UPS Cargo Jet Crashes Near Birmingham Shuttlesworth International Airport

[Dispatch Dog]

N-ONE, that IS the question. It is a well-recognized problem. The industry is working to remedy this tendency to continue a wonky approach and I think we are going to see a stepped up effort soon. Reasons might include schedule pressure and fatigue, a task-orientation and confidence common to pilots, a bad safety culture (past accident rates do not necessarily reveal that), cockpit gradient, weak training on stabilized criteria or simply a failure to monitor. We've had three accidents this month that might all be the result of a reluctance to call a go-around. I would say this issue is one of the greatest threats to safety right now. It is not because we can't define stabilized and unstabilized. It is because pilots are failing to either recognize it or correctly react to it. The answer lies in training, culture and human factors.

I would say that human factors best describes the root cause of the issue in this group of accidents. Namely target fixation.

Ironically, often referred to in a positive way as determination as it is easily transposed.

The fixation relieves the concious mind of the burden of peripheral information and SA as they concemtrate their attention on only certain criteria. Not only is it common to all humans as a fundamental feature of our modern minds, it is trained in at school and especially in things like flight training. The ability to focus on a few critical issues while maintaining peripheral awareness is the part of latter or more advanced training and then gets better with accumulated experience. But not all humans are in control of their tunnel vision aspects as others.

In the extreme, target fixation can cause us to be aware of a danger, even the exceptional and iminent severity of risk and still push through to the objective in the delusion that we are expanding our personal envelope of performance.

On the whole, modern pilots are as good as it gets at overcoming this and their overall safety record proves it. But, as in many things, fatigue and overload can make it worse as well as complacency and lack of experience.

Miss a vital process one day, and the next many other processes suffer the consequence of slight neglect as you 'make sure it doesn't happen again'.

A rather simplistic analysis and just my opinion.

[3WE]

Gabe's approach plate questions.

...I do not know what "S-18" means...

Thanks to www.internet.com...

S-18 means "Straight in to Runway 18"

Other approach plates will often have a row labled circling minimums to other runways.

[AVION1]

The FDR looks in very bad shape !



It is supposed to look like this:

[AVION1]

Hi Evan,
With cockpit indications giving an unstable approach down to minimums, including a Ground Prox Warning at the last, why would an experienced crew not perform a go-around?

Hind sight is a wonderful thing for us all, but there is more yet to this puzzle. There always is.

RIP

I would go around if I don't see the Threshold !
and I fly single engine junks !

[snydersnapshots]

If you really want me to do the math, just ask.
But I don't see the point since there was no ILS available here.

If (again: IF) they were using an FMS generated glide slope, those ones don't follow the conic geometry of the ILS.

As far as the math goes, I figured you'd let me know if I was way out of the ballpark; I don't see any reason to do any heavy-duty ciphering for my example. DD asked how far below the glideslope you would have to be to hit the trees and so I thought I'd give him some data with the approximate thickness of an ILS GS about that far out for reference. Of course, my data doesn't account for rising terrain, etc. And you're exactly right about the FMS generated GS.

[Gabriel]

As far as the math goes, I figured you'd let me know if I was way out of the ballpark; I don't see any reason to do any heavy-duty ciphering for my example. DD asked how far below the glideslope you would have to be to hit the trees and so I thought I'd give him some data with the approximate thickness of an ILS GS about that far out for reference. Of course, my data doesn't account for rising terrain, etc. And you're exactly right about the FMS generated GS.

Now you've made me check it

Your math was correct: +/- 0.7° = +/- 70ft/NM

What's really interesting is that I've learned a new and very easy way to convert between slope and gradient: each degree = 100ft per each NM.

I should have known it, since that means that 3° = 300ft / NM, what explains the 3x1 rule.

[3WE]

8/18/2013...NTSB reports that the plane was on autopilot until "the last second before impact"

Gabe scores one for CFIT.

And this tidbit: "The crash occurred shortly before dawn in rainy conditions as low-lying clouds hung over Birmingham."

Sounds like "we" (the media, us, everyone) all fixated on the lateral visibility being OK without thinking about micrometeorology and topography!

Parlour talk moment- knowing there's so much topography in the area, could someone have a propensity to dismiss GPWS alerts- assuming you were confidently on the glidepath / approach profile?

[3WE]

I would go around if I don't see the Threshold !

Me too, and I'm sitting in an armchair going 0 miles per hour after, typing on a computer after a good 8 hours sleep with virtually no complacency or over-familiarity whatsoever since I haven't done a zillion instrument approaches in an A-300 at 5:00 in the morning.

[Evan]

A rather simplistic analysis and just my opinion.

I think you are hitting the nail there. I think the industry is well aware of this phenomena. Accident reports have also investigated the phenomena of confirmation bias, where a pilot will assess a situation wrongly by placing more credence on those things that tend to support his/her confidence and dismiss those things that might go against it.

The issue of target focus should be easily overcome by installing something called a PNF. The PNF would then monitor the conditions of the approach while the PF concentrated on the target. There is still the question of why this is not working correctly. We've heard about cockpit gradient, procedural distractions and of course fatigue, but in those critical minutes where the earth is rising up to greet you, I don't think it is so much to ask that the PNF be all about SA.

...NTSB reports that the plane was on autopilot until "the last second before impact"

...at which point the thrust levers were pushed to TOGA but no thrust was available due to an extended period at flight idle?

And what modes I would love to know. I don't know how much the A300-600F has been modernized with respect to autoflight. I assume there is now RNAV GPS. What about FPA?

[Dispatch Dog]

I think you are hitting the nail there. I think the industry is well aware of this phenomena. Accident reports have also investigated the phenomena of confirmation bias, where a pilot will assess a situation wrongly by placing more credence on those things that tend to support his/her confidence and dismiss those things that might go against it.

The issue of target focus should be easily overcome by installing something called a PNF. The PNF would then monitor the conditions of the approach while the PF concentrated on the target. There is still the question of why this is not working correctly. We've heard about cockpit gradient, procedural distractions and of course fatigue, but in those critical minutes where the earth is rising up to greet you, I don't think it is so much to ask that the PNF be all about SA.


...at which point the thrust levers were pushed to TOGA but no thrust was available due to an extended period at flight idle?

Doesn't "on autopilot until the last second" suggest that the pilots assumed all was going to plan and they thought the aircraft was in the correct position on the approach?

So what other info was telling them that all was correct until they saw trees ground?

CVR and FDR will produce only so much factual evidence, but for me, the real telling evidence will be what they were actually looking at during those last minutes/seconds, what was their scan including and what did they believe or not believe in the information that was being presented.

As in AF447, I think the pilots were selecting certain info that agreed with their existing perception and mentally attempting to explain information that conflicted. So when they heard "sink rate, sink rate", I can imagine the first reaction would be 'why?' then the scan of other information may (in their split second view) confirm this as an erroneous system call.

On approach over irregular topography, is it the norm that 'sink rate' is activated as they come close to hills that they regularly pass safely over on a correct approach path? If so, would they dismiss the call as the system detecting a known hill or area of raised ground?

So how would they know the difference?

I think these questions can be specifically answered by cockpit cams. I know it would be a nightmare to get them installed because of employment legislation and unions, but how often to final accident reports have to assume or speculate on what the pilots were thinking and looking at?

I'm happy with the usual security cams looking at me, but a camera in my face? I wouldn't like it so I understand the objection completely, but we could learn so much about the human factors that lead to incorrect perceptions, decisions and ultimately accidents.
They would have confirmed Egypt Air 991, Silk Air and many other questionable report assessments.

[Evan]

On approach over irregular topography, is it the norm that 'sink rate' is activated as they come close to hills that they regularly pass safely over on a correct approach path? If so, would they dismiss the call as the system detecting a known hill or area of raised ground?

So how would they know the difference?

Wouldn't the warning there be TERRAIN or ALTITUDE or PULL UP instead of SINK RATE? I assume these newer a/c have EGPWS, so maybe those warnings are still active away from the runway. But SINK RATE is about V/S, meaning there is terrain and you are descending too fast into it rather than flying to close over it, right?

[Dispatch Dog]

Wouldn't the warning there be TERRAIN or ALTITUDE or PULL UP instead of SINK RATE? I assume these newer a/c have EGPWS, so maybe those warnings are still active away from the runway. But SINK RATE is about V/S, meaning there is terrain and you are descending too fast into it rather than flying to close over it, right?

I thought so too, but here we have an aircraft that didn't seem to have a high VS but hit trees, so I was trying to be specific in relation to the actual warnings this crew got.

[N-ONE]

Hi Evan,

...at which point the thrust levers were pushed to TOGA but no thrust was available due to an extended period at flight idle?

I don't read that NTSB comment in quite the same way, however, if the throttles are pushed TOGA, engines should accelerate to max rated within aprox. 7 seconds from approach idle.

....I don't know how much the A300-600F has been modernized with respect to autoflight. I assume there is now RNAV GPS. What about FPA??

Depending in the build date and mod spec of the aircraft, the auto flight system for the A300-600 is capable of flying a managed Non-Precision approach.

[Dispatch Dog]

Hi Evan,

I don't read that NTSB comment in quite the same way, however, if the throttles are pushed TOGA, engines should accelerate to max rated within aprox. 7 seconds from approach idle.

Unless they are full of tree branches.

I am assuming that "impact" in the NTSB releases is relating to the trees, not the ground.

Again, that picture of one of the engines does not look like it was turning much when it hit the ground.

[Gabriel]

Evan is right.

Three things:

First, the "sink rate" warning is triggered by an excessive barometric vertical speed, not by an excessive radalt vertical speed (which is called "terrain closure rate" and triggers the "terrain" warning, not the "sink rate").
The "sink rate" warning is still dependent on the radalt, because the barometric sink rate threshold that triggers the "sink rate" warning depends on the radio altitude (AGL = "above ground level"). The "sink rate" is inhibited above 2500ft AGL. At 2500ft, the sink rate needed to trigger a sink rate alarm is 5000fpm. That threshold vertical speed is linearly reduced to 1000fpm at 0ft AGL. So, for example, it's 1800fpm at 500ft AGL and 2600fpm at 1000ft AGL.
This means that if the "sink rate" warning sounds, it tyipically means that either you have an excessive V/S, or that you are lower than you thought (for example, if you are intentionally holding a VS of -1200fpm and the "sink rate" warning is triggered, that would mean that you are just 125ft above the ground and that (if the terrain was even and you were to hold that VS) you are going to die in 6.5 seconds.

Second, I am against the unstabilized approaches, not like 3WE that defends them as a valid pilot's prerogative, but let's be fair: Even if the approach was unstabilized (and it was, as in any CFIT), the sink rate warning sounded just 2 times in a row 9 seconds before impact. That this warning stopped and didn't sound again means that they reduced the vertical speed and by a lot, since reducing it just a bit would have triggered the "sink rate" warning again as the radalt reduced. The VS at impact (0ft AGL) must have been less than 1000fpm.

Third, in the "old" (not "Enhanced") GPWS, the "terrain" warning, I don't know it so well, but it's similar to the "sink rate" one except that is triggered by the terrain closure rate and not by the VS, used different thresholds, and has some weighting function so it's not triggered if you overfly for example a supermarket, which generates an instantaneous almost infinite but very brief closure rate as the altitude AGL is immediately reduced by the height of the building. There are some approaches that overfly somehow steep raising terrain that could trigger a false "terrain" warning. The approach plates sometime warns about the possibility of false "terrain" warning in these cases. However, with the newer EGPWS, there is a more "fuzzy" logic. The GPWS doesn't just use radar altitude for this warning but also the 3D speed vector of the plane, the 3D position of the plane, and an internal 3D database that contains terrain, obstacles and runways. This system has two main advantages over the older "non E" GPWS: First, it provides a "look ahead" feature that eliminates the blind spot of the previous version, which would not detect flying towards a wall because it only has a "look down" capability. Second, it greatly reduces the false alarms.

We don't know yet what kind of GPWS was installed.