|CASB Minority Report|
No Ice Contamination
No Ice on the Aircraft
The findings of the majority with respect to ice contamination are based on theoretical possibilities, We confuted these in detail in our paper Critique of the Ice Contamination Hypothesis Presented in Conditional Draft No. 1 (presented to the Board in May 1988).
The majority has adduced no direct evidence of ice on the aerodynamic surfaces of the Arrow Air DC-8, the only evidence of ice anywhere on the aircraft is one reference to a small amount on an unheated edge of a windshield. This reference was made by a refueller who spoke with the flight engineer in the cockpit just before departure. His words were as follows:
"I noticed some ice buildup around the edges of the cockpit window, and I asked him if they picked up much ice on the way in. He said, 'No it wasn't too bad, there's a tiny bit left around the window.'"
To us, this means that the crew had monitored ice during the approach, had used the airframe de-ice equipment if needed, and that the flight engineer knew from his inspection there was no ice on the lifting surfaces.
The Boeing 737 that took off from Gander shortly before the accident landed in St. John's within an hour. This aircraft, which had to descend through the same cloud conditions as the Arrow Air DC-8 on its approach to Gander, did not need de-icing at St. John's. The Boeing 737 that landed at Gander shortly after the accident did not pick up any ice during the approach.
The captain of the Arrow Air DC-8, an instructor and check pilot, was universally lauded for his professionalism and meticulous attention to detail. A close professional colleague testified that the captain was aware and wary of the effects of ice contamination, having cited these, for example, in discussions of the 1982 Air Florida crash at Washington.
The Human Factors Group Factual Report concluded that the flight engineer was "extremely conscientious and thorough in his approach to his professional duties" and that his "adherence to standards was noted by many sources."
Given the evidence of the crew's professionalism. we conclude that the captain checked the wing as he left the aircraft and that the flight engineer conducted a thorough external inspection. Had there been ice on the leading edge, the crew would have detected it and had the aircraft de-iced.
Two refuellers who could not fail to see the leading edge while connecting the fuel hose to the refuelling panels (see Figure DO1) testified that they saw no ice. Had de-icing been necessary, it could have been provided on a fee-for-service basis. None of the four ground handlers who would have done the work noticed any ice. Earlier that morning, one of these workers recommended de-icing to the captain of an aircraft that had been at the airport for some time. This witness stated that the Arrow Air DC-8 did not need de-icing because there wasn't any ice on it.
The witness testimony and the detailed meteorological evidence (as presented at the Board's public inquiry and discussed in our previously cited paper) establish that the wing of the Arrow Air DC-8 could not have been contaminated with ice during the take-off run at Gander on 12 December 1985.
The Aircraft Did Not Stall
The majority based its finding that the aircraft stalled mainly on interpretations of the heading, vertical acceleration, and altitude traces from the flight data recorder. The aircraft's attitude at first contact is cited as supporting evidence. Our paper Critique of the Ice Contamination Hypothesis .... particularizes why both categories of evidence are unconvincing.
We observed, for example, that the alteration in heading is more consistent with a gentle turn than with a stall. We also submitted that the "substandard" vertical acceleration trace (discussed but not reproduced by the majority) provides no support either for or against the notion of a stall.
The majority assessed fluctuations near the end of the altitude trace as an indication of stall buffet. Our paper noted that analogous fluctuations on the trace from the previous take-off disappeared when the aircraft reached about 100 feet above ground. Subsequent analysis by several consultants found: "A study of the altitude stylus marks during take-off going back to twenty flights before the accident ... suggests that such fluctuations tend to be associated with the longer range flights that were made at higher gross weights" That is, fluctuations in the altitude trace near lift-off are characteristic of the installation and not of the accident flight.
The majority cites the high angle of attack at first contact (estimated from the "tree model") as collateral evidence of a stall. Our paper cited a possible estimation error of about six degrees, along with the reduction of apparent angle of attack due to pitch rate, as reasons to be wary of this interpretation. The full upward deflection of the elevator at impact indicates that the pilot may have pulled the control column back in a last ditch effort to reduce speed and delay the crash.
We also note that ground effect could have had no substantive influence on the occurrence or non-occurrence of a stall during the accident flight. This is readily demonstrated by modifying the computer code used for the performance calculations cited by the majority. Recalculation of test cases with no allowance for ground effect produces minor differences (typically: reductions of some 15 feet in maximum altitude and some 350 feet in distance covered) without affecting the presumed stall.
The conclusion that the aircraft did not stall can be drawn from evidence of a number of witnesses about the level attitude of the aircraft as it crossed the TransCanada Highway. ("It was a normal departure... but that levelling-off effect was abnormal"; "The aircraft's attitude appeared to be level as it crossed the Trans-Canada Highway"; "The plane was levelled off. The plane wasn't nose up or nose down. It was level"; "It looked very flat. Just two or three degrees"; "The nose was not pointed up": "The aircraft was pretty level... Very level.")
In conclusion, the Arrow Air DC-8 did not stall before it crashed.
Performance Not Consistent With Ice On Wings
Our paper Critique of the Ice Contamination Hypothesis ... also points to misinterpretations of the performance calculations reported by the majority.
The computer models (and the flight simulator modifications) extrapolated lift and drag values beyond the range of experimental data. The resulting steeply rising drag curve generated very large increases in drag for modest increments of angle of attack. Thus, the calculations allow minute amounts of "equivalent roughness" to overpower all four engines at take-off power.
Even if we were to accept the assumed effect of ice on lift and drag, we would have to reject the computed results because they depend on unrealistic assumptions about the crew's reactions. To make the computed trajectories "crash" at about the right distance. it was necessary to Further assume that, when confronted with decaying airspeed and negative rate of climb, the crew would pull up and hold the aircraft at an angle of attack of 18 degrees. At this angle, aerodynamic buffet would warn the crew to lower the nose - with or without ice on the wings, with or without synthetic stall warning. In any event, corresponding fuselage attitudes of about 10 degrees contradict the observations of the tower controller ("It was a normal departure... but that levelling [sic] off effect was abnormal") and a number of witnesses ("The aircraft's attitude appeared to be level ... "; "The plane was leveled off. The plane wasn't nose up or nose down"; "It looked very flat, just two or three degrees", etc.)
The calculations cited by the majority take no account of the turn and sideslip which we feel are essential features of the accident flight path. The stall presumed in the calculations should at least coincide with the beginning of the turn. Those cases that lead to approximately correct altitude gain and distance show no such coincidence.
We noted in our (previously cited) paper that the flight data recorder indicates a deceleration near the end of the short flight on the order of what would be produced by aerodynamic drag on the standard (i.e., not iced up) aircraft with all engines stopped. The computer program cited by the majority can be used, not only to verify this, but also to find a better fit to the known characteristics of the accident flight - through the assumption that the engines start to spool down shortly after liftoff.
The most natural, tractable assumption for computing the observed performance is that of a massive power loss followed by the expected crew reaction of lowering the nose to try to maintain airspeed. The turn to the right may indicate that the power loss was most severe on that side. It may also indicate additional control problems.