I HAVE ADDED A VERY IMPORTANT UPDATE AT THE BOTTOM OF THIS!
If you have not previously read this post, please read it as is, then continue to the update which follows it.
If you have previously read this post, feel free to skip to the bottom to read the update.
The NTSB usually makes their initial report within a week of a crash. Their final follows a year later.
We posted our preliminary last week, and here’s our latest. Our first post’s focus was on chain of events issues based on my CRM [crew resource management] training and philosophy. CRM teaches us that accidents generally are caused by multiple interrelated factors forming a chain that leads to disaster. No one ”link” by itself would have caused such a crash, but joined together they make the crash almost inevitable.
Even where we find pilot error, we avoid assessing pilot blame. Although we must “calls ‘em as we sees ‘em,” we have to ask why? – endlessly why? – if we want to prevent a repetition. In quality improvement, we call this process root cause analysis – with the approach that you can’t eliminate the symptoms without eliminating the disease and/or the conditions that bred it. In other words, the root cause is the cause you can fix which will eliminate a recurrence of a similar accident.
In this approach, usually – not always – pilot error results from management error. Even if we find the crew were sleazy, drunk, incompetent, etc.,. – which in this case they were NOT – we’d have to ask how were they hired, trained, scheduled, evaluated, and coached? Those are all under management’s control, not the crew’s. Therefore, fixing human performance issues entails examining any management error which contributed to the pilot’s error and modifying it appropriately.
This accident’s “chain of events,” updated a week later, includes: [Note: This section is still being revised.]
- overscheduling [12 hr plus hour duty day (not confirmed) + 2 hour delay = 14 hour duty day] = pilot fatigue
- minimally experienced captain [less than 100 hrs tt in type and limited previous flying experience in the N.E. in winter]
- icing
- deicing system turned on 10 minutes after takeoff – never turned off
- autopilot use in icing up to stall onset
- airspeed 45 K below recommended for that phase of flight [no gear or flaps, outside outer marker]
- incomplete readbacks to ATC and “fuzzy” rushed radio transmissions
- jumpseat captain on board [seat location unknown]
- radar vectors by ATC to turn final close in to marker
- previously reported glideslope anomalies
- localizer anomaly [suspected by ATC immediately after crash and confirmed by DL 1998, the aircraft following 3407]
.
But meanwhile, based on Jerry Zremski’s Buffalo News article on 2/20/09, there may be machine [design, maintenance, or performance] issues that could be root causes of this accident. Therefore, we’ve decided to give the benefit of the doubt to the crew, and added 2 Machine theories to the huMan error theory. No doubt they overlap in some areas.
Let’s start with 4 assumptions:
1. Weather-related items like “icing” cannot be given as a root cause per ICAO standards. "Failure to handle the weather" by huMan or Machine can be, but weather items like icing, for example, are considered a given in commercial aviation; you shouldn’t be out in it if the huMan or Machine is not certified / trained / designed to be capable of handling it. If “caught” in such a situation, you should have exited it as quickly as possible. Thunderstorms provide a clearer example. During World War II, the US military tried to find a way to safely penetrate and fly through significant thunderstorms. They were unable to, and to this day pilots are forbidden to attempt it. Therefore weather radar systems aboard airliners are for thunderstorm avoidance, not penetration.
2. A stall in this type aircraft under the conditions 3407 was experiencing:
- less than 1500’ AGL
- night
- IMC
- turbulence [forecast]
- icing
- and 49 souls aboard
is unrecoverable; therefore we’re not considering events after stall onset to be relevant. Stall avoidance is what would have counted here; failure to do that doomed the flight. [See the email from my colleague Captain Bud (below) for more information.]
3. The most significant pre-stall event per the FDR information releases so far is airspeed readout of 134K – just prior to flap and landing gear extension. Based on information from the FAA type certificate DHC-8-400 series, PW150A:
Airspeed limits:
VFE (Flaps extended)
Flap 5 - 200 kts
Flap 10 - 181 kts
Flap 15 - 172 kts
Flap 35 - 158 kts
[posted by 'little alex" on www.jetphotos.net]
as well as other Colgan Q400 captains and my own experience flying props in a high density regional airline operation, the appropriate airspeed for that phase of flight [no flaps or gear; outside the marker] is 180 K [give or take 15 – it really doesn’t matter]. The readout speed is about 45 K too low, making the subsequent events virtually inevitable. We reject out of hand the reasoning that somehow the crew wanted to be at 134 instead of 180 – no matter how fatigued or marginally experienced, there's no way professional pilots fly 45 K below where they should be, if only because they, too, want to get to home as quickly as possible and do it in one piece. When you're doing props in the soup on a 14-16 hour day -- and I have, lots -- you may miss small things and/or your decision making may not be the greatest. But you don't forget to keep the plane flying.
4. Lest there be doubt about this last assertion, although the captain’s experience in type and in Northeast winter flying was marginal at best, he had already flown as a captain on a smaller Houston-based jet prop for a year. While he may not have been a great pilot, there’s nothing to indicate he was irresponsible or incompetent in any way. According to the Buffalo News, Captain Renslow had worked his way up, was a dedicated husband and employee, passed his checkride, had never flown intoxicated, and was not a member of a weird death-seeking cult. So he at least knew that flying 45 K below the appropriate speed is not good, for all the reasons I explained. We’d have no problem adding "dumb, poorly selected and trained idiot" to the huMan category [1] above, if we thought it fit, but we don't.
We’ve discussed the first major cause types possible – huMan. The only other possible major cause is Machine.
In other words, what Machine causes[s] would have put this aircraft 45 K below where it should have been?
We only see 2 possibilities here, both related to machine system design flaw or malfunction, as follows:
2A. Instrumentation failed to inform the crew how slow they were.
The crew could have been unaware of how slow they were due to incorrect information from the airspeed indicator[s]. There is a history of prior problems with this system, as evidenced from the airworthiness directive [A/D] note the FAA issued regarding it in 2006.
[Thank you to “AA7772ER,” who posted it on www.jetphotos.net].
http://www.airweb.faa.gov/Regulatory_and_Guidance_Library/rgAD.nsf/0/34c4557f2807655c862572450052f0ba/$FILE/2006-25-17.pdf
Even if this problem has now been corrected, it’s possible that where’s there smoke there’s fire – another design flaw has presented itself, or the previous fix was not effective, the A/D not accomplished timely, or – if accomplished –not accomplished properly. Have you ever had the mechanic not fix your car correctly on the first try? Thus, obviously, the crew would only realize how slow they were when the stick shaker and nose pusher activated – and by then the die was cast.
and/or
2B. The machine itself could not be prevented from slowing rapidly to an inappropriate / hazardous speed, and it could not regain a safe speed rapidly enough to forestall this accident.
On a normal descent into an airport like BUF, an aircraft like the Q400 would slow to 250 K below 10,000’ to comply with the speed limit mandated in the FAR’s. On a high performance jet prop transport like the Q400 the crew would likely reduce power to idle or perhaps a little above. [The Colgan Q400’s are not equipped with auto throttles, which adjust thrust to maintain a pre selected airspeed.]. The crew would want the airspeed to be under 200 K [the maximum speed for flap extension] once they leveled off close to the marker at 2300’ MSL like 3407 was doing so they could begin their approach with initial flap extension. Normally an aircraft descending at 250 K with low or no power is going to slow below 200 K pretty rapidly once it levels off. Therefore, the crew may choose to feed some power in slightly before level off, but they may not want to if they’re being vectored to intercept their final approach course close in to the outer marker. Either way, once the flaps start down at 180 –190, say, – followed by landing gear extension – the Q400 would slow down in a hurry, as long as it’s level outside the marker. Therefore its pilots would be ready to feed in power as required to hold altitude and at least 160 K until passing the marker inbound. [160 is desirable because it provides a margin of safety and stability, especially if there may be ice and/or turbulence possible. And it also is about the slowest speed the larger turbojets fly to the marker; slowing below their slowest speed means delaying them, risking inadequate spacing, or being taken off the approach by ATC and sent to the back of the line. [I've been there / done that – only once thank you – that was quite enough!]
Given that the crew knew all this, what could have slowed them down to 134 K so quickly they had no time to recognize, react, and recover from it? I emphasize that at 134 if you’re still slowing rapidly, you could be to stall speed in a heartbeat – and then it’s over. Again, the crew had to know this.
But what if something was acting like a big speed brake slowing 3407 down much more rapidly than normal, and when they leveled off -- especially if they hadn't led the level off with power [intentionally or not] – it really brought them to a halt.
What could this have been?
Posted by jehm on www.flightaware.com
The attached picture shows ice accumulated on the wing of a DHC-6 Twin Otter [which I flew] which preceded the Dash 8 - Q400. This picture raises the obvious question: What about the dictum that you can’t fly with that much ice? This aircraft seems to be flying just fine. But the DHC-6 is a STOL aircraft with fixed gear and a short stubby wing – made to produce a lot of lift, fly slow, and carry a lot. What if that wing were replaced with a long, thin high performance wing built for speed – like the Q400 wing? And what if the de-icing boots on the wing’s leading edge were working fine – so the aircraft was flying fine, since the essential aerodynamic surfaces were clear? And what if, while the leading edge were clear and the sensors were reassuring the crew that all was well, ice remained further back on the wing?
There’d be no way to know it was there in a high wing aircraft like the Q400 -- but there it be, acting like a huge speed brake. And when the Q400 leveled off, with power reduced or at idle to enable that tight turn in, what if this “speed brake” really kicked in and brought everything to a crashing halt – aggravated, of course, as flaps and gear were extended? Extending flaps and gear, chasing an errant localizer and/or glideslope, autopilot continuing to trim nose up [see below] would only enhance an already unredeemingly fatal situation. It's possible that had the crew known that such was the situation, they would have kept power and speed up to descend a few hundred feet lower on the glideslope – and once below the freezing level [which was just about at 2300’ MSL], all their ice would have melted prior to landing, as it did for US Airways 1452 15 minutes later [See ATC transcript.] On the other hand, it may have been too late, no matter how much power were added. Only wind tunnel tests and more information from the FDR and CVR can confirm or deny this theory. Again, Captain Renslow could not have known about this “speed brake” by looking or via the sensors – he could only have suspected it by experience. And while experience like I had as an understudy for some of the best pilots in the business certainly would have helped, it may not have been enough in this case.
So – putting the huMan and Machine chain links together, we can construct a scenario of the final moments of Colgan flight 4307:
The aircraft was had slowed to 134 K prematurely, either because the crew was unaware of it or was unable to prevent it or recover from it.
Thus, the following events ensured – per the FDR information released so far:
Initial flap extension – perhaps to add lift to compensate for the low airspeed. This could cause the nose to pitch up if only a bit. And given the aircraft’s predicament – low airspeed, possible wing contamination – the flaps may have caused the aircraft’s angle of attack to increase a little – perhaps not enough for the crew to notice immediately, given that they were still in the clouds without outside visual cues.
But next, gear extension, combined with the slight pitch up further slowed the 4307.
[Normally flaps and landing gear extension are done by the non flying pilot, hopefully not until commanded by the p/f. But what if in this case that convention was not observed, further surprising the p/f? We cannot confirm or deny this possibility without obtaining the CVR data.]
Meanwhile, we know the crew had previously advanced stall response speed [stick shaker and nose pusher] to occur 20 K above normal. If stall speed on the Q400 is around 105, and by then they had slowed another 9 K below the 134 last reported, they’d be right at 125 K – and 105 + 20 = 125.
Bingo!
As stick shaker and nose pusher engage, the autopilot clicks off. There are several theories that as the wing loading had increased due to icing and as the aircraft had just leveled off – possibly without a major power increase, the auto pilot would have gradually increased nose up trim, trying to maintain altitude. It’s possible that while the autopilot was banking ever more sharply to intercept / make the tight turn required for close in localizer interception or to chase an unreliable, unstable localizer, it further increased nose up trim – to an extreme degree. And thus, when the autopilot clicked off due to stick shaker activation, the nose pitched up to 30 degrees [with or without crew control input], resulting in an accelerated stall. Initial bank may increase [45 degrees left per FDR] on the side the aircraft is turning toward [in this case – left, perhaps to finish joining the localizer,] followed by sharp wing drop off to the opposite side [100 degree bank right reported.]
This last scenario may or may not accurately describe what kind of stall 3407 was in at this point – it may have been a combined wing and tail stall; it may not have been an accelerated stall, it may have been a mild stall aggravated by or not by ice. It doesn’t matter – the stall rapidly became a spin, likely a flat spin. We know this because we know at this point the aircraft had a very high descent rate [-10.000 fpm] and a very low airspeed [100 K, less than its stall speed]. Thus, it was no longer flying, it was dropping.
There is no recovery from such a drop, beginning no more than 1500’ above the ground [1500’ / 10,000 fpm = .15 minute = 9 seconds maximum recognition and recovery time.]
So, we reach journey’s end. Barring substantial new NTSB disclosures, we have constructed the most likely scenario combining the huMan chain links and the Machine possibilities about 1 week after the crash. NTSB will take a year to develop theirs.
And they should. Hopefully, theories like ours will be tested and confirmed or denied. We can't go any further without more FDR information, the CVR raw information, a wind tunnel, the wreckage, access to other pilots, management, and ATC people, the aircraft’s A/D and maintenance logs, the pilots’ schedules and training records, etc., etc.
Obviously, the NTSB has all these items and more. That's why we’re done now, and their work is just beginning. Hopefully they'll view ours as a good start.
Captain DT
(c) 2009, airline-crash-analysis.com
IMPORTANT UPDATE, POSTED ON 2/24/09
This is a summary; I hope to expand this in the next few days.
As I reflect upon the post I made yesterday, I am troubled by one question: How could Captain Renslow not have realized he was dangerously slow for that phase of flight and not have acted accordingly by increasing power rapidly? Why would he have lowered flaps and extended gear, thus compounding the airspeed problem, if he had thought he was close to a stall?
A malfunctioning ASI – as I explained yesterday – does not explain this. For if it were indeed malfunctioning, the ASI would likely read low, not high, alerting Captain Renslow to the problem more vividly instead of masking it.[Thanks to Falconer's comment at jetphotos.net for reminding me of this.] Moreover, as the conscientious and well trained pilot he was, Captain Renslow likely was aware of the A/D note and the problems the Q400’s airspeed systems had had in the past, so he was likely all the more on the lookout for them.
Therein lies the explanation: The ASI was functioning properly, and the perilously low airspeed was staring Capt. Renslow in the face! But because he was aware that that ASI system had had problems, he chose to ignore it! Quite possibly, he thought that just as his windscreen had iced over, moisture had contaminated or ice had blocked the pitot-static system -- which provides raw data for the ASI.
Thus, it’s not that the ASI failed to warn him. It’s that because ice on the wing’s surface [where there is no deicing system or sensor] is not visible from the cabin, and the leading edge and its deicing boots – which he could see – looked free of ice, the de-ice system appeared to be operating properly [which it was]. Thus, based on his knowledge and his perception, there was no reason for the ASI to indicate such a low airspeed -- unless it was malfunctioning once again. And if it had been, his best course of action may well have been to continue the approach another minute or 2 until the moisture cleared and the ASI returned to normal. {Been there / done that -- see below.]
Unfortunately, somewhat like the deicing system on the ATR-42 [American Eagle, Roselawn, In], the Q400’s de-icing system’s design did not allow it to clear as much ice off the wing as necessary. Capt. Renslow may have compounded this design flaw by remaining in the ice for almost an hour, but again, there was no way he could have seen evidence of that.
This is not the first time a pilot has chosen to ignore what his instruments are telling him and compounded an already perilous situation. Indeed, often -- ignoring seemingly inaccurate instrument indications is advisable. [See my personal note, below].
Unfortunately, without assessing blame, we have to at least admit the possibility that -– possibly for the best of reasons –- Captain Renslow ignored the low airspeed reading he was getting [134 K] and chose to assume that his Q400 was still flying at the appropriate and safe airspeed [180 K] for that phase of the flight.
There is no other good explanation for his actions.
A personal note: I recall landing a single engine Cessna 172 on a dark, short runway in a strong gusty wind at night after a long trip home from my grandmother’s funeral in Buffalo. As I turned from base leg to final, I heard a thump, and my ASI stopped working. It was the old mechanical type, and the dial read 0. Initiate stall recovery or go around? Heck no! - I knew that a certain pitch attitude at a certain power setting would give a certain airspeed. I may have lowered my nose a little and/or added some power to increase my margin [sound familiar?], but I knew I would be fine and kept up with the landing. I was ignoring my airspeed indicator, wasn't I, because I knew better? In that case, it was the right thing to do - especially because unlike 3407's it was totally inoperative and I was in visual contact with the ground and runway. Also I had felt that thump on the wing when this had all begun. So I had lots of good reasons to ignore the defunct ASI -- what choice did I have? ;-]
As I was tying down, the reason for the ASI’s malfunction became apparent - I had speared a bird with the pitot tube [ASI probe].
Captain DT
(c) 2009, airline-crash-analysis.com
