Wx Watch: Deadly droplets

Discovering large-droplet icing--the hard way

November 1, 2012


I was going over some weather documents the other day and came across an icing accident that will go down in history as one of the most notorious. It was a true landmark accident, and it occurred at 3:59 p.m. CST on October 31, 1994, at Roselawn, Indiana ( “Safety Pilot Landmark Accident: An Inconvenient Departure,” January 2007 AOPA Pilot).

The fallout from that accident still has repercussions today, and still gives us many valuable lessons. But enough time has passed that many pilots aware of it at the time have forgotten many details, and another generation of pilots may never have heard of it.

For an in-depth look at what has come to be known as “the Roselawn crash,” I’d refer you to the NTSB official accident report. Just Google “Roselawn crash” and select the appropriate links. The identification numbers for the reports are NTSB/AAR-96/01 and NTSB/AAR-96/02. That’s right, there are two reports. More on that later.

To be exceedingly concise, here is what happened. American Eagle Flight 4184—an ATR 72—held for 24 to 32 minutes on its way into Chicago’s O’Hare airport. The weather was IMC and this was causing delays. At the time of the crash Chicago was 1,100 feet broken, visibility two miles in rain and fog. The ATR began the hold at 10,000 feet and soon would be cleared to descend to 8,000 feet, still in the holding pattern. And the whole time the airplane was picking up ice—so much ice that it was building in ridges aft of the deice boots. As ice built, lift decreased and the airplane’s autopilot began trimming nose up in order to maintain altitude in the hold.

The crew noticed the high deck angle and lowered the flaps 15 degrees—the first notch. This lowered the nose—for the moment. At one point the crew was talking to a flight attendant. Shortly thereafter the captain left the flight deck to go to the lavatory.

During the descent to 8,000 feet the airplane accelerated past 186 KIAS, which triggered the aural flap overspeed warning. So the flaps were retracted. As the flaps retracted to the zero position the angle of attack increased and the ailerons deflected to the right. Then the autopilot automatically disconnected.

Things quickly got worse. The ATR rolled 77 degrees to the right, but the pilots corrected for this. Then came another, more violent rolling moment, with the ship rolling back to the right and the angle of attack increasing. The roll rate reached 50 degrees per second, the airplane’s G-load hit plus 3.6 Gs, and airspeed shot up to 375 KIAS as the pilots applied nose-up elevator forces. A total of 68 people were killed by the subsequent impact.

Probable cause

The NTSB determined that the probable cause of the accident was a loss of control caused by an unexpected aileron hinge moment reversal that occurred after a ridge of ice accreted aft of the wing deice boots. This reversal happened because ATR failed to adequately educate pilots about the effects of flying this type of airplane in icing conditions. More blame went to the Direction Générale de L’Aviation Civile (DGAC)—the French counterpart to the FAA—for inadequate oversight of the ATR 72, and its failure to provide the FAA with airworthiness information developed from previous ATR incidents and accidents in icing conditions. It seems that there were several prior icing incidents, known by their locations: the Mosinee, Newark, and Burlington incidents in the United States, and the Ryanair and Air Mauritius incidents overseas.

Contributing factors included the FAA’s failure to ensure that flight into known icing (FIKI) certification accounted for flight into freezing rain and “other icing conditions not specified in the Code of Federal Regulations (CFR) Part 25, Appendix C.” In order to earn FIKI certification, an airplane must prove it can fly in the droplet sizes, liquid water contents, and temperatures set out in Appendix C.

The DGAC took issue with the “inadequate oversight” charge, and the French Bureau d’Enquete Accidents (BEA—the French counterpart to the NTSB) published its own findings. This is the second volume of the Roselawn accident report. The DGAC did provide information about the ATR’s behavior in ice, the BEA report said. It also said that the accident occurred as a result of “prolonged operation in freezing drizzle/rain conditions well beyond the certification envelope for all aircraft.” Also, the pilots were inattentive, used an unapproved flap setting for the hold, knew they were in icing conditions, and could have asked for a clearance to exit the area. Those were just a few of the BEA’s exculpatory statements. A back-and-forth debate ensued over the years, and it was not until September 2002 that the NTSB softened its critique of the DGAC.


The Roselawn crash, at a terrible cost, advanced our icing savvy. Everyone knew that ice accretions could cause premature stalls and other flight anomalies, but aileron hinge moment reversals? For most of us this was the first mention of the problem—in which disturbed air flowing over the wing in effect sucks ailerons into the up position, causing uncommanded rolls.

After this accident, a new form of icing—supercooled large droplet (SLD) icing—first came to mention outside the research community. Turns out that SLD, with its 140-micron-plus droplet diameters, exists way outside the certification envelopes’ maximum of 50 microns. No airplane had ever been tested under those conditions, so there is no way to determine how any airplane—even one with FIKI approval—would behave. SLD icing, like clear icing and freezing rain, causes supercooled precipitation to run back well aft of leading edges. This runback also causes distinctive ridges to form aft of deice boots and on propeller hubs, windshields, and wiper blades.

By the way, to date the icing envelopes remain the same, in spite of efforts to expand them to include SLD.

For pilots, the principal lesson remains the same: avoid icing conditions in the first place, and exit icing conditions as soon as possible at the first sign of ice. Climb, descend, reverse course, or steer a heading to cloud-free skies—anything to escape.

And don’t use an autopilot in icing conditions; it can mask the warning signs of an impending loss of control.


The Roselawn crash spurred a burst of activity in the icing research community, especially at the NASA Glenn Research Center in Cleveland. Flight testing behind icing tankers duplicated SLD and the result was a modification of the ATR 72’s boots, extending their coverage farther aft on the wing chord.

But the damage was done. You don’t see many ATR 72s plying the skies over the northern United States these days. Most have migrated to warmer climates.

Finally, a poll taken by the FAA’s John P. Dow Sr. reported on the experiences of 23 flight crews operating in large-droplet icing. It turns out that most encounters took place when total air temperatures were between zero and plus-7 Celsius. (Total air temperature is higher than static, outside air temperature because it reflects the heating caused by friction and compressibility effects at airspeeds above approximately 180 KIAS).

So the dangerous temperatures are right around zero Celsius—something to remember this winter when flying in instrument meteorological conditions. Most pilots in the study also reported a mix of rime and clear ice, with most icing encounters happening in the cruise and descent phases of flight.

As for escape maneuvers, 70 percent made descents to get out of ice. In three cases, however, icing became more severe after taking action to escape.

Lest you think these findings apply only to regional airliners, remember this: these airplanes spend a lot of time grinding around below 20,000 feet—sometimes well below. In other words, they do a lot of instrument flying in the altitudes that piston-powered general aviation pilots fly. We share the same airspace, as well as the same weather hazards.

Email tom.horne@aopa.org.