The weather conditions at the time of the accident and the similarity of this accident to others involving aircraft with ice-contaminated wings caused the Board to examine, in detail, the possibility that the accident was the result of ice accretion. The Board's analysis determined that the performance of the aircraft was consistent with the known effects of wing icing. The theoretical performance analysis conducted by the Board determined that a reduction in lift production and increase in drag were necessary to produce the performance of the aircraft observed during the accident take-off. Furthermore, the Board determined that the aircraft stalled at an airspeed above the stall speed calculated for the applicable weight and configuration.

As demonstrated in previous research and by previous accidents, seemingly insignificant amounts of ice can be sufficient to significantly degrade an aircraft's performance and flight characteristics. This performance degradation is the result of reduced lift production and increased drag. Of particular significance is the increase in stall speed and decrease in stall angle of attack caused by changes in the leading edge shape of the wing and surface roughness. The Board believes that the failure of the aircraft to accelerate following lift-off, its failure to achieve a sustained climb, and the stall at a higher than normal airspeed exemplify the known effects of ice-contaminated wings.

Calculations performed by the Board during its analysis determined that the increase in drag and decrease in lift production were consistent with that demonstrated to occur with wing surface roughness elements of about 0.03 inches or an amount of leading edge ice contamination with equivalent effects.

The fact that the aircraft did initially achieve a climb and continued to accelerate for a very brief period after rotation could be attributed to the enhanced aerodynamic efficiency (increased lift and reduced drag) provided in ground effect. However, as the aircraft climbed away, the benefit of ground effect would have been quickly reduced. As the aircraft crossed the departure end of the runway, ground effect would have been lost because of the rough, downsloping terrain. Analysis of the aircraft flight profile indicated that the aircraft entered stall buffet and stalled soon after it crossed the departure end of the runway.

The conclusions based on the computer simulations conducted by UDRI and the simulator tests conducted by the Board were consistent with those of the theoretical analysis. Both demonstrated that lift and drag values consistent with ice accretion on the aircraft wings duplicated the takeoff performance of the aircraft.

The performance simulations conducted by DND, on behalf of the Board, also confirmed that the performance of the aircraft was consistent with that which results from ice-contaminated wings. Although the simulations were limited in that the aircraft pitch history of the brief flight and inputs by the pilots were not known and thus could not be considered, there was close similarity between the observed performance of the aircraft and simulations of take-offs with the wings contaminated with surface roughness elements of 0.04 inches or an amount of leading edge icing with equivalent effects, or with wings contaminated with surface roughness elements of 0.02 inches, or an amount of leading edge ice with equivalent effects, compounded by the loss of thrust from one engine. Furthermore, the simulations demonstrated that, with ice contamina- tion present, aircraft take-off performance is very sensitive to small changes in aircraft pitch and airspeed. The differences between a successful take-off and an unsuccessful take-off were only one degree and two to three knots respectively.

The simulator tests showed that it was possible to complete a take-off successfully with CL and CD values consistent with ice-contaminated wings. However, to be successful, it was necessary to use significantly lower than normal pitch angles during rotation and initial climb in order to maintain the angle of attack below the lower than usual angle of attack at which a stall would occur. Such an action would require advanced knowledge of the degraded performance. In this regard, the simulator tests confirmed the sensitivity of aircraft performance to changes in aircraft pitch demonstrated in the computer performance simulations.

The precise amount, type, and location of any ice adhering to the surfaces of the aircraft during the take-off could not be determined. Nevertheless, based on the prevailing weather conditions, the Board believes that some ice would have accreted on the leading edge of the wing. Under the prevailing conditions for the aircraft's approach to Gander, it was calculated that the most probable maximum amount of ice accretion on the leading edge of the wing would vary from about 8.7 millimetres (0.34 inches) at 85 per cent span, through 6.5 millimetres (0.26 inches) at 53 per cent span, to 5.0 millimetres (0.20 inches) at 26 per cent span. This accretion would represent a full span, narrow ridge, or disturbance on the leading edge of the wing with the greatest accretion on the outboard section of the wing. This calculation did not include any ice that would have accumulated below cloud in the approximate one and one-half minutes of additional flight to touchdown. In view of the freezing precipitation occurring when the aircraft landed, it is prob- able that additional ice would have accreted on the leading edge during the approach, although the quantity could not be calculated.

The calculated ice accretion was consistent with the pilot reports made by the captain of the Boeing 737 which departed Gander about 45 minutes after MF1285R had landed and the pilot of the PA-31 which landed just after the accident. During his brief climb through the same cloud layer, the 737 captain reported moderate icing. He estimated that approximately one-quarter inch of ice accumulated on the centre post of the windscreen. The PA-31 pilot reported icing on approach sufficient to significantly obscure visibility through the cockpit windshield.

With the exception of the one refueller who reported seeing ice on the edge of the windscreen, none of the ground service personnel who assisted in servicing the aircraft reported observing ice on the aircraft. However, the Board notes that most of these personnel were not in position to observe, at close range, the aircraft wings. Further, in their interviews with CASB investigators, those personnel who did approach the wings of the aircraft reported that they did not specifically inspect the aircraft for ice and that ice may have been present. In considering the lack of witness reports of ice on the aircraft wing, the Board also notes that the leading edge of the wing is between approximately 10 and 16 feet above ground and that the station stop was made during the hours of darkness. Both of these factors would have made it difficult to detect small amounts of glaze ice on the leading edge, particularly on the outboard sections of the wing, when no specific effort was being made to look for ice. None of the ground service personnel were in position to observe any ice contamination that may have been on the upper surface of the wings.

The quantity of ice which would have accreted on the leading edge of the wing would be dependent on the use of wing ice protection. Although the Board cannot conclude with absolute certainty that ice protection was not used during the approach, normal industry practice suggests that it would not be usual for the crew to employ ice protection for such a brief descent through cloud. Pilots who were interviewed from Arrow Air and other operators concurred that it would be unusual for airframe ice protection to be used on approach in the prevailing circumstances.

As a result, the Board considers it likely that ice was present on the leading edge of the wings when the aircraft landed at Gander. The greatest quantity of leading edge ice would have been on the outboard section of the wings. The approach and landing at Gander would have been completed without incident because they were flown at angles of attack below those used for take-off and because of the aerodynamic benefits of ground effect experienced during the landing flare.

Data provided by Douglas Aircraft enabled the Board to estimate the decrease in coefficient of lift maximum which would result from the calculated leading edge accretion amounts. As seen in Figure 1.16., the per cent reduction of maximum lift coefficient which results from a localized, spanwise disturbance or narrow band of roughness located at the leading edge is a function of the roughness height divided by chord length.

At 85 per cent semi-span, the chord length is 125.5 inches, thus the 0.34-inch calculated accretion divided by the chord length is 0.00271, which, according to the Douglas data, results in a maximum lift coefficient reduction of about 27 per cent.

At 53 per cent semi-span, the chord length is 226.4 inches, thus the 0.26-inch calculated accretion divided by the chord length is 0.00115, which, according to the Douglas data, results in a maximum lift coefficient reduction of about 23 per cent.

At 26 per cent semi-span, the chord length is 312.9 inches, thus the 0.20-inch calculated accretion divided by the chord length is 0.00064, which, according to the Douglas data, results in a maximum lift coefficient reduction of about 18 percent.

From Figure 1.16. it can be seen that the reduction of maximum lift coefficient determined at the 85, 53, and 26 per cent semi-spans equates to full upper surface contamination with roughness elements of 0.052 inches, 0.033 inches, and 0.022 inches respectively.

The weather conditions during the technical stop at Gander were conducive to the accumulation of additional ice on the wings of the aircraft. Freezing precipitation in the form of very light freezing drizzle and snow grains was reported between 0900 and 0945. At 0930, the observer noted freezing drizzle and snow grains adhering to the accretion indicator. He described the precipitation as a thin, rough layer, covering less than 30 per cent of the indicator's surface. After 0945, no further freezing drizzle was noted; however, snow grains continued to be observed on the indicator until after the accident. The time of the aircraft's landing at Gander corresponded closely with the 0900 surface observation taken by the weather observer. Thus, the Board believes that the type and quantity of ice which accumulated on the aircraft would be closely reflected by the freezing precipitation observed on the ice accretion indicators at 0930, 0945, and 1000.

Based on these observations, the Board concludes that the upper surface of the wings would have been roughened by the cumulative effects of the freezing drizzle and snow grains. The texture of the precipitation which adhered to the indicators was further described by the meteorological observer as resembling medium grit sandpaper. This description is often used in the research documentation to describe the magnitude of roughness necessary to significantly degrade an aircraft's performance and flight characteristics.

In addition, it is considered possible that some frost may have formed on the upper surface of the wing as a result of interaction between the cold wing surface and the near saturated atmosphere. Although the amount of frost that may have formed is not considered large, it could have resulted in further roughening of the upper wine surface.

The Board concludes that the combination of leading edge ice, which accreted during the approach, and upper surface roughening, which occurred during the station stop, was probably sufficient to result in aircraft performance degradation equivalent to that which occurs with the entire wing upper surface roughened with roughness elements of between 0.03 and 0.04 inches.

The flight engineer was observed to conduct a visual inspection of some portions of the aircraft while at Gander. It is not known if he observed any ice on the wings of the aircraft. From his vantage point on the ground, it should have been possible to see ice left on the wing leading edge from the approach to land. However, it was dark at the time, and, although the ramp area was lighted, without close inspection, the darkness would have made such an observation more difficult, particularly on the outboard sections of the wings. Furthermore, it is possible that his inspection was confined to areas of the aircraft under the wings such as the landing gear and engines. If this was the case, ice on the leading edge would not have been detected. Alternatively, it is possible that he did observe ice on the wing leading edge but considered its effects insignificant. The Board could not determine whether the crew knowingly, or unknowingly, attempted the take-off with ice contamination on the wings.

The freezing precipitation which fell during the station stop at Gander was a signal that there was a high potential for ice accretion on the upper surface of the wings. Unfortunately, the absence of a useful cockpit voice recording precluded the Board from establishing what, if any, discussion took place between the flight crew members regarding ice on the aircraft.

Although regulatory requirements, company procedures, training, and advisory material stressed the importance of the clean wing concept, experience has shown that some pilots do not fully appreciate the extent to which small amounts of contaminant can degrade an aircraft's performance, especially swept wing aircraft and, in particular, those not equipped with leading edge devices. Thus, it is possible that the flight crew was aware of the ice contamination and underestimated its effects. Had the crew determined that de-icing was necessary, suitable equipment and facilities were available at Gander. A review of records determined that Arrow Air flights had utilized these facilities on previous occasions.