traffic to the Far East, they need the longer 28 runways to get airborne. Despite the fact those runways point the takeoffs right at a narrow pass between two hills covered with houses.
As I said earlier, these maneuvers are done correctly every couple of minutes at SFO alone hour after hour day after day year after year. But once in awhile they aren’t. SFO is but one of many places in the US airport ecosystem where every day the deck starts out mildly stacked against everyone operating there, both ATC & pilot.
Apparently my estimate of 3 seconds was wildly optimistic. CBS News says:
and
Yeah, easy, if you can’t distinguish between blue (taxiway) and white (runway) lights, you aren’t using your instruments, and your co-pilot isn’t awake.
It really isn’t as easy as it might seem. Different colours of lights aren’t all that obvious from a distance at night, it’s more the general picture that you’re looking for plus cross checking with whatever instruments are available. As an example, I was cleared a night visual approach on the PAPI (visual slope guidance lights) recently. The ILS was out of service and there was an airport vehicle conducting a runway inspection. The vehicle, with headlights on, was adjacent to the PAPIs just as we were supposed to be starting our final descent. The car headlights swamped the PAPIs, even though there was a good 20 meters of lateral separation and I couldn’t distinguish between the red ones and the white ones very easily at all.
Check out the above video, how many blue and green taxiway lights can you see? The airport is probably swimming in the things, but they are designed to be seen by taxiing aircraft, not flying aircraft.
The runway itself should have been pretty obvious, but the taxiway, with four aircraft sitting on it with their collection of beacons, nav lights and taxi lights, wouldn’t have looked like much other than a bunch of lights. Obviously I wasn’t there and don’t know what went on, but if the crew were convinced that both sets of runway lights were on then it would have been very easy to convince yourself that the mess of lights to the right of an obvious runway is the other runway, maybe with some work vehicles on it.
It seems from the TSB Canada preliminary report that the A320 had already commenced a go-around when ATC instructed them to. They overflew the first two aircraft by ~100’, the third by ~200’, and the last by ~300’. That means they were well and truly in the go-around before getting the ATC instruction.
Edit: I don’t want to sound like I’m mindlessly defending the crew, I’m not, I’m just trying to explain how this type of error is a lot more understandable than a layman might think.
Here is the actual radio interchange with ATC, combined with an animated video of the incident. If the animation is at all accurate, it was a damn close thing.
Echoing **Richard **here. They screwed up. Period.
But anyone who says the mistake is inconceivable or impossible or they must have been asleep is under-informed about the realities of night visual approaches to big airports loaded with thousands of steady, blinking, and moving lights of every color. Many of which simply look yellowish through the haze.
This math is invalid. The 30 meters is vertical distance and the 140 mph is horizontal speed.
For very round numbers on final approach we descend at ~750 feet per minute = 12 feet per second. So ~100 feet of altitude is ~8 seconds of descent.
Eight seconds is a friggin’ eternity in jets.
A bottom line point: Aviation, as the saying goes, is terribly unforgiving of any incapacity, error, or neglect. It is easy for outsiders to fall into the trap of thinking that big bad consequences necessarily must have flowed only from big bad mistakes. Not so. A tiny easy-to-make error can create an epic disaster.
They *did *nearly create an epic disaster. That does *not *imply they made a huge mistake when measured in terms of how easy or hard it is to make.
I have hunted for such a doc and can’t find it. Do you have a link? I tried going through that PPRuNe thread, hoping to find something useful. I gave up after 6 pages; it was a hard slog with a lot of people saying silly stuff based on not even looking at any docs on SFO, much less ever having operated there.
As is usually the case at this stage nowadays, there’s a lot of publicly available data on the path the aircraft took. And on who said what when over the radio. What’s mostly absent is the actual configuration of the environment (weather, lighting, NOTAMS) and of the configuration of the aircraft (modes, FMS programming, flap/weights/speeds, etc.).
What’s utterly absent at this stage is any info on what the people involved were thinking or doing instead of what they should have been thinking and doing.
Ultimately this near-accident probably has the large majority of its roots in human factors. With some minority roots in systemic decisions to use less than ideal approach aids to shoehorn more traffic into an inadequate airport.
Until the authorities or the pilot community or, especially the lay media and lay public, know what the crew was thinking and saying to one another as the event unfolded, we lack the information needed to generate an informed opinion. The other kind of opinion are, as always, readily available and loudly promoted.
Human eyes have a better range of sensitivity than does a vid cam. But overall your impression is about right.
At an easy-to-operate airport, the runway is a distinctive well-lit space in the midst of the almost-blackness of the rest of the airside of the airfield surrounded by the randomly-lit terminals, hangars, and the surrounding city. At a hard-to-operate airport, the runway is a slightly-better-lit space in the midst of a slightly-less-well lit space surrounded by the randomly-lit rest of the city.
At SFO it’s easy to tell the bay from dry land, except for the dark nearby hills. After that the details get harder to find reliably. A couple hundred crews do that right every night for many years straight. Then the statistics catch up.
Have been thinking about the above comment a bit lately.
In my experience, pilots who are the most dismissive of other’s errors are the most likely to make errors themselves.
“I wouldn’t do what they did because I am good and they are not” is not a useful thought process to have. If pilots have the attitude that another pilot ballsed up because they were different in some fundamental way, they lose the opportunity to learn something from the incident.
My opinion is that I would be very unlikely to make an approach to a taxiway. It is important, however, that I question why I have this opinion. What is it that I actually do in the aeroplane that prevents this type of error from developing? It is no good to say, “I am a professional, therefore I don’t make this kind of error.” Instead I must say, “I am human therefore I am capable of making this error and must take steps to catch it before it develops in to a PPRuNe thread.”
To get the most out of the discussion around these types of incidents, you really need to know what the crew were thinking, or at least what they were talking about and what they’d done in the cockpit. What was it that lead them down the garden path? If you can learn what led to the incident occurring you can hopefully identify whether your own procedures would’ve been enough to stop it happening to you or if maybe you need to make some changes.
We’re in the midst of a sea change in how we teach and talk about this. In the latest model, opportunities for error occur many times on every single flight without exception. And errors are made on every single flight without exception. It’s the uncaught errors that increase risk or, at the limit, cause an accident.
The goal is to accept all the above facts, then learn to identify errors promptly & correct appropriately. As well as to develop the procedures and techniques that will put more and better barriers upstream between the opportunities for error and the development of the proto-error.
When you’re actively seeking the opportunities for error and proactively taking steps to first prevent errors, then detect and correct those that happen anyway, you’re playing the game right.
The attitude “I’m too skilled to make error X” is totally antithetical to that model.
I know the other pros know what follows, but for the benefit of the rest of the readers…
In the case of the night visual approach, error prevention includes things like having the best available FMS set-up and navaids dialed up so you have as much computer and electronic help as may be available. Ensure per the NOTAMs and ATIS and published airfield info that you *know *the actual lighting configuration today of all parallel runways and nearby taxiways. On the approach positively identify each and every runway and taxiway and that it completely matches the info you’ve read. Actively question any discrepancies, even minor ones. Verbalize what you see, what decision you’ve made about which bright spot you’re aiming at, and obtain verbal concurrence from your partner that he/she came to the same conclusion independently. If you’re the partner, don’t accept any differences in mental models; dig until every shred of difference is uncovered and resolved.
Then keep a questioning attitude someplace in your mind as you get closer and your view of the situation improves. We often have to make tentative decisions based on partial information and then refine them as we get closer. It’s vital to stay very aware of just how tentative any given decision is; folks get suckered when a preliminary decision becomes locked in their mind and they subconsciously discard all the slowly accumulating evidence that their original decision was faulty. The human mind has a strong bias towards finalizing decisions too early. It takes active measures to compensate for that.
All that sounds fairly easy bordering on obvious just sitting here reading about it.
But has to be done in addition to actually driving the machine, managing the computers & events, and dealing with continuous radio chatter. After 12 hours at work and several time zone shifts. Being tired is IMO the largest obstacle to the mental discipline needed to do all of it diligently every single time, not just most of it most of the time. No matter how busy or rushed things get, this process can’t be relegated to such a low priority that it gets short shrift or is skipped entirely.
