Ever since the crash of American Eagle Flight 4184 (an ATR 72) in Roselawn, Ind., on Oct. 31, 1994, the NTSB has been recommending that the FAA enact a new rule that the Board believes might have prevented these accidents. As a result of the crash of Flight 4184, the NTSB recommended that the FAA “prohibit the use of the autopilot” during encounters with icing conditions. The idea behind this recommendation is that the autopilot could mask a change in flight control feel during icing conditions and that pilots might be better able to detect the effects of icing on flight controls if they were hand-flying rather than using the autopilot.

In accidents where the NTSB makes this recommendation and others, the airplane stalled and crashed after the autopilot suddenly switched off after exceeding pitch and/or bank limits, leaving the pilots in a situation where they had to take control of an airplane in an attitude from which it was difficult to recover. There are also some incidents where pilots lost control after icing encounters with the autopilot engaged.

Five noteworthy accidents where the autopilot suddenly relinquished control during icing conditions have killed 162 people (see box on facing page), and there have been other accidents under the same circumstances. Yet the FAA has never considered making the NTSB’s recommendation to turn off the autopilot in icing conditions a general regulation, and there doesn’t appear to be any such rulemaking on the horizon.

The FAA, it should be noted, has imposed the no-autopilot-in-icing requirement on Cessna’s Caravan single-engine turboprop, but via an Airworthiness Directive applicable to that specific airplane (AD 2006-06-06, March 2006). This raises the question, if the AD-based requirement is a good idea for the Caravan, why not for other aircraft types?

The FAA is well aware of autopilot and icing issues because in AD 2000-09-15, the FAA mandated that Mitsubishi MU-2 operators install an autopilot disconnect system “to prevent the autopilot from applying nose-up elevator control with pitch trim until the airplane stalls.” This is intended, the agency noted, to force the pilot “to take control of the airplane before the autopilot applies pitch control in the full trimmed nose-up direction. The automatic autopilot disconnect would, prior to a stall condition, give control to the pilot and allow detection and prevention of a stall.” Again, one might ask the question: why is this important for the MU-2 but not other airplanes, including the ATR 72 and Brasilia?

Following the Monroe, Mich., crash of a Comair Brasilia in January 1997, the NTSB recommended that “the FAA require all operators of turbopropeller-driven air carrier airplanes to require pilots to disengage the autopilot and fly the airplane manually when they activate the anti-ice systems.” The Safety Board has repeated this recommendation often, but interestingly, while the NTSB’s list of most-wanted transportation safety improvements includes an icing category, banning use of the autopilot in icing conditions is not listed as one of the desired improvements. This is surprising in light of more recent autopilot-in-icing accidents such as the February 2005 Circuit City Citation 560 accident in Pueblo, Colo., and the Colgan Air Dash 8-Q400 crash near Buffalo, N.Y. in February.

The final report on the Colgan accident is not complete, but in its report on the Circuit City accident the NTSB did not highlight use of the autopilot in icing conditions as a significant factor. It mentioned only that the pilots failed to note deteriorating airspeed before the airplane stalled.

The FAA’s advisory circular for pilots flying in icing conditions–91-74–has been cancelled and there is no current circular available from the FAA. However, a copy of 91-74 contains this advice, so clearly the agency is on board with the idea of turning off the autopilot while flying in icing conditions:

“Care should be exercised when using an autopilot in icing conditions, whether in cruise or other phases of flight. When the autopilot is engaged, it can mask changes in handling characteristics due to aerodynamic effects of icing that would be detected by the pilot if the airplane were being hand flown. In an aircraft that relies on aerodynamic balance for trim, the autopilot may mask control anomalies that would otherwise be detected at an early stage. If the aircraft has non-boosted controls, a situation may develop in which autopilot servo-control power is exceeded, the autopilot disconnects abruptly, and the pilot is suddenly confronted by an unexpected control deflection.

“Pilots may consider periodically disengaging the autopilot and hand flying the airplane when operating in icing conditions. If this is not desirable because of cockpit workload levels, pilots should monitor the autopilot closely for abnormal trim, trim rate or airplane attitude.”

That last bit of advice is important, because as an airline pilot told AIN, writing a rule that mandates turning the autopilot off in icing conditions could have unintended consequences.

There are times when flying in icing conditions that the workload is extremely high, he noted, and, after all, the autopilot is there to help ease the workload, so why not use it? But when doing so in icing conditions, the pilots must closely monitor the state of the airplane.