Near Crash From Icing Prompts Call for Better Stall Warning

The twin-turboprop cleared the ground by about 112 feet as the pilots recovered from the stall. They then climbed to a safe altitude and a few moments later completed an uneventful landing.

The incident underscores the continuing hazard of in-flight icing, the need for aircrews to be extremely conscious of the danger posed by exposure to icing conditions, and the need for greater progress in the industry overall to improve training and equipment for coping with in-flight icing.

As the saying goes, a turboprop aircraft flying in heavy icing conditions can become an icicle enroute to becoming a hailstone.

The incident involved an icing-induced stall – actually two stalls in rapid succession – of a July 28, 2002, Hazelton Airlines flight from Sydney to Bathurst, Australia. The early night flight of the Saab 340 with two pilots and 30 passengers came within a hairsbreadth of ending in disaster. Indeed, the case might not ever have come to light had not a passenger, advised of the seriousness of the situation by the passenger and experienced pilot sitting next to him, later submitted a confidential report of the frightening event to the Australian Transport Safety Bureau (ATSB). This account galvanized the agency into action. The ATSB’s report of investigation was issued Jan. 28. The event was classified as a serious incident, defined as one where an accident nearly occurred.

The pilot-in-command (PIC) of the flight reported a “wing drop” on the approach into Bathurst to the chief pilot the next morning but attributed it to turbulence, not icing. The chief pilot asked for a written report, which the PIC submitted four days later. The matter would have gone no further without the passenger’s letter to the authorities.

What emerges from the ATSB report is the need for improved preparedness for icing encounters. In-flight icing continues to catch aircrews by surprise, and basic design and warning systems may not be adequate to alert crews to impending danger.

Details of the case read like numerous other accounts of in-flight icing incidents involving twin-turboprop aircraft:

The setup. The weather forecasted by the Bureau of Meteorology (BOM) for the Sydney to Bathurst flight noted freezing temperatures above 4,000 feet and moderate icing in clouds above the freezing level. Snow showers and broken clouds down to 800 ft. were forecast for the arrival at Bathurst. In other words, the weather report available before the flight put the crew on notice.

A specialist area BOM report, which the ATSB received some weeks after the incident, indicated that the clouds contained large supercooled droplets of water. As such, the conditions were conducive to clear ice accretion on the airframe during the descent.

The descent. With the PIC flying, descent was begun at 12,000 feet. To slow the aircraft, power was retarded to about 17 percent. The PIC recalled to ATSB investigators that the airplane passed through clouds a number of times and that ice accretion was noted on the windshield wiper, but not on the wings. Engine anti-icing had been turned on, but propeller de-icing and wing boot de-icers were not activated.

The autopilot was on and in the altitude (ALT) mode to capture the selected minimum descent altitude (MDA). The airline’s standard operating procedure (SOP) encouraged use of the autopilot, although restricting its use in icing conditions to the IAS (indicated airspeed) mode. This policy was intended to prevent penetrating stall speed margins. Also, the SOP called for the half bank mode (13.5� instead of the full-bank mode of 27�) for approaches. In this case, the autopilot was in the full bank mode.

The airplane leveled off at the 3,800-foot MDA. However, with the low power setting and added drag of ice accretion, airspeed was below the minimum for operations in icing conditions. The trim system was automatically adding nose-up trim to compensate as the autopilot attempted to hold altitude.

At a speed of 133 knots and a radio altitude of 1,264 feet, the PIC commanded an autopilot turn to the right (which was in the full bank mode, recall).

The first stall. The aircraft rolled 28� right. As speed continued to decrease, the copilot called “speed,” and the PIC added power. But not enough. With engine power at 38 percent and 109 knots airspeed, and with the pre-stall buffet reported by passenger Kenny, the airplane pitched into an uncommanded left bank of 109� and 7� nose-down pitch.

The stick shaker did not activate. The reason why will be dealt with momentarily.

The second stall. The pilot responded with a roll command to the right. The airplane stalled again, rolling through wings level to 56� right wing down. At an altitude just a few hundred feet above the ground, maximum nose-down pitch of 27� was achieved. When angle of attack (AOA) increased from 12� to 16.2� during this second roll, the autopilot snapped off.

At one point, the airplane was descending at a rate of nearly 5,000-ft. per minute.

The recovery. Applying full power, the PIC pulled out of the dive a scant 112 feet above ground level at an airspeed of 165 knots. Landing gear and flaps were retracted, a missed approach was called, and the aircraft subsequently landed at Bathurst.

Related factors. The ATSB identified a number of elements contributing to this incident. We focus on the three most prominent:

Crew conditioning. The captain said the stall warning system always worked in training. By implication, he was expecting a warning to sound in the event of an approach to stall. The crew said they did not detect any ice buildup on the wings. However, they did observe ice accumulating on the windshield wipers. Supercooled droplets can be the most threatening, leading to ice that slaps and sticks on the aircraft. It does not take much ice buildup to lead to a big loss of lift, as much as 80 percent in some cases.

Muddle in the manuals. Three manuals are relevant: Hazelton’s SOP, its aircraft operations manual (AOM), and the manufacturer’s airplane flight manual (AFM). The ATSB found that “all three manuals contained sections dealing with flight during icing conditions, but these sections were scattered through [them].” Some of the material contradicted the AFM to boot.

For example, the company SOP instructed aircrews to activate the boots when ice accumulated to an estimated thickness of 10 mm (about fourth tenths of an inch). The aerodynamic behavior of the incident aircraft corresponded to about 12 mm (nearly 1/2 inch) of horn-shaped ice on the protected surfaces.

The AOM advised crews to activate boots when 6 mm of ice accumulated on the wing leading edges (about 1/4 inch). Yet Volume 2 of the same manual instructed crews to activate the boots at the first sign of ice formation anywhere on the aircraft (which in this case would have been the observed ice buildup on the windshield wipers, since the pilots said they did not see ice buildup on the wings. The situation was in darkness and clear ice would have been hard to detect anyway.

The manufacturer’s AFM, which takes precedence over the SOP and the AOM, called for use of the boots in known icing conditions. A Saab official said, “The ATSB report supports our previous conclusion that if proper published procedures were followed, this incident would not have occurred.”

“Saab’s manuals state that at the first sign of icing anywhere on the aircraft, the de-ice boots should be activated. Had the crew performed this action and yet still allowed speed to decay towards minimums, they would have received a warning prior to stall,” the Saab official maintained.

Stall warnings. Remember that in icing conditions the stall margin is decreased, and therefore angle of attack in icing conditions must be restricted to a smaller range in order to avoid stalling the airplane.

To explain why the PIC did not receive a stall warning, a brief system description is in order. According to the ATSB report, the stick shaker activates at 12.5� AOA (including autopilot disconnect).

In the case of the incident aircraft, which was configured with 20� flaps at the time, stick shaker was programmed to activate at 13.1� AOA.

Stick pusher will activate when one of the two AOA sensors is at 19� and the other is at 12.5�. The stick pusher forces the control column forward to drop the nose, thereby reducing angle of attack to recover from the incipient stall.

The AFM cautions that a stall may occur before the stick shaker activates, with an attendant roll of more than 90� and excessive nose drop. How is this possible? After all, when the wing de-ice boots are inflated, the AOA signal is reduced by 0.4� to increase the stall margin by 1-2 knots. However, the incident airplane stalled at 9.5� AOA and, besides, the wing boots had not been activated, even though the AFM called for boot activation in icing conditions.

As the ATSB noted succinctly:

“The SOP did not require the activation of the de-ice boots in known icing conditions as per the manufacturer’s documentation.”

“The SOP allowed for the use of the autopilot during icing conditions which masked tactile cues regarding the increasing nose up attitude.”

Of note, as a consequence of its investigation into the fatal icing-related 1997 crash of a Comair EMB-120 twin-turboprop, where masking of ice-buildup occurred, the National Transportation Safety Board (NTSB) strongly recommended that the autopilot be switched off when anti-ice systems are activated (see ASW, Aug. 31, 1998 and Sept. 7, 1998). This recommendation was debated vigorously by experts at an FAA-sponsored icing conference in 1999 (see ASW, Feb. 8, 1999).

Some experts maintained that if there is some area of flight where the crew should not use the autopilot, that information should be provided in the AFM. Others urge cautious consideration before issuing a mandate to turn the autopilot off and hand-flying the airplane to provide a better “feel” for its response in icing conditions. Unless the pilot flying has experienced the tactile cues of ice accretion and approach to stall, he won’t know what it feels like. Simulator training does not replicate the conditions of altered control response and buffet. Secondly, the autopilot is considered an aid in high workload conditions. Periodically disconnecting the autopilot while flying in icing conditions to check the handling characteristics with subtle control inputs, and then re-engaging the autopilot, may be an alternate – provided there is some means in pilot training to demonstrate the feel of the airplane in this abnormal and untested flight regime.

Most icing events such as the one at Bathurst are part of a pattern. The ATSB noted a number of similar cases involving the Saab 340 – in the United States, in Europe, the United Kingdom, New Zealand, and a Nov. 11, 1998, Saab 340 event at Eildon Weir, Australia. As a result of ice accretion, the airplane in this 1998 incident stalled, rolled 126� into a 35� pitch down attitude, and it took the crew fully 10 seconds to recover the aircraft. The resulting investigation into this Kendall Airlines incident at Eildon Weir produced a number of recommendations, some of which five years later have not been resolved to the ATSB’s satisfaction. The NTSB has faced similar challenges getting some of its recommendations coming out of the Comair crash investigation implemented.

One of the more significant ATSB recommendations dealt with a requirement imposed on the Saab 340 for operators in Canada. Transport Canada (TC) required the installation of an ICE SPEED function on the Saab 340. It provides an added input to the stall warning computers. Basically, when the crew activates the ICE SPEED switch, the triggering thresholds for stall warning are lowered (i.e., in AOA terms, reflecting the increased stalling airspeed of a contaminated wing). For example, the stick shaker will activate at 5.9� AOA at zero flaps, as opposed to the standard 12.5� AOA. The stall identification (e.g., stick pusher function) activates at 11� AOA regardless of flap setting. Saab advises that this optional ice speed modification is available to any customer.

The Saab 340s in Canada are operating with this added protective feature, according to the ATSB. This is the second instance observed recently where Canadian regulatory authorities have taken a more aggressive stance than counterpart agencies such as the U.S. Federal Aviation Administration (FAA). Recall that the FAA did not emulate TC’s requirement that engine magnetic chip detectors for Pilatus PC-12 aircraft should display warnings for all phases of flight. The FAA allows the warnings to be disabled in flight (see ASW, Jan. 19).

Saab does not believe the ice speed mode would have been of much help in the Bathurst incident. After all, the crew had not activated the wing de-icing boots, so it may be fair to presume the pilots would not have activated the ice speed mode, either. The modification also imposes some drawbacks, notably in higher landing speeds to avoid triggering warnings. Saab believes the drawbacks outweigh the benefits.

As for the Hazelton Airlines incident at Bathurst, the airline reported that simulator sessions in icing conditions are now a formal item, and the inconsistencies between the SOP, AOM and AFM were rectified with a revised SOP.

As for the Civil Aviation Safety Authority (CASA) in Australia, the ATSB urged it take “appropriate action” regarding the stall without warning that occurred in the Bathurst flight. And to manufacturer Saab, the ATSB urged, “as a matter of priority,” that the stall warning system be modified to alert crews of an impending stall when flying in icing conditions.

These recommendations notwithstanding, the ATSB findings in the Bathurst incident reflect a global problem, as evidenced by the similarity of NTSB recommendations from its investigations of icing-related crashes, and as evidenced by the fatal crash Dec. 21, 2002 of a Trans Asia Airways ATR-72 in the Taiwan Strait, where the crew realized too late the great peril of the icing conditions in which they were flying. In that case, too, the crew was flying with the autopilot on (see ASW, Nov. 10, 2003).

The ATSB report raises two seminal questions: Are aircrews sufficiently alert to the danger of ice accumulation, and are aircraft being fitted with protective systems to provide crews with sufficient warning time to escape an icing-induced stall? After all, traffic alert and collision avoidance (TCAS) is designed to alert crews to a potential mid- air collision in plenty of time to take action, and enhanced ground proximity warning systems (EGPWS) are designed to provide timely alerts of impending crashes into terrain. Yet for icing, the ATSB is obviously concerned that airplanes can stall on unwary crews without warning.

In other words, the current risk of progressing from an icicle to a hailstone is unacceptably high.

(ASW note: The full ATSB report may be viewed at www.atsb.gov.au/aviation/pdf/vh-olm.pdf)

Stall Without Warning

The human challenge: “The investigation was unable to determine if the flight crew later realized the seriousness of the incident.” Supporting findings:

• “The flight crew had activated the engine ice protection systems, however, they did not activate the wing de-ice system as they perceived that there was no ice present on the wings.”
• “The flight crew were operating the autopilot in full-bank mode. The manufacturer does not recommend the use of this mode in icing conditions.”
• “The flight crew did not notice a significant increase in pitch attitude or decrease in speed until just prior to the stall.”

The machine challenge: “The situation where it is possible for an aircraft to stall prior to a stall warning, indicates that procedures and/or technical safeguards, at the very least, need to be in place to prevent flight crews from being exposed to such a dangerous situation.” Supporting findings:

• “The Saab 340 aircraft is capable of stalling with an ice deposit on the wings, without providing aircrew with any artificial warning of the impending stall.”
• “There is a modified stall warning system fitted to Canadian registered Saab 340 aircraft and this different stall warning system (when activated) provides an increased warning margin to the stall, when operating in icing conditions.”
• “The aircraft was not fitted with an audible trim warning to alert crews when the autopilot was trimming the aircraft [into a stall], nor was this required by certification. The only visual indication to crews of trim movement is the movement of an index on the trim position indicator, which is not located in the flight crew’s primary field of vision.” (ASW note: Saab maintains that its 340 model airplane “meets all current certification requirements for operation in icing conditions.”) Source: ATSB