We used to takeoff with a good coating of ice in the T-37 and KC-135A to no ill effect because the airplanes (and crews) were expendable. Having dodged very many bullets it was easy to look at the "clean aircraft" concept with a jaundiced eye and perhaps that was the case here. But just because one airplane can take it doesn't mean the next can.

— James Albright





CL-604 N90AG

The CL-604 is incredibly sensitive to wing contamination. There was a story about a CL-604 coming out of paint with a strip of masking tape left on the forward edge of a flap causing the airplane to roll during stall maneuvers. I thought that a bit too much to believe at the time. After this mishap I believe it. And now so does the FAA, they have removed the old caveat allowing take off with ice "polished smooth." The key takeaway: you cannot ensure both wings are equal with any amount of contamination so they both have to be contamination free. (See the postscript below for expert testimony about this.

1 — Accident report

2 — Narrative

3 — Analysis

4 — Cause

5 — Postscript



Accident report

  • Date: 04 JAN 2002
  • Time: 12:07
  • Type: Canadair CL-600-2B16 Challenger 604
  • Operator: Epps Air Service
  • Registration: N90AG
  • Fatalities: 2 of 2 crew, 3 of 3 passengers
  • Aircraft Fate: Destroyed
  • Phase: Takeoff
  • Airports: (Departure) Birmingham International Airport (BHX) (BHX/EGBB), United Kingdom; (Destination) Bangor International Airport, ME (BGR) (BGR/KBGR), United States of America



The aircraft arrived at Birmingham Airport on 3 January 2002 at 2039 hrs after a non-stop international flight from West Palm Beach Airport, Florida, USA. The Birmingham METAR for 2050 hrs indicated that the surface wind was from 120° at 6 kt, visibility 8 km, no significant weather, temperature minus 1°C, dew point minus 2°C, QNH 1026 mb.

On arrival, the commander stated that the refueling could be done the following morning in time for the planned 1200 hrs departure to Bangor Airport, Maine, USA. The aircraft was parked overnight on the Western Apron. The same 2 pilots and 3 passengers were to board the aircraft for the accident flight the following day.

There was no precipitation while the aircraft was at Birmingham. The air temperature remained below zero with a minimum temperature of minus 9°C at 0550 hrs. Initially, the sky was clear but the amount of cloud increased to give variable cloud cover after midnight. The surface wind remained south-easterly at less than 5 kt.

The next morning, one of the crew, together with one of the passengers, arrived at the aircraft at approximately 1040 hrs. The commander arrived at approximately 1100 hrs. At different times, each crew member was seen to carry out an independent external inspection of the aircraft. Aircraft refueling commenced at about 1105 hrs and the aircraft fuel tanks were reported full (20,000 lb) at about 1140 hrs. Then, following the arrival of the remaining two passengers, the aircraft doors were closed. During the morning, other witnesses stated that they had seen frost or ice on the wing surfaces of N90AG prior to departure.

Other aircraft, which had been parked overnight, were de-iced during the morning, with associated reports of moderate to severe ice or frost accumulations. Neither crew member requested de-icing, so N90AG was not de-iced prior to departure. The Birmingham METAR at 1150 hrs indicated that the surface wind was from 150° at 6 kt, visibility 8 km, scattered cloud, base 700 feet agl and broken cloud, base 800 feet agl. The outside air temperature was minus 2°C with dew point minus 3°C. The QNH was 1027 mb.

There were no ATC flow restrictions to affect the flight and ATC start clearance was obtained immediately on request. The engines were started at 1156 hrs and N90AG was cleared to taxy at 1201 hrs.

The preceding aircraft movement on Runway 15 was a landing BAe 146 aircraft about one minute before N90AG's departure. The previous departing aircraft had been an Embraer 145 a few minutes before that.

At 1206 hrs, N90AG was cleared to line up on Runway 15 and at 1207 hrs take-off clearance was issued with a surface wind from 140° at 8 kt.

The pilot in the left seat was handling the controls and the commander was seated in the right seat. Flap 20 had been selected for take-off and the following speeds had been calculated and briefed by the pilots: V1 137 kt; VR 140 kt; V2 147 kt. Initial post-accident assessment has shown these speeds to be appropriate to the estimated weight of the aircraft.

The take-off appeared normal up to the time of lift-off. Rotation was initiated at about 146 kt with an initial pitch rate of approximately 4°/second. Lift-off occurred 2 seconds later, at about 153 kt, with a pitch attitude of about 8° nose-up. Once airborne, the elevator position reduced from 8° to 3° (aircraft nose-up) whilst the pitch rate increased to about 5°/second.

Immediately after lift-off, the aircraft started to bank to the left. The rate of bank increased rapidly and 2 seconds after lift-off the aircraft had reached 50° left wing down. At that point, the aircraft heading had diverged about 10° to the left. Opposite aileron, followed closely by rudder, had been applied as the aircraft started banking; full right aileron and rudder had been applied within 1 second and were maintained until the end of the recording. As the bank angle continued to increase, progressively more aircraft nose-up elevator was applied. The last recorded aircraft attitude was approximately 111° left bank and 13° nose-down pitch.

Initial analysis of the FDR parameters indicated that both engines were functioning normally from start-up until the end of the recording. There was no indication of any abnormality which could have accounted for the uncommanded left roll.

Source: Aircraft Accident Report, History of the flight




The effect of the environmental conditions at Birmingham Airport on the formation of frost/ice on the surfaces of a parked aircraft, during 3/4 January 2002, was considered by specialists from The Meteorological Office at Bracknell. Their estimate was that, with the aircraft parked exposed to a minimum temperature/dewpoint of -9°C/-9°C, at dawn, there would have been a deposit of frost on the upper surfaces of wings and horizontal stabiliser.

This was confirmed by witness reports of frost on N90AG and other aircraft. One of these was a Canadair CRJ, which had been positioned, on the evening of 3 January 2002, close to where N90AG was parked later that evening. The CRJ was towed from that location the next morning at about 1030 hrs and was then found by the captain on his pre-flight inspection to require de-icing; he considered that the frost was some 1 to 2 mm thick over the aircraft surfaces.

The following sources of heat could have affected the amount of frost/ice remaining on N90AG’s surfaces at the time of the accident:

  1. Engine and/or APU Exhaust - The APU was started at approximately 1050 hrs.
  2. Fuel - The aircraft was refueled at 1105 hrs.
  3. Solar Radiation - No direct measurement of the ground level solar radiation at Birmingham Airport for the morning of the accident was available. Evidence from nearby observations indicated that the sunlight would probably have been diffused through a localised layer of thin cloud, rather than direct.

Information from the handling agent determined the position on the Western Apron in which N90AG had been parked overnight prior to the accident flight. Post accident observation of this area showed that the aircraft would not have been either partially or wholly in shadow from any ground object on the morning of the accident between shortly after sunrise until the time that it departed for takeoff. During this period the sun would have been generally behind and moving to the left of the aircraft as the sun’s elevation increased and it appeared that no significant shading of the left wing by the fuselage or empennage would have occurred.

Source: Accident Report, ¶1.7.4

Conditions at Birmingham on the night before the accident were conducive to the formation of frost on N90AG. During the morning of 4 January, all other aircraft with flights originating from Birmingham were de-iced and reports related to these aircraft, one of which had been parked adjacent to N90AG, indicated that they had accumulated an extensive covering of frost on their external surfaces. It was inevitable that N90AG had been similarly affected.

Both of N90AG’s pilots were seen to make independent external inspections of the aircraft and, from evidence on the CVR, it was clear that the commander was aware of frost on the wing leading edge. In addition, some witnesses stated that the left wing had some frost on its surface; one witness (the refueller) stated that the right wing upper surface was clear of frost and that the light frost on the right wing leading edge was melting as refueling progressed. Thus it was evident that there was frost contamination on at least some of the wing surfaces of N90AG during the preparations for the accident flight, although there was no evidence to enable the extent of the coverage or its thickness or roughness to be quantified.

N90AG happened to be subjected to a slight tailwind while parked and this could have drifted a warm mixture of engine exhaust flow and ambient air forward over the airframe. However, the wind was light and any such effect would have ceased almost immediately the aircraft started taxiing, three minutes after the start of the first engine. The influence could therefore have been present for a relatively short time only and it was judged that the effect would probably not have been major.

This was not necessarily the case for the exhaust gas from the APU. It was clear from observations during the investigation that the APU exhaust gas flow could appreciably raise the ambient temperature around parts of the parked aircraft and that in a tailwind situation the predominant effect would be on the right wing. N90AG’s APU was operating for a little over an hour, until shortly before the aircraft began taxiing. The testing conducted on a similar aircraft in generally similar conditions showed that in a comparable time period the mean surface temperatures on the right wing rose by around 3°C at 1/3 semispan, 5°C at 2/3 semi-span and 8°C in the tip region. Over the same period there was generally a much smaller increase in surface temperature for the left wing. As the reported ambient temperature at the time N90AG was being prepared was -2°C, the test results indicated that heat from the APU exhaust gas could have reduced, eliminated or smoothed the frost on parts of the right wing, while hardly affecting that on most of the left wing. This would have been generally consistent with the evidence from the refueller. It was likely that the outboard portions of the right wing would have been particularly susceptible to any such de-icing or smoothing effect, as any fuel heat sink effect reduced with the wing taper, or was absent near the tip, and the temperature increment was higher in the outboard regions.

In summary, it was evident that there was frost contamination on at least some of N90AG’s wing surfaces during the preparations for the accident flight. It was also probable that frost contamination was present on the empennage. If the frost were sufficient to trigger the aircraft ice detector system, an initial warning should have been given, but probably would have gone unnoticed and would not have been repeated (see By the time of the takeoff, the frost on the empennage and the left wing had probably not altered greatly, but the frost on the right wing, particularly over the outboard regions, may have been reduced, eliminated or smoothed by heating from the APU exhaust gas.

Source: Accident Report, ¶

Medication and jet lag

Diphenhydramine is a sedative anti-histamine used in a number of cold and allergy preparations on sale to the public. It is also used in a number of products used to aid sleep. Examination of the luggage removed from the wreckage site revealed a number of medications within the baggage belonging to the crew. In the handling pilot’s bag there was a quantity of ‘Excedrin PM - aspirin free’; this medication contains 500 mg of acetaminophen & 38 mg of diphenhydramine citrate per tablet.

The aviation pathologist who carried out the autopsy undertook further research to determine the possible significance of the toxicology findings concerning diphenhydramine. He concluded that both pilots had disturbed and inadequate sleep for the two nights preceding the accident and that it was possible that they were suffering from circadian dysrhythmia (jet-lag). Evidence indicated that both had consumed some alcohol on the evening of 3 January and diphenhydramine was found in their tissues. The Pathologist concluded that it was possible that the tiredness, possible jet-lag and diphenhydramine had all combined to impair the ability of the pilots to deal with the situation with which they were faced.

Following the results of the toxicology tests, a Principle Psychologist was briefed on the circumstances of the accident and contracted to report on the possible human factors aspects of the accident. This included possible fatigue, drug and social factors.

In his conclusion, the Principle Psychologist stated that two errors had occurred. Firstly, the handling pilot had failed to arrive at a proper appreciation of the icing situation during his external inspection. Secondly, the discussion initiated by the commander did not adequately address the issue or arrive at a proper conclusion. The evidence for causal factors underlying these errors was slight. The available evidence suggested that both pilots were probably suffering fatigue on the morning of the accident flight and that this could have predisposed them to errors of judgment and reasoning. This factor probably contributed to the second error and may have contributed to the first.

Source: Accident Report, ¶1.13

The wing


Challenger 604 air foil,
Accident Report, fig. 5

The Challenger wing has supercritical aerofoil sections, ie sections designed to operate efficiently with substantial regions of supersonic flow while at the design cruise mach number (M). For high-speed subsonic aircraft in the transonic flight regime (M 0.7 to M 1.0) acceleration of the airflow over the wing causes the speed of the airflow over parts of the wing to exceed the speed of sound. A standing shock wave is formed towards the rear of the aerofoil where the airflow decelerates to subsonic speeds. A strong normal shock, with an associated high pressure gradient that can induce boundary layer separation, can result in a large increase in drag.

In comparison with previous types of aerofoil, supercritical aerofoils have a reduced camber, an increased leading edge radius, reduced curvature on the upper (suction) surface, and a concavity in the rear part of the lower (pressure) surface. At cruise conditions the profile maintains supersonic flow over a large part of the upper, suction surface of the aerofoil that is then decelerated towards the rear by a weak shock wave. As well as improving aerodynamic efficiency, the design allows a thicker wing section for a given aircraft critical mach number, providing for a more efficient wing structure and additional wing tank fuel capacity compared to previous types of high-speed aerofoil.

Source: Accident Report, ¶

Wing stall

The airflow in proximity to the wing surface is slowed by its passage over the surface to form a boundary layer (BL), ie a layer of air, generally adjacent to the surface, within which velocities are less than the free-stream velocity. Within the BL the local stream velocity increases with distance from the surface, with the outer edge of the BL defined as the location where the local velocity reaches 99% of free-stream velocity. In order to generate wing lift efficiently, the BL must remain relatively thin and generally attached to the surface. Separation of the BL from an extensive area of the upper surface of a wing results in a stall, causing substantial lift loss and drag increase. The lift coefficient CL, (a nondimensionalised measure of lift) reaches a maximum (CLmax) at an AOA just below the stall angle.

Loss of energy from the BL can, if sufficient, cause flow separation. This can result from an excessive adverse pressure gradient, ie an excessive rate of increase in static pressure with distance from the point of peak suction (generally on the upper surface of the aerofoil leading edge). The BL over an aerofoil in an undisturbed airflow is initially laminar, ie streamwise flow within the BL without mixing across the BL. In practice, transition to turbulent flow (with mixing across the BL occurring) generally occurs after a short distance. Available evidence suggests that transition for the Challenger wing typically occurs at around 5% chord.

Supercritical aerofoils tend to generate a relatively flat suction profile over most of the wing upper surface at typical cruise angles of attack. At lower speeds, on both supercritical and conventional aerofoil sections with a relatively low t/c (in the order of 9-12%) the upper surface pressure profile develops an increasingly high and steep peak just behind the leading edge as the AOA increases. At high AOA the high adverse pressure gradient associated with a steep peak can cause the BL flow in the region to separate from the surface for part of its chordwise travel before transitioning to turbulent flow and reattaching to the wing upper surface, forming a laminar separation bubble.

The chordwise extent of the bubble increases with increasing AOA and the stall AOA at a particular spanwise wing station is reached when the flow does not reattach, ie the separation bubble bursts, producing a major reduction in the net suction over the upper profile of the section. Depending on the wing geometry, this change in flow pattern in one spanwise region can influence the flow at adjacent wing stations. This can cause the collapse to extend rapidly across the whole of one wing, causing a sudden major reduction in the lift and increase in the drag generated by that wing. The flow separation near the leading edge produces a more abrupt loss of lift than for the thicker aerofoil types (greater than 15% t/c) more typical of previous, lower-speed aerofoils, where the initial separation tends to occur nearer the trailing edge.

The upper surface suction peak becomes higher and steeper as the curvature of the flow around the leading edge becomes tighter. The upper surface curvature can be reduced at high AOA by use of an aerofoil with a drooped leading edge, rather than a generally symmetrical leading edge, as used on the Challenger wing. The manufacturer noted that appreciable droop would severely degrade the high speed performance of the wing and would therefore be impracticable. The suction peak is also increased by deflection of trailing edge flaps. A common means of suppressing the peak is the addition of leading edge slats or flaps, deploying in concert with trailing edge flaps.

It is traditional practice on lower performance aircraft to arrange the wing geometry such that the wing root regions stall at a lower AOA than the tip regions in order to minimise the rolling moment due to lift asymmetry between the wings at incipient stall. However, this is usually impractical for a wing with the appreciable sweep that is normally used for modern transport aircraft in order to increase the transonic cruise speed. Such designs typically aim to force the initial stall to occur near mid span. The aircraft manufacturer reported that flight testing has shown that the wing stall for the Challenger typically begins at the leading edge of the outboard section of one of the wings. The manufacturer noted that flight testing had shown that, with N90AG’s take-off configuration and Mach Number, the uncontaminated Challenger wing would stall in free air (ie out of ground effect) at a sensor AOA of 25-26°, in the absence of appreciable sideslip. Ground effect would reduce the angle by an estimated 3-4° when the aircraft height above ground was 0-10 feet.

The manufacturer noted that the aircraft’s response is highly dependent on pilot actions once the stall is detected and that large variations in roll rate and peak roll angle can occur. Aileron inputs would normally be ineffective in controlling the roll rate once a wing had stalled, until it was unstalled by reducing the overall AOA. Data for sample Challenger flight test stalls with flaps at 20° showed appreciable wing drops, with roll rates up to 70°/second and bank angles up to 85°. Other information indicated that a rapid wing-drop at the stall would be likely to occur in most cases, irrespective of pilot technique. It was intended that the SPS would prevent the aircraft from reaching a wing stall condition.

Source: Accident Report, ¶



  1. The crew did not ensure that N90AG’s wings were clear of frost prior to takeoff.
  2. Reduction of the wing stall angle of attack, due to the surface roughness associated with frost contamination, to below that at which the stall protection system was effective.
  3. Possible impairment of crew performance by the combined effects of a non-prescription drug, jet-lag and fatigue.

Source: Accident Report, ¶3.(b)



Date: 1 Nov 2014

Hello James, Love this site and visit often always learning something. I'm a technician currently flying with a corporate operator in CT (GX & F900). Anyway, I've been around Bombardier A/C since the mid 90's when I worked at the BDL service center. So I always pay attn to what you write about Bombardier stuff. I couldn't believe when reading the article about the CL604 incident that you mentioned the story about the Challenger roll problem due to tape being left on the flap. I couldn't believe it because I was there and worked on that particular squawk! It's all true and I couldn't believe it at the time either. But it was not a piece of tape, It was (get this!) a paint run on the left inbd flap. It had just come out of the paint shop. We spent days going thru the rigging of those flaps even putting contour boards on the wings. I didn't see the paint run so I can't say how big it was, but I saw the area that was repainted and it was probably 6"X6". Since that incident I now take ANY imperfection on ANY part of ANY wing as something that needs to be addressed. Just wanted to let you know that its a true story. Thanks for such a great site.

If you are ever tempted to takeoff with a slight coating of ice on one wing take a look at the other wing and ask yourself if they are identically coated. Of course you have no way of knowing that. If you decide to takeoff, you have no way of knowing if your ailerons can outroll the difference in lift. This technician has said it better than I could ever hope to: your wings need to be clean prior to takeoff.


(Source material)

United Kingdom Air Accidents Investigation Branch, Aircraft Accident Report 5/2004, Report on the accident to Bombardier CL-600-2B16 Series 604, N90AG at Birmingham International Airport, 4 January 2002