The Real Reason Behind Regional TurboProp Icing Crashes

Considering prop rotational directions and the resulting asymmetry of lift and drag due to icing buildup under helical prop slipstream influence

Whilst initial focus upon this Colgan Air DHC-8 accident was upon the possibility that icing was involved, this proved not to be the case, and yet it kicked the whole issue of turboprop icing concerns back into prime time visibility - thus the following recapitulation of Regional Turboprop icing concerns is warranted.

BUFFALO, New York (USA) - In the minutes before a turboprop plane plunged to the earth killing all 49 people aboard and one person on the ground, the pilot and crew were recorded discussing "significant ice build-up" on the plane's windshield and the leading edge of the wings, federal investigators said today.

The "black boxes" recovered from the burning remains of Continental Express Flight 3407 also indicated that the de-icing button in the cockpit had been in the "on" position.

Shortly after that conversation, Capt. Marvin Renslow deployed the plane's landing gears and the wing flaps to slow down the plane in preparation for landing.

"Severe pitch and roll [began] within seconds" of the flaps being deployed, said Steven Chealander, spokesman for the National Transportation Safety Board. Chealander said that means the plane's nose bucked up and down while the wing's dipped and rose violently.

The plane plummeted to the earth so rapidly that Renslow and his crew never had time to radio a mayday alert.

Sudden Onsets - the question is why?

(Tail Icing? - or spanwise lateral icing asymmetry due to same direction prop rotation?)

main article at bottom (within this cell) is an extract from this link

IASA Comment:

Note that this editorial response (below) by Airline Safety.com's Editor Robert J. Boser reflects the bottom line arrived at by the NTSB's (post final report) Review of the Roselawn accident i.e. the first major event to draw global attention to the dangers of regional turboprop icing in rain-ice, freezing rain (or the euphemistic SLD -Supercooled Large Droplet). His text and the quoted extracts supports the NTSB party line that the French aircraft type certification was somehow remiss and that fundamentally, the formation of "runback" ice aft of the de-icer boots allowed a masking of the aileron function. The Editorial comment is useful for the observation that aft extensions of the de-icer boots were ineffectual as a rectification, that auto-trim compounded the pilots' sudden dilemma - and for faithfully depicting the very revealing onset of the incidents (i.e. the sudden rolls - see yellow highlit sections).

Bose says: "They couldn't be certain if it was a wing stall or a tail stall, but the most likely scenario was a tail stall, since the computer simulation showed a wing stall as producing only a 35-degree nose down pitch, whereas the tail stall would match the actual 50 to 60 degree down pitch." Apparently, nobody can get their minds around the possibility of left and right wings gradually developing dissimilar stall characteristics (and stall-speeds) due asymm icing - a condition that can lead to a sudden unrecoverable rolling wing drop (with an accompanying consequent severe nose-drop >35º).

However later aerodynamic testing and more recent accidents and incidents have tended to validate the alternative (and much more logical) premise i.e. that "spanwise lateral icing asymmetry due to same direction prop rotation" will ultimately lead to a pilot instinctively and continually opposing a (stalling) wing-drop with opposite aileron - thereby embedding the unequally contaminated L & R wings (and tail and fuselage) in an unrecoverable autorotative condition." The insidious nature of this unequal lateral build-up and spanwise distribution of ice is that the scenario can happen during cruise or gentle manoeuvre - as the higher iced-up (and increasing) stalling speeds coincide with the reducing cruise speeds. This coming together of speeds is due both to icing drag and the power decrement caused by using wing and engine anti-ice. Pilots never expect to stall during that normally uneventful cruise phase of flight. But once they are into that sudden flick-roll and nose-drop, whether or not they can recover is dependent upon whether the autopilot will click out, they remember to deselect it or whether the aircraft transitions into a spiral dive with speed rapidly increasing through VNe. In any of these cases, the fact that the autopilot has trimmed the aircraft INTO the stall will be working very much against them recovering from either a spin or spiral.

Unfortunately, bureaucracy being what it is, neither the NTSB nor the FAA, BEA etc can bring themselves to sign off on the real cause of turboprop icing accidents as being due to lateral asymmetry resulting in a spin entry when the more heavily iced side stalls first. Whilst there can be no argument against the obvious, that heavy icing will increase both wings' stall speeds significantly, no recognition is being given that when the stall occurs, either the autopilot or the pilot will inadvertently compound the lateral asymmetry by automatically (or reflexively) doing the one thing that a pilot should never do - i.e. attempt to pick up a dropping wing with aileron at the point of stall/incipient spin. That the sudden autorotative roll is caused by the more heavily iced side's wing stalling first is undeniable. However the cause is non-rectifiable (both props rotating same direction) - so the fallacy of ice-ridge masked ailerons lives on.... as a facile and socially acceptable cause of the turboprop icing "departure" phenomenon.

The evidence for ice-induced tail stalling remains less than sparse..... by comparison with that for prop "same direction" rotation induced icing asymmetry. But "modelling" can be (and indeed oft "must" be) made to support whatever you wish, or expect, the outcome to be.

IASA Safety

By Robert J. Boser (editor Airline Safety.com)

On Oct. 31st, 1994, an ATR-72-210, American Eagle flight 4184, crashed near Roselawn, Indiana, killing all 68 on board. The plane was in a holding pattern in icing conditions, when it suddenly rolled over and dove into the ground.

Ice, forming on the airframe and in particular aft of the de-icing boots---which are located on the leading edge of the wings---was the cause of the sudden rollover. The existing regulations, in respect to design certification of an airliner, were not adequate when it came to the ability of the plane to operate in light to moderate icing conditions.

Extensive airborne testing, following that accident, revealed it is possible for airliners to encounter water droplets exceeding 200 microns in average diameter, which is probably what happened to AMR 4184. The existing FAA design certification rules for Transport Category aircraft (Part 25, Appendix C), at that time, only required the plane to handle droplets with an average diameter of 20 to 50 microns.

The ATR was designed and manufactured in France by Aerospatiale. Mr. Fredrick's book, Unheeded Warning, reveals how political considerations overruled the safety mandate that is supposed to govern the FAA's design certification decisions on aircraft of foreign manufacture.

Fredrick details strong evidence that some experts in the FAA, knew the plane would be dangerous in icing conditions, but they were overruled by higher officials because the French might have been offended if certification was denied by the FAA. He also demonstrates that numerous "close-call" incidents and one accident (10-15-87, near Lake Como, Italy with no survivors), which preceded the Roselawn crash, were known to be the result of in-flight icing, but Aerospatiale effectively covered-up and did not circulate that information, lest the reputation of its ATR aircraft be damaged. The FAA also failed to follow up on the findings in the accident report of the Italian crash. If it had, the 68 that died at Roselawn might well be alive today.

On December 26, 1989, a United Express, flight 2415, suddenly pitched nose down, while on final approach to the Pasco, Washington Airport. The pilots, of that British Aerospace BA-3101 turboprop aircraft, were not able to recover and all died in the ensuing crash.

As in most accidents, there were multiple factors that combined until they culminated in a crash. At the previous enroute stop, the captain did not have the plane de-iced; instead crewmembers manually removed some of the accumulated ice from the wings and tail. There was also evidence that the deice distribution valve, which controlled the amount of air available to the de-icing boots on the leading edges of the airfoils, was partially defective and may not have allowed the boots to inflate to their full capability.

The NTSB concluded, in its final accident report, that during the descent into Pasco, the airplane accumulated from 1/2 to 1 inch of mixed rime and clear ice and that it most likely took the "mushroom" or "ram's horn" shape that is very detrimental to airflow over both wing and empennage (tail) airfoil surfaces. Since the plane did not carry any "black boxes," (cockpit voice or air data recorders), the NTSB could not be certain of what control actions the pilots took, while on that fatal approach. However, the Board was able to come to a conclusion, with aid of computer simulation of the sudden pitch-down (50 to 60 degrees nose down at impact), combined with the ATC radar tape (which recorded ground speed, course and glide slope tracks), that the plane stalled because of ice accumulation.

They couldn't be certain if it was a wing stall or a tail stall, but the most likely scenario was a tail stall, since the computer simulation showed a wing stall as producing only a 35-degree nose down pitch, whereas the tail stall would match the actual 50 to 60 degree down pitch. It matches the classic tail-stall profile that has been seen in other accident investigations over many years. It usually happens suddenly and without time for recovery (because they are too close to the ground on that stage of the final approach), when the flaps are extended to their maximum range.

On March 4, 1993, a Continental Express ATR-42 experienced a sudden roll, but the pilot was able to recover and make a safe landing. There were other such incidents, prior to the Roselawn crash, but they were passed off as turbulence encounters, even though icing conditions existed at the time. It wasn't until after the Roselawn accident that it became clear that those incidents too, had been near-disasters precipitated by the inability of that airfoil design to handle more than light icing conditions.

On Jan. 9th, 1997, an EMB-120 Brasilia turboprop, (Comair 3272) crashed near Detroit, killing all 29 on board.

The NTSB found the probable cause of that accident to be:

...the FAA's failure to establish adequate aircraft certification standards for flight in icing conditions, the FAA's failure to ensure that a Centro Tecnico Aeroespacial/FAA-approved procedure for the accident airplane's deice system operation was implemented by U.S.-based air carriers, and the FAA's failure to require the establishment of adequate minimum airspeeds for icing conditions, which led to the loss of control when the airplane accumulated a thin, rough accretion of ice on its lifting surfaces.

Contributing to the accident were the flightcrew's decision to operate in icing conditions near the lower margin of the operating airspeed envelope (with flaps retracted), and Comair's failure to establish and adequately disseminate unambiguous minimum airspeed values for flap configurations and for flight in icing conditions.

In March, 1998, a WestAir EMB-120 Brasilia, departed Sacramento and was immediately placed into a holding pattern in icing conditions. It was only in its second turn, in that holding pattern, when it suddenly rolled and dove towards the ground. The pilots immediately extended the flaps and were able to recover before they hit the ground. The pilots, in the Roselawn crash, also tried to extend the flaps, but the French computer, on that plane, would not allow flap extension because the computer determined the speed exceeded the limit for that degree of flap extension.

On March 19, 2001, a Comair EMB-120 Brasilia, experienced an upset event after encountering icing conditions. The aircraft, flying from Nassau, Bahamas, to Orlando, Florida, made an emergency diversion to West Palm Beach, Florida. A ground examination revealed substantial damage to the elevators and the horizontal stabilizer. Fortunately, the captain was able to recover from the dive and there were no injuries to the 28 onboard. Comments from the NTSB summary:

The crew reported that the airplane's systems, including its ice detection and anti ice systems, functioned normally before the upset.

They indicated that the airplane was being controlled by the autopilot at about 18,000 feet when they encountered instrument meteorological conditions that rapidly led to the windscreen being covered by a layer of ice.

The crew turned ice protection systems on and the ice on the windshield was cleared. The first officer observed ice on the right wing's boots and the right prop's spinner that extended farther back than he had previously experienced. The first officer switched the ice protection systems to their highest settings.

The first officer notified the captain of a decrease in airspeed from about 175 to 160 knots.

The captain disconnected the autopilot, applied power, and initiated nose down pitch inputs to arrest the airspeed loss. They indicated that these actions were unsuccessful and the speed further deteriorated to about 130 knots at which point the airplane experienced oscillations about its pitch, yaw, and roll axes and subsequently rolled sharply to the right and entered a steep descent.

During the descent, the electronic attitude display indicators in the airplane were observed to intermittently present no useful information. The captain stated that, "when we needed it [the electronic attitude display indicators] the most we didn't have it."

The airplane descended between cloud layers into visual conditions where recovery occurred about 10,000 feet.

After the recovery, no anomalies with the airplane, its handling characteristics, or its systems were noted.

The flight data recorder and the cockpit voice recorder have been examined at the Safety Board's laboratory. The cockpit voice recorder continued to run after the landing and did not provide any useful information regarding the upset. The accident airplane's solid state 25-hour FDR captured the event and functioned until power was removed on the ground. Preliminary review of the FDR data indicate the following sequence of events:

The airplane was at about 17,000 feet, with the airspeed stabilized around 200 knots indicated airspeed (kias). The autopilot was engaged.

The airspeed slowed from about 200 kias to 180, and the airplane began trimming nose-up. The airspeed continued to decrease to about 140 kias while trimming to a nearly full nose-up position.

The autopilot disconnected and the airplane rolled about 90 degrees to the left, and then back to near level. In the next 24 seconds, the airplane again rolled about 110 degrees to the left, back to level, then about 120 degrees to the right, back to level, and then rolled 360 degrees to the right, back to near wings level. Since the crew reported trouble with the flight attitude instruments, the roll angles recorded on the FDR are being further investigated.

The maximum nose down pitch attitude was 60 degrees, the maximum recorded airspeed was about 240 kias, and maximum vertical acceleration during recovery was about +3.6 g.

In April, 1996, following the Roselawn crash, the FAA issued 18 new airworthiness directives (ADs) affecting 29 models of turboprop aircraft. Those aircraft all have the same common features in their designs:

---- Unpowered flight controls.

---- Pneumatic deicing boots.

---- NACA five-digit sharp-stall airfoils (which were made obsolete by the more modern soft-stall designs).

The ADs require extensive instruction, to pilots flying the affected aircraft, on how to fly in freezing rain and drizzle (including the prohibition of the use of the autopilot in icing conditions), how to recognize indications of severe icing, and then require an immediate exit from icing areas. In addition, both ATR-42 and ATR-72 aircraft had their de-icing boots modified to extend the boot area to reach back to 12.5 % of the chord. Previously, they had extended only to 5 % and 7 %, respectively. In theory, that should solve the problem of the tendency of ice ridge formation at the 9% chord position of those obsolete sharp-stall airfoils.

However, it still doesn't deal with the results of the Bascombe-Downs tests, conducted by the British, which demonstrated ice could form as far back on the wing as 23% of the chord, and on the tail at 30% of chord. Both percentages remain well beyond the limits of the deicing boots. Those tests limited the size of the droplets to 40 microns, near the maximum limit of the archaic FAA design certification rules for Transport Category aircraft (Part 25, Appendix C), still in effect at that time of the Roselawn crash.

That is why I still believe that some, if not all, turboprop airliners still have a serious problem in regards to tail icing. To the best of my knowledge, nothing has been done to make tail de-icing more effective.

STEPHEN A. FREDRICK, the author of Unheeded Warning, came very close to "buying the farm," when he was flying the ATR in icing conditions. After the Roselawn crash, his conscience compelled him to work surreptitiously to expose the deficiencies of the ATR aircraft, in icing conditions, and the history of how those deficiencies were known and covered up by the French manufacturer, the FAA and airline officials. Fredrick is a rare person of genuine conscience. He was willing to give up his job as an airline pilot (actually -- his entire career, since he cannot hope to ever be hired again by any airline), to expose the facts about the ATR.

My recommendation to the flying public is the same as that of the courageous American Eagle pilots who wrote and distributed an anonymous pamphlet at Chicago's O'Hare airport, after the Roselawn crash (Fredrick was one of those pilots):

"...If the weather is clear this winter, sit back and relax because this is a good aircraft. If the weather is cloudy, snowy, or cool and rainy, think about alternate transportation methods..."

August, 1998, revised June, 2000 and September, 2002

Robert J. Boser
Editor-in-Chief
AirlineSafety.Com

Re the revelations above, two theories are worth pursuing (but they're not mutually exclusive):

The asymm icing scenario (as induced spanwise by same-direction prop rotation on each engine

a. firstly, that the flap extension suddenly exacerbated the differential between the left and right wing's icing asymmetry (i.e. by rapidly changing each wing's stalling angle-of-attack - and the L&R tailplane's also). Think and concede "effective wing leading-edge boot de-icing" but, in freezing rain, areas further back on each wing's chord would contain varying (and significant) thicknesses of spanwise-ridged ice accretions on particularly the upper (but also on lower) wing surfaces. At lower (approach) speeds the difference between the L&R wing's elevated stalling speeds (at the higher AoA) would be even more critical. With flap extension, those ice ridges would have an increased asymmetry effect upon the power of a differential frieze aileron to impose (or oppose) roll. Additionally (and more lethally) however, there'd be an increased likelihood that a large aileron deflection (turning onto the localizer) would be more likely to stall the upgoing wing and induce autorotation. The slipstream effect of added engine power at lower speed (due to gear/flap extension drag) would also accentuate the differing lift and drag coefficients between the two wings (and across the horizontal stabilizer).

The TailPlane Stall Scenario

b. Secondly, in respect of tailplane stall due to ice build-up on the undersurface of the horizontal stabilizer and elevator (whether asymmetric or not). Note that Q400 does not have a variable incidence tailplane:

The problem is that wing and tailplane stalls have opposite recovery procedures. Recovering from a tailplane stall entails full aft yoke, raising flaps and decreasing airspeed. Tough situation to find yourself in, particularly at low speeds.

Unfortunately, these recovery procedures are generally not part of airline training. Recognizing instantly that these wholly different recovery measures are required? Another whole new ball-game." It's like the 737 pilot's reaction to the rudder hardover in USAIR 427. If he'd eased off the rudder (and pedalled it L/R), that may have undone the jammed PCU's reversal. But instinctive sustained inputs always rule the day when instant inputs are required to resolve underway doomsday scenarios.

c. It's also noteworthy that the Q400 isn't difficult to load correctly but it does have a pretty narrow CofG range (despite what the Bombardier website says) and it's quite trim sensitive due to its length. Both spin and spiral instability are adversely affected by an aft CofG. An aft CofG will promote a flattish spin. It's worth noting that an aft CofG, even if within CofG limits, can become an adverse condition once tailplane/elevator authority has been partly compromised (and even excessive elevator trim travel caused?) by icing (on the tailplane and elsewhere).

So, to restate the case for an asymmetric icing cause for Control Loss: "Sudden unexpected autorotative roll during a turn, at well above the normal stall speed, is highly likely to generate an instinctive pilot reaction of opposite ("held-in") aileron. It's a well-known fact that use of aileron to re-instate wings level at the stall will more deeply embed the aircraft in the autorotative condition. This is why, even for wing drop in a wings level stall, the only solution (to prevent further wing-drop whilst pushing the yoke fwd to unstall the wings) is the secondary effect of rudder (i.e. roll) to prevent further wing-drop." That use of rudder to stop roll has never been instinctive in an unexpected sudden wing-drop/uncommanded roll scenario. In fact it got a bad rep for being the instigator of the A300 fin detachment in AA587 at Belle Harbor, Queens. If you're unaware that the wing-drop is the result of one wing stalling and you maintain a desperate wing-levelling aileron input (in an attempt to stop that roll), the result will be a flat spin. Use of increased engine power will also flatten the spin's pitch attitude.

Preferred Theory? If a tail stalls on any aircraft, the nose will drop, but the wings do not stall. The aircraft will continue to have considerable forward

movement. I suppose it might theoretically even continue to nose over until inverted - or, more likely, nose-drop ONLY UNTIL the higher IAS restores tailplane effectiveness - i.e. unstalls it as the speed increase lessens its AoA. Neither outcome or development was evident in this accident. The stick-shaker and stick-pusher actuation suggests a simple wing-stall. It's hard enough to stop a wingdrop in a clean, 1g, full back-stick induced stall; but give each wing different aerodynamic characteristics and superimpose the pilot's natural reaction (of immediate opposite aileron) and you've got pro-spin controls (even though he went nowhere near the full backstick that is characteristic of stall training - so he's not thinking "stall/spin" nor contemplating the use of rudder to prevent further roll).

Moreover, the FDR, the accident scene, and eyewitness accounts, do suggest a classic stall/spin scenario. Witnesses say that the aircraft's impact attitude was flat with roll and yaw - the classic flat spin impact attitude resulting from autorotation with a sustained aileron input and moderate to full power.

There's been ample precedent amongst turboprops for this asymmetric icing scenario resulting in the stall speed between L&R wings differing significantly (as induced spanwise by same-direction prop rotation). Such a large stall-speed differential between port and starboard wings is conducive to a very rapid snap-roll entry into an unrecoverable spin. Pilots, whether on autopilot or not will be totally caught out and will instinctively try to oppose the roll with aileron. That input will merely EMBED the dropping wing more deeply into a stalled condition. Application of power will just flatten the spin. Spins entered on autopilot will have the added penalty of having been trimmed (pitch and roll) deeply into the condition - and this will really complicate the already impossible task of attempted recovery.

Roselawn ATR / Aero Commander in NZ / ATR freighter in the Taiwan Straits /United Express, flight 2415 (BA-3101 turboprop)/ March 4, 1993, a Continental Express ATR-42/ EMB-120 Brasilia turboprop, (Comair 3272)/ a number of MU-2 accidents etc etc (see third cell down, in the table below).

Colgan 3407 update: Aircraft appears to have landed in flat attitude

By John Croft

Federal investigators say the Colgan Air Bombardier Q400 that crashed on approach to the Buffalo Niagara International Airport Thursday night landed in a relatively flat attitude, despite eyewitness accounts that suggested a nose-dive.

The aircraft was also oriented in the opposite direction of the instrument approach to Runway 23 at the airport.

Colgan 3407, enroute to Buffalo from Newark as a Continental Express flight, crashed into a house in a neighborhood about 5mi from the airport after the crew experienced violent pitch and roll excursions after deploying the first increment of flaps (15 degrees) in preparation for landing, information gained from the flight data recorder.

NTSB board member Steven Chealander, speaking to reporters today from Buffalo, says further review of the cockpit voice recorder (CVR) also shows that stick shaker and stick pusher, devices that attempt to prevent pilots from entering an aerodynamic (i.e. wing) stall, had activated after the upset.

Chealander yesterday reported that initial review of the CVR showed pilots discussing "significant" ice build-up on the wings and windshield.

That evidence, combined with the aircraft attitude, may suggest the aircraft remained in a stalled state from upset at approximately 701m (2,300ft) altitude to the crash, a result that is not uncommon in fatal turboprop accidents.

An April 2006 report by the Flight Safety Foundation of stall recovery events in turboprops reveals that in three accidents that killed 134, pilots did not initially reduce the angle-of-attack on the aircraft by moving the control column to the nose-down position early in the upset sequence.

The pilots of Colgan 3407 had extended the landing gear 20 seconds before deploying the flaps, and had attempted to retract both during the upset that followed the flap extension. All 49 onboard perished in the accident as well as one person in the house.

The picture below, taken by the Buffalo News, shows the relative orientation of the aircraft at the crash site.

Chealander says investigators are reviewing maintenance records at Colgan’s base in Virginia and that earlier reports that the aircraft was delayed out of Newark because of a mechanical problem were false. High winds at Newark had delayed the flight however. The aircraft’s Pratt & Whitney Canada PW150 engines appear to have been generating torque at the time of the accident, investigators say.

The crew did not discuss any caution lights related to the pneumatic deicing boots on the wing leading edges as well as horizontal and vertical stabilizer leading edges, says Chealander, indicating that the system, as well as systems for the engine inlet cowlings and propellers, appears to have been operating properly.

A cursory look at the US Federal Aviation Administration’s service difficulty report (SDR) database for the Q400 model reveals at least three instances, all in 2002, where Q400 wing leading edge pneumatic deicing systems failed to work properly due to a faulty dual distributing valve built by Aerazur, a subsidiary of Zodiac. In each case the caution lights were illuminated however. All Nippon airways had earlier reported a high removal rate for the distributing valves, in most cases because the caution light had illuminated. Aerazur in 2007 switched parts to address the problem.

Colgan, a subsidiary of Pinnacle Airlines, purchased N200WQ nine months ago.

Another aspect of investigating a roll upset icing accident should consider the possible effect of ice asymmetry and ice protection coverage. Propellers impart a swirling motion to the air that results in an asymmetric ice formation on the wing leading edges immediately downstream of the propellers. The downgoing blade side will cause more ice to accumulate farther aft on the upper surface and less further aft on the underside on that side. Conversely, the upgoing blade side will tend to cause more ice to accumulate farther aft on the lower surface and less on the upper surface. Apply this theory to each wing and then factor in the significance of the left and right props rotating in the same direction. The spanwise ice distribution will be lopsided to say the least. The Left and Right wings' upper and lower surfaces will end up with significantly different profiles and stalling speeds and characteristics. There will also be additional disymmetry in accretions on the port (versus starboard) fuselage and tail surfaces.

The computational fluid dynamics analysis derived graphic below visualizes this effect clearly on the horizontal surface. (from Newmerical Technologies International)

The blue area in the image below shows the area of the DHC-8- Q400 wing immersed in the propwash. It is approximately 43% of the wing area and represents a substantial percentage of the wing lift - so asymmetry due to ice formation in this area is worthy of careful consideration.

In certain icing conditions the helical flow field may result in not only asymmetry of ice formation, but may also form asymmetrically aft of the ice protection system coverage. When the engine power is reduced, the velocity of the propwash and the beneficial effects of the propeller flow will be reduced and the local angle of attack behind the propeller can increase suddenly. If the ice has caused a serious erosion of the normal safety margins, the result can be a lateral imbalance of lift and can contribute to a roll. In some cases the wing stall can spread across the wing and the ailerons changing the control characteristics of the airplane.

The picture below (from FAA FFFSCR report) was taken in an icing test behind an aerial tanker spraying water from a large “shower head” to determine the formation of ice. The water contains sea marker dye that appears yellow in the solid (ice) phase and fluorescent green in the liquid phase. Some green is visible on the viewer’s right at the very leading edge. Note the ice formation aft of the active area of the deicing boots on the viewers left (airplane’s right wing). This ice cannot be removed by the ice protection system. The DHC-8-Q400 propellers both rotate in the opposite direction so the effect would tend to be the mirror image of this.

If icing is found to be a causal factor, it's not known how the manufacturer addressed this issue on the accident airplane design and if, and/or how much, of a factor this may be.