Colgan 3407, icing, and turboprops

Early information released by the US National Transportation Safety Board about the Colgan Air Bombardier Dash 8 Q400 that crashed on approach to Buffalo on 12 February made it look as if the agency had more or less decided that icing was the cause.

This is definitely no longer so. Icing now looks like just one factor – maybe not even the most important one - in a more complex series of events and influences. All the information available at present indicates that the icing that night was at worst moderate and perhaps only slight, neither of which would have been a problem for the Q400′s icing management systems if they were fully serviceable. They had certainly been selected.

Although circumstantial evidence (classic icing conditions prevailed, the stick-pusher kicked in, the aircraft pitched up rapidly), suggests icing-caused aircraft behaviour, there are still some alternative possibilities that are unresearched. These include the possibility of an aft centre of gravity exacerbated by other factors. The other factors could include the forward movement of the centre of lift on the wings caused by a combination of light ice contamination followed by the deployment of the flaps. Combinations like these cannot be ruled out until the investigation has had time to analyse the accident data they are still collecting.

For that reason I’m disappointed that the NTSB has handed the press a piece of information that has been interpreted by much of the media as meaning that the aircraft crashed because the aircraft was on autopilot when the crew should, in icing conditions, have been flying it manually. That is horrifyingly misleading at this stage.

One thing that this investigation will definitely NOT conclude, when the final report is published, is that the aircraft crashed because it was on autopilot when it went out of control.

Since the outcome of extensive research by the FAA on icing as it affects turboprops following the investigation into the October 1994 crash of an American Eagle ATR72 twin turboprop at Roselawn, Indiana, crews have been advised (not required) to trip out the autopilot in known icing conditions so they are more aware of abnormal or unusual control forces building up.

That is basically a sound piece of advice but it does not guarantee that, just because the pilots are manually controlling the aircraft, they would necessarily be able to prevent the departure from controlled flight of an aircraft whose de-icing systems are being overwhelmed by the prevailing conditions.

Let’s also just have a generic look at what icing protection turboprop aircraft have, especially those with only one or two engines. The first point is that it is not as comprehensive and powerful as the systems jets have, because turboprops are so efficient at converting fuel into power to drive the aircraft forward that they have a smaller proportion of engine power left over to generate electricity or provide hot bleed air – the two most commonly used forms of energy used for anti-icing.

The large surfaces that need icing protection on turboprops are the leading edges of the wings, tailplane, and fin. While smaller areas like engine intakes, propeller roots and spinners can be (and some are) anti-iced by electrical heating, achieving that over the large areas would be beyond the capability of the aircraft’s electrical generation system.

So the system most turboprops - including the Dash 8 and ATR series - use is the rubber de-icing “boot”. These black “boots” that wrap around the leading edges of all the lift-producing surfaces work by employing engine bleed air as economically as possible. The de-icing system provides pulses of pneumatic energy to cause the boots to inflate slightly, then deflate, then re-inflate and so on. This is intended to break off ice that has already formed on the leading edges.

Yes, you did read that correctly: the ice has to form first and then be broken off, and this is called de-icing, not anti-icing.

If the system is operated before ice starts to build up, there is a risk that, when buildup begins, the pulsed inflations of the boot will cause a gap to form between the ice buildup and the deflated boot, so when it re-inflates it has little effect on the ice that can quickly wrap around the entire leading edge to points beyond the boot itself.

So although the system works well if used at the right time, it can have reduced effectiveness if used at the wrong time. Knowing the difference between the right time and the wrong time is difficult because we are talking about weather here, and no two sets of conditions are ever the same.

In the end, turboprop pilots have to take entry into icing conditions even more seriously – in fact much more seriously - than jet pilots do. And one of the things they need to be able to make good decisions is good forecasts and actual weather reports. Avoiding icing conditions if you are unsure of their intensity is the best policy, but if you trust the information you are given and it turns out to be on the optimistic side of inaccurate, you are in trouble.

Icing is, perhaps above all, the weather condition that remains the most unpredictable and uncontrollable, and the airborne systems for dealing with it are still basically the same as they were in the 1950s.

22 Responses to Colgan 3407, icing, and turboprops

  1. Not a pilot but happened to read this today 16 February, 2009 at 7:43 pm #

    In an interesting coincidence, the NTSB calls ice-bridging a “myth” and advises use of the boots on entering icing conditions. Disagreement has been the result:

  2. Not a pilot but happened to read this today 16 February, 2009 at 7:45 pm #

    Correcting my post above, the NTSB did not use the word “myth.” That was the article’s interpretation of the statement that “Ice bridging is extremely rare, if it exists at all.”

  3. Airline pilot 17 February, 2009 at 4:11 am #


    You have old information.

    The FAA has proven and wants everyone to know that deice boot bridging is a myth.

    They [the FAA] have even issued an airworthiness directive requiring that all transport aircraft deicing system be operated the auto or cycling mode if so equipped while in icing conditions.

  4. Airline pilot/former test pilot 17 February, 2009 at 3:41 pm #


    Your story is incomplete. Boot use is advised at any time ice is suspected. Please note that tail ice may occur long before and much thicker than wing ice. This can lead to autopilot trimming to maximum, usually nose up. Any large elevator input to the down position can lead to tailplane stall resulting in total loss of pitch control. Extensive research was done by the FAA and NASA on this subject during 1992 and 1993. It was extensively discussed in Air Line Pilot of February 1993 titled “Tailplane Stall, the rime is one reason.

  5. David Learmount 17 February, 2009 at 4:44 pm #

    You’re absolutely right about the tailplane stall issue, but we don’t know for sure yet that this accident was purely about icing, let alone tailplane stall. However, Flightblogger has been digging up the research you have been talking about, and you can find it, plus NASA video, at—february-16—n.html

  6. Frederick 17 February, 2009 at 7:30 pm #


    The NASA video you are referring to is very interesting . It explains that tail stall on turboprops due to icing is often initiated by extending the flaps. The loss of control for the accident in Buffalo was at flaps extension. In addition, the tail produces a downlift that is required for stability to counter the pitch up moment created by the lift being applied aft of the center of gravity. Removing the tail lift would generate a large pitch up moment (like it was the case for flight 3407). As we can only speculate on the cause of the accident, to me it seems to lead to a tail stall issue. We will see what comes out of the investigation. Moreover, I think you are right that the NTSB should not gives that kind of bones to the media that are not able to properly interpret them. I heard a CNN guy stating that he couldn’t understand why this aircraft had problems with ice since it was built in Canada where people are used to severe weather conditions. It made me laugh.

  7. charleston 18 February, 2009 at 2:40 am #

    Perhaps it’s your lack of currency.

    You have old information.
    The FAA has proven and wants everyone to know that deice boot bridging is a myth. Additional research by NASA confirms same.

  8. David Learmount 18 February, 2009 at 9:48 am #

    You undoubtedly know that there are also those who believe that the FAA and the NTSB are wrong in their contention that bridging never happens, being concerned that if crews believe it’s possible it may discourage them from following procedure in icing conditions.

    No-one disputes that it’s rare, and that with today’s boots extending further aft from the leading edge than they used to that it’s more unlikely than it was.

    Meanwhile go back to the beginning of the blog and check that I said icing in this case has not been proven to be the culprit – there are other plausible reasons why this accident could have happened, and at this early stage these have not been examined as they surely will be.

  9. X - NTSB 18 February, 2009 at 4:28 pm #

    Pilot inexperience with a Dash 8 was the cause. He flew it like it was a SAAB, hence his actions caused the stall.

    That’s why the Dash 8 following 47 minutes on the same flightpath landed with no issues.

  10. David Learmount 18 February, 2009 at 4:48 pm #

    X-NTSB – While we can’t rule anything out, I don’t think anyone can, at this stage, reasonably be as confident as you are that this was a simple case of crew mishandling or misjudgement. This is early days, and I suspect there is much more yet to be revealed about the multiple circumstances surrounding this accident.

  11. Line Pilot 19 February, 2009 at 2:35 am #

    You make a good point, but are incorrect about the result of a tail stall. The tail provides a DOWNWARD lifting component (ie keeps the nose up). A tail stall would result in a rapid nose down pitching moment, but this could still shed some light on what may have happened. Transport category turbo prop flight crews have received a lot of information on the topic of inflight icing and tail plane stalls in the last 5 or 6 years (hence all the posts on the existence/myth of ice bridging), including the NASA video Dave posted above. One of the interesting points made in that video, which always stuck with me, was a tail plane icing scenario in which the ice formation leads to the low pressure (lifting) area on the bottom of the tail to move so far aft that it passes the elevator hinge point and actually deflects the elevator full nose down. They warn that the possibility exists that the control column could suddenly move to the full nose down position and require very high control pressure to pull the column back to a normal position.
    Consider the NTSB report in this case, in which the stick pusher activated shortly after flap extension, followed by an extreme pitch up. A pilot familiar with tail plane stalling characteristics could easily interpret a stick pusher activation (especially after flap extension) to be a tail ice induced elevator deflection and therefore react by pulling back hard on the control column. While this is the correct response for a tail plane stall, it would be disastrous if the situation turns out to be a wing stall. It would be very difficult to differentiate the stick pusher, from a tail ice inducted elevator defection.
    This is entirely speculation at this point but it would explain why the pilot flying may have pitched up to such an extreme attitude after encountering a stick pusher activation (an action people are already dismissing as pilot error or lack of experience).
    Any thoughts?

  12. Sergio 20 February, 2009 at 11:15 am #

    Honestly, with all these comments, I and every one else can get confused. So if someone knows:

    1) how to identify if it is a wing stall or tail stall
    2) which actions are required to solve the problem!

    Everyone knows a normal stall, identifying if it is normal or abnormal is the life-or-death problem, can you answer Frederick? I would really appreciate it!!

  13. Al 20 February, 2009 at 9:04 pm #

    Flight Safety has circulated a Nasa report done in 1993 where they examined ice-contaminated tailplane stall conditions.Nasa says that increased elevator buffet, elevator or pitch force lightening or reversal, reduction of tailplane lift and a reduction of pitch stability are all symptoms of tailplane icing.

    Recovery is that opposite for wing stall icing. Decrease power initially, pull up or at least level flight, and select flaps to previous lesser setting.
    This was taken from a Nasa study done in 1997. The report cautions that failure to correctly diagnose a wing from a tailplane ice situation on approach may prove fatal.

  14. Steve 21 February, 2009 at 1:31 pm #

    The main wing stall may exhibit buffeting felt through the airframe, roll off and/or pitch down. It generally occurs near the low end of the speed range for the airplane configuration. If equipped with a stick shaker, the shaker is designed to precede the onset of stall in order to provide stall warning. If equipped with a stick pusher, the pusher is designed to activate at or slightly prior to the actual stall in order to provide stall recognition. The pusher is generally required when the airplane’s normal stall recognition characteristics are not definitive.

    If ice has accreted, it is possible for a full stall of the contaminated wing to develop prior to either shaker of pusher functioning. Some systems have an ice bias applied to counter this; but the bias is based on one particular ice condition, and may not account for all.

    The tail stall may exhibit buffeting felt through the elevators only. There is normally no roll off directly associated with it. It generally occurs near the high end of the speed range for the airplane configuration and is almost always associated with a landing flap setting (full flaps). It’s effects may begin with a loss of longitudinal stability, a change in stick force gradient, ultimately an elevator “snatch” to the nose down position and finally a loss of lift from the horizontal stabilizer resulting in an extreme pitch over.

    The elevator behavior is most likely a result of a flow separation bubble extending over the elevator hinge line and altering the aerodynamic balance. The reduction in longitudinal stability may lead to a larger than normal pitch correction to maintain profile. The stick force gradient change may then lead to a much larger elevator deflection than the pilot intended, since the stick force resistance to deflection is now reduced or even reversed. All of this leads to a huge camber change for the worse, and the situation dramatically deteriorates.

    Tailplane issues with ice do not occur with “irreversible”, or fully powered, flight controls.

    The usual result of a tailplane stall is a rather vertical impact. In most cases, the landing flaps were selected fairly close to the ground, allowing little or no time for recovery. In some cases, the elevator stick force due to the “snatch” can be enormous.

    All of this should indicate that the Colgan accident bears none of the hallmarks of ice contaminated tailplane stall. However, the possibility of pilot misinterpretation of stick shaker followed by pusher remains. This was, in my opinion, a very likely cause of the United Express Jetstream 41 accident at Columbus. The similarity between the shaker/pusher sequence and the elevator buffet/elevator snatch sequence is significant.

    The recoveries are quite opposite. A normal stall would leave flap where it is, add full power and pitch down to regain the angle of attack. A tailplane stall requires a strong pitch up, the re-camber, the tail, along with flap retraction and careful power management (full power may exacerbate the problem). I believe this explains the Jetstream captain’s response to the shaker/pusher, which was partial flap retraction and mid-range power application while pulling the nose up. As it was actually a main wing stall, this was catastrophic.

    The applicability of this mis-perception by the crew in the Colgan accident cannot possibly be known yet.

  15. Frederick 24 February, 2009 at 3:08 pm #

    The latest info provided by the NTSB seems to lead to a pilot error. CVR shows that pilots discussed about ice accumulation on the windshield and the wing leading edge. It is known that ice builds up more on the tail leading edge, so when you see ice on the wing, there is usually more on the tail. The flight crew may have increased their workload because of that situation, which occurred in an already high workload situation that is landing configuration and preparation including localizer and glideslope capture.

    The pilot was probably very sensitive to every aircraft reactions in order to detect a stall because of icing and he probably already though of a possible tailplane stall. With that in mind, like Steve said, he probably misinterpreted the stick shaker/stick pusher sequence as it is a fact that he retracted the flaps at that moment, which is what would be the correct action in a tail stall situation, but not in a wing stall situation as Steve pointed out.

    To answer Sergio concerns, one way to detect which stall situation you are in, is to fly manually. The pilot would then be able to detect the buffeting and lightening of the yoke that precede a tail stall, but in the Colgan flight, the autopilot was engaged. However, having fly it manually may not had allowed the pilot to recover the aircraft in time, but could have helped him identify properly the issue.

    The stall shaker and pusher activation, although related to a wing stall, takes into account tail stall parameters. As the AOA increases, the airflow reaching the tail is more and more perturbed by the wing and engines starting to be in line between the incoming airflow in front of the aicraft and the tail. This is especially true on T-tail aircraft and therefore, the maximum AOA takes into account this parameter.

    Also note that the Dash-8 Q400 is equipped with a function allowing the crew to increase the stall speed by 20kts in icing conditions in order to cope with ice build up that could change the stall limits. However, this is a constant bias added that may not account for all foreseeable conditions. I think that the NTSB confirmed that the bias was set during flight 3407.

    Finally, I don’t like all the bad press that is out concerning the Dash-8 aircraft. Many people requested that all Dash-8 be grounded or at least the Colgan’s Dash-8 be grounded because it is not safe for flight in icing conditions. The NTSB reports that will be out in at least a year may conclude that it was a pilot error in a chain of events, but until then, Bombardier may suffer a lot from this bad press even if Dash-8 are operated by a lot Canadian airlines and also SAS (scandivian airline) and they face icing condition almost every winter day since more than 10 years without any incidents.

  16. Magnar Nordal 28 February, 2009 at 3:46 am #

    I am an ATR instructor and have flown this turboprop in winter condions for several years.
    First, ice bridging is not an issue for modern turboprops, since the de-icing boots are aligned along the wing leading edge and not horizontally. ATR procedure is to select the boots on at first indication of ice accretion and keep them on until the aircraft has left icing conditions.
    The Buffalo accident may have similarities with the ATR-72 accident in Roselawn in 1994, but I think NTSB should give the conclusions first. At Roselawn, the crew was not aware of the severity of the ice and used wrong procedures.
    Both ATR and Dash-8 have procedures for increased approach speeds in icing conditions. One thing that surprises me, is that the crew started to configure the aircraft for landing at a relatively low speed. In comparision, minimum flaps 0 speed for ATR-72 in icing condition is about 150-160 knots, depending on the weight. Could this have allowed ice to form behind the de-icing boots on the Q400?

  17. mark 2 March, 2009 at 8:11 pm #

    My union ( major US Airline ) recently circulated a memo regarding the BUF GS which if intercepted beyond certain parameters, could cause a 30 degree pitch UP ! ( A good reason not to use the AP during GS/LOC intercept if any…)

    Low speed + ice + autopilot potentially yanking a 15-30 degree bank or more at LOC intercept + a 30 degree pitch UP….?

    Too many things.


  18. Ed 3 January, 2010 at 6:38 am #

    Pilot but no airline experience and not at all familiar with Dash-8 Q-400.

    Based on preliminary report from NTSB:

    There was a marked decrease in airspeed immediately following deployment of the gear – flaps had been set only slightly, but were increased just before the stickshaker activated and the AP went off. Then there was a surge of power and a pull on the yoke resulting in a pitch-up and a full stall. Airspeed dropped as low as 30 kts.

    My first question: Wouldn’t it be likely in icy conditions that the gear would accumulate a substantial amount of ice immediately following deployment? Even though the main gear is located near the center of gravity, as ice accumulates wouldn’t drag increase and require some additional power to maintain required glidepath? The engines apparently stayed at idle even after flaps were moved from 5 to 10.

    Apparently, airspeed dropped markedly with little change in altitude, yet the pilot did not increase power. He did increase prop pitch but not throttle. Is that normal procedure or doesn’t it suggest that the pilot relied on the AP to maintain heading and simply failed to notice decrease in airspeed. Is it also possible that he forgot that his AP did not have auto-throttle capability?

    Second question: Has anyone observed a delay in response to throttle when calling for more power in a turboprop engine after it has remained at idle for several minutes or are the electrical heaters so effective that they keep compressors warm enough to give a pilot any torque he needs as soon as he asks for it?

  19. Black Sandals 5 February, 2010 at 1:39 am #

    I am thankful for the enormus , of course I’m really loving the blog, and I think this, as well as the important reviews some other guys have posted, will help you fix on if it’s the right pick for you.

  20. pumps shoes 10 February, 2010 at 2:31 am #

    A friend of mine talked about your blog, and now I am here. Aftera read a couple of your post, I must say that it’s really great.

  21. pumps shoes 10 February, 2010 at 3:21 am #

    Hopefully more people can learn from this post what I have. A great contribution to the world at large.

  22. Donnie Anno 19 July, 2010 at 8:15 pm #

    I am normally not the guy to write comments on people’s blogs, but for this post I just had to do it. I’ve been digging around your weblog an excellent nowadays and I’m extremely impressed, I imagine you can really emerge as the main voices for your market. Not certain what your load is like in life, but in the event you started devoting additional effort to posting on this website, I would guess you’d start seeing a bunch of traffic soon. With ads, it may become a sweet second revenue stream. Just an idea to ponder. Good luck!