Air Safety — DC-3 Ditches in Botany Bay

The IFR flight plan which the captain submitted at the airport showed that the DC-3 was to fly from Sydney (Kingsford-Smith) Airport to Norfolk Island, with an intermediate landing at Lord Howe Island to refuel. The estimated time of departure from Sydney was 9:00 AM, and the aircraft, carrying 21 passengers, would be crewed by two pilots, a supernumerary pilot, and a flight attendant.

When the passengers, their luggage, and their band instruments arrived at the airport terminal being used by the DC-3, both the captain and the copilot were concerned to see the amount of freight they were expected to carry. But although the captain saw that the total weight would obviously be more than he had allowed for, he considered it would still not exceed the 127 kg which, according to his load sheet, remained available.

Preparations for departure were completed shortly before 9:00 AM, the engines were started, and the DC-3 was cleared to taxi for Runway 16. The captain occupied the left-hand seat, the copilot was in the right-hand seat, and the supernumerary pilot, a flying instructor who had completed a DC-3 copilot endorsement three months before and was aboard to gain further experience, was in the jump seat.

The sunny autumn day in Sydney was fine and almost cloudless, the wind was light and variable with no turbulence, the temperature was 16°C, and the QNH was reported on the ATIS as 1026. As the aircraft was taxying, the flight attendant briefed the passengers on the location of emergency exits, life rafts, and life jackets, and a demonstration of how to fit and inflate a life jacket. She also directed their attention to the safety briefing card.

It had been agreed that the copilot would conduct the departure from Sydney, and at the runway holding point he gave the other crew members a general emergency response briefing, referring to the takeoff safety speed of 81 kt (150 km/h), and stating they would return for a landing in the event of an engine failure. After being cleared to line up on Runway 16, the DC-3 was cleared for takeoff at 9:08 AM.

All engine indications remained normal as the throttles were opened and throughout the takeoff roll, and at 81 kt (150 km/h) the copilot lifted the aircraft off. But during the initial climb, at a height of about 200 feet, with the flaps up and the undercarriage in the process of retracting, the crew heard popping noises above the roar of the engines. Almost immediately, the aircraft began to yaw to the left and the captain called the tower to report they “had a problem”.

The copilot determined that the malfunction was in the port engine, and the captain, having verified this, closed the left throttle and pushed the propeller feathering button (though a lever on most modern twins, on the DC-3 it is a button), at the same time maintaining full power on the starboard engine.

The captain judged that a landing back on the remaining runway might just be possible at this stage of the takeoff, but as the aircraft was capable of climbing on one engine, he quickly decided it would be safer to continue the takeoff over Botany Bay and, at a safe height, to return for an asymmetric landing back on the airport. But although the instrument indications showed that the starboard engine was continuing to deliver full power, the airspeed began to decay.

Realising the DC-3 was failing to climb, and seeing the airspeed was falling below 81 kt (150 km/h), the captain took over control from the copilot. But his own efforts to maintain height also proved to no avail and, less than two minutes after they had been cleared for takeoff, the captain reported to the tower they would be ditching the aircraft.

Positioning the DC-3 for a touchdown in the water as close as possible to the southern end of the partly constructed parallel Runway 16L which extended well out into Botany Bay, he also manoeuvred to avoid small pleasure craft on the approach path he had selected. The aircraft ditched 46 seconds after the captain reported the emergency to the tower.

Because of the rapid onset of the emergency, the flight crew had no time to make any announcement over the DC-3’s public address system. The supernumerary pilot attempted to signal to the flight attendant that there was a problem, but evidently without success. As a result, the passengers were not directed to adopt the brace position for the ditching.

Although the surface of the bay was calm, the initial impact with the water was considerable, both captain and copilot being thrown forward against the windscreen, sustaining minor injuries. Both control seats were fitted with shoulder harnesses incorporating an inertia reel, but neither pilot was wearing the shoulder restraint. Even so, the crew had no difficulty leaving their seats. The captain and the supernumerary pilot immediately went back into the cabin to assist the evacuation of the passengers, while the copilot climbed through the cockpit roof escape hatch and signalled nearby small craft to come to their aid.

The 21 passengers were seated in five rows of four seats, a pair on either side of the central aisle, and in the rear row of two seats on the port side of the cabin. None of them had braced for impact, despite having been referred to the safety briefing card which provided details of the brace position. But although three seat rows broke free from the outboard seat rail in the impact and one passenger’s seat belt became detached, no passengers were injured.

But the flight attendant, who just before takeoff had checked that seat belts were fastened and advised the pilots accordingly before returning to her station, was hurled over the rearmost passenger seats on to the next row of seats as the aircraft struck the water. Although the flight attendant’s seat at the rear of the cabin near the main door was fitted with a shoulder harness, it was evidently not fastened. Obviously concussed, the flight attendant was unable to help with the evacuation.

The DC-3 was carrying sufficient life rafts and life jackets for all passengers and crew, but many of the life jackets were displaced in the impact. Some passengers had difficulty finding a life jacket, others experienced difficulty in donning them because some jackets differed from the type the flight attendant had demonstrated. Even so, all but the crew and one passenger donned a life jacket before leaving the aircraft.

One of the passengers opened the rear main door, and they began evacuating the aircraft in an orderly manner. Deploying a life raft, they assisted the injured flight attendant into it, and two passengers paddled it to nearby pleasure craft. By the time this transfer had been completed, the aircraft was taking water through the forward fuselage and the captain urged the remaining passengers to leave without delay.

Several boats were now on hand to help take the passengers off. When all the remaining passengers were taken off, the captain and the supernumerary pilot also left through the rear door. The copilot meanwhile had walked aft along the top of the fuselage to the tailplane and was taken off by a boat shortly before the aircraft sank.

The tower controller had activated the crash alarm immediately the DC-3 reported that it was ditching, and the airport’s Rescue Firefighting Service, Police, Ambulance, Federal Airports Corporation and the NSW Fire Brigade were all notified at once by common crash call. But by the time Water Police and Maritime Services Board launches reached the aircraft, the majority of the passengers and crew had been taken on board pleasure craft. Altogether eight boats were used to transfer the passengers and crew to shore.

All were then transported to various hospitals for examination. With the exception of the flight attendant, who had sustained serious injuries, all were discharged by early afternoon.

There was some confusion as to the number of persons on board. The flight plan showed 25 passengers and crew, but the captain told Water Police there were only 24 occupants. He had been advised earlier of a late cancellation by one passenger, but did not know that another passenger had taken his place. It took an hour to confirm that all occupants had left the aircraft safely.

Sydney (Kingsford-Smith) Airport is situated on the northern shore of Botany Bay, with Runway 16 extending some 1800m southwards into the bay. The parallel Runway 16 Left, under construction at the time of the accident, lies to the east of Runway 16, and extends about 2500m southwards into Botany Bay. The distance from the taxiway intersection where the DC-3 commenced takeoff to the end of Runway 16 is 3330m.

The aircraft sank near the end of Runway 16L, in about 16m of water and 100m from the sea wall. It was salvaged using inflatable airbags to float it to the surface, then moved to an airport hangar for examination. Except for the separated starboard engine and propeller assembly, which was recovered later, the aircraft was intact. Damage to the airframe was consistent with its impact with the water and the effects of the subsequent recovery operation.

No evidence was found of any pre-existing defect or malfunction in any part of the flight control system. The undercarriage was retracted when the aircraft ditched and the flaps were extended about 20°. The feathered port propeller was at 65 to 66° of pitch instead of the 88° pitch of the fully feathered position.

The starboard propeller, when it was recovered from the seabed, was still attached to its engine. The three blades were bent symmetrically rearwards at the mid position as a result of contact with the water while under power.

The DC-3, constructed in 1944 as a military C-47A, was first registered on the Australian civil register in November 1949 and had flown a total of 42,000 hours. It was fitted with two Pratt & Whitney R-1830-92 radial engines and Hamilton Standard three blade hydromatic propellers. At the time of the accident, both engines were operating on CAA approved concessions to overrun their published overhaul life of 1000 hours.

When the failed port engine was dismantled, excessive wear was found where the No. 3 cylinder inlet valve pushrod enters the valve rocker housing, and on the pushrod itself. Further inspection revealed that one of the thrust washers, normally fitted to the rocker shaft on either side of the rocker arm, was not on the shaft. The loose thrust washer, which subsequently fell out of the valve rocker housing, was oval in shape, having sustained damage during engine operation.

The No. 3 cylinder, complete with inlet and exhaust valve assemblies, inlet valve pushrod and cover and the damaged washer, were then examined to determine how the washer was damaged, and the likelihood of it jamming the inlet valve in the open position.

The rocker arm end of the pushrod exhibited heavy rub marks which penetrated the pushrod to about 25% of the wall thickness, the curved edges of which matched the deformed washer. There was also a curved depression within the rocker housing, adjacent to the inlet pushrod tube, which matched the shape of the deformed washer. Plastic deformation of the cylinder head material into this depression indicated the nature of the compressive loads applied to the pushrod.

The amount of material rolled inside the depression mark suggested that the washer had been in this position for a considerable time. With the washer in the depression, it could become jammed between the cylinder head and the pushrod, thus preventing the inlet valve from closing.

Firing of the spark plugs while the inlet valve was jammed open would cause the fuel-air mixture in the common induction system to ignite. In addition to the loss of power from the No. 3 cylinder, the disruption of the fuel-air mixture to the remaining cylinders would cause irregular engine operation and a loss of up to 50% power, accompanied by backfiring through the intake manifold and carburettor. Soot deposits in the No. 3 cylinder intake cavity, the adjacent intake manifold, and the carburettor air intake assembly, were indicative of backfiring, and possible intake fire.

The investigation was unable to determine where or when the assembly error occurred.

Other substandard conditions found in the port engine were:

The splines of the accessories driving gear to the left magneto were worn well beyond service limits and failure of the magnetic drive was imminent.

The propeller governor pitch control cable 90° pulley block securing bolt was excessively worn.

During removal of the No. 12 cylinder, two of the 16 cylinder base studs were found to be sheared. Light pressure applied to the nut of an adjacent stud resulted in that stud breaking also. Removal of the cylinder revealed an area of galling and fretting in the vicinity of the broken studs, indicating cylinder movement on its mounting pad. The recovered broken stud had failed in fatigue, initiating from around the stud circumference.

The spark plug gap settings were inconsistent between plugs, and the majority of plugs showed evidence of electrode wear beyond normal life. Under testing some plugs were found to be electrically breaking down.

Examination of the port propeller to establish the reason for it not being in the fully feathered position revealed the following:

Torque required to turn individual blades within the hub was 55 lb/ft, 40 lb/ft, and 25 lb/ft respectively, compared with the manufacturer’s specification of 30 to 40 lb/ft.

Each blade butt had been fitted with a plastic sleeve during hub assembly to prevent water ingress. The sleeve of the blade which required the highest torque to turn was dislodged from its position, permitting water penetration and consequent corrosion.

When removed from the propeller dome, the cam assembly with the piston attached did not respond to a force applied to move it towards the fine or feathered pitch positions. Correctly assembled cams with the piston attached move freely throughout the range when propelled by their own weight, without any outside force applied.

Despite in-service cam wear, when the cam assembly preload nut was released by about 40°, the mechanism achieved unrestricted movement of both cams throughout their range, driven by their own weight. There was no damage or abnormal wear within the hub which would have prevented the blades from reaching the fully feathered position had the cam assembly preload nut not been overtightened.

There was nothing to indicate the propeller had been desludged in accordance with an Airworthiness Directive which requires this at 500 hourly intervals.

It was concluded from the examination that the propeller had not been able to operate to the full feather position for some time preceding the accident.

During the investigation, a weight and balance summary of the aircraft’s loading was compiled from known and estimated data gathered. This indicated that the aircraft’s weight at takeoff was 12,446 kg, 562 kg or 4.7% above the MTOW.

Aircraft performance charts relevant to the accident were:

Chart DCA PK16.1/1 (takeoff weight chart) and DCA PK16.1/2 (landing weight chart): These were developmental services charts used during operations in Papua New Guinea and were not approved for use in Australia. However, they were included in the company’s operations manual originally accepted by the CAA.

Chart TAA P19 Issue 1 (takeoff chart) and TAA P20 Issue 1 (landing weight chart): These were originally produced to permit operations up to a MTOW of 12,202 kg (26,900 lb). At the time of the accident these charts were valid for use up to the CAA approved MTOW of 11,884 kg (26,200 lb).

The operations manual required that charts P19 Issue 1 and P20 Issue 1 be used for all normal operations, but gave the captain discretion to use PK16.1/1 and PK16.1/2. The circumstances in which the captain might exercise such discretion were not identified, however. The use of Chart P19 Issue 1 would have precluded takeoff from Lord Howe Island, whereas use of PK.16.1/1 allowed the operation, with the caution that the accelerate/stop distance would not always be available.

Prior to the accident flight, the CAA’s Bankstown Office had advised the captain both orally and in writing to include only the approved performance charts in the operations manual. Although he amended Chart P19 Issue 1 to reflect the maximum takeoff weight of 11,884 kg (26,200 lb), he did not remove Chart PK.16.1/1 or Chart PK.16.1/2 from the manual as advised.

When the operator carried out initial task planning some time prior to the flight, providing the charterer with a load availability of 2160 kg, this did not take into account the life rafts, additional drums of oil, or spare parts and tools to be carried. Furthermore, no provision was made for the supernumerary crew member. The load availability quoted thus exceeded that actually available by approximately 600 kg.

Performance calculations using Chart PK16.1/1 were made by the captain, despite the fact that this performance chart was not approved. To complete the flight to Norfolk Island, the aircraft had to stage through Lord Howe Island requiring takeoffs at the Chart PK16.1/1 MTOW from Lord Howe on both the outbound and inbound legs. The operator was awarded the charter contract after offering payload availability on the basis of using that performance chart.

The company which owned and operated the DC-3 was based at Camden, NSW, but one of its principals was also a partner in a company based at Moorabbin, Vic, which held an Air Operator’s Certificate for the operation of normal category aircraft.

To enable commercial operations with the DC-3, the Moorabbin based company requested the CAA’s Moorabbin office to include DC-3 aircraft on their AOC. The CAA at Moorabbin thus became responsible for surveillance of the DC-3 operation.

But because the Moorabbin based company’s chief pilot had limited multi-engine experience and no DC-3 experience, CAA Moorabbin agreed that a DC-3 flight captain, based at Camden, could be appointed to exercise some of the chief pilot’s responsibilities. This function was delegated to the captain of the accident flight and he was later approved as check and training captain. However, overall supervision of the operation remained the responsibility of the chief pilot based at Moorabbin.

When assessing an operator for approval of an AOC or for the addition of a new aircraft type to an existing AOC, compliance with both operational and airworthiness regulatory requirements are to be assessed. These include a system of maintenance, appropriate facilities and equipment, and documentation. Yet the inclusion of the DC-3 on the AOC was approved by CAA Moorabbin without any inspection being conducted to ensure these airworthiness requirements were met.

Subsequently, because the DC-3 and its operator were based at Camden, the Bankstown CAA office, at the request of CAA Moorabbin Office, accepted airworthiness surveillance responsibility, as well as specific operational surveillance on request from the Moorabbin Office.

Most of the CAA Bankstown’s involvement with the DC-3 operator during the period leading up to the accident focused on the DC-3 flight captain, check and training approvals, and operations and performance limitations. No flight operations surveillance, responsibility for which had been retained by CAA Moorabbin, was conducted. Actual airworthiness surveillance was limited to one inspection, when the aircraft was flown to Bankstown for radio maintenance.

The certificate of airworthiness for the aircraft was issued in the transport category, requiring the aircraft to be maintained to Class A standard. However, the aircraft continued to be maintained as a Class B aircraft. The Class A system of maintenance is more structured and accountable for quality assurance than the Class B system.

The investigation found that within the CAA, there was documentation which gave conflicting information as to whether this and other DC-3 aircraft were classified as transport or normal category aircraft, and therefore subject to Class A or Class B maintenance respectively. This resulted in some confusion among those administering the system.

Ten months before the accident the operator submitted an application to CAA Bankstown for a concession to exceed the published engine overhaul life for the starboard engine. It included two oil sample analysis reports and a compression/ground run test report. The CAA approved an extension of 100 hours.

Six months later, the operator submitted a further application for the port engine to exceed its overhaul life. Again two oil sample analysis reports and a compression/ground run test report were included. The CAA airworthiness inspector assigned to the operator expressed the opinion that both engines should not operate on the same aircraft in the overrun period at the same time, but the CAA did not have a policy to this effect.

Despite indications that the port engine was suffering mechanical distress, together with marginal cylinder compression recordings, further information was not sought by either CAA Bankstown Office or the CAA Central Office and the concession was approved. The inspector had been advised verbally that the starboard engine had been changed and when the port engine overrun approval was granted, he assumed that the starboard engine had been changed.

Interviewed after the accident, the pilots expressed differing views on exactly when the engine malfunction occurred. The captain thought it occurred after the undercarriage was selected up, at a height of about 200 feet with the airspeed in excess of 100 kt (185 km/h) and probably close to the normal climb speed of 113 kt (209 km/h). But despite this and power indications on the starboard engine of 48 inches Hg and 2700 rpm, the aircraft would neither climb nor maintain altitude.

The copilot thought the aircraft was at 200 feet with the undercarriage retracting when the failure occurred, but was not clear about the airspeed reached. He considered that the propeller was slow to stop after feathering action and could remember trying to maintain 81 kt (150 km/h). He was unable to do so and the aircraft diverged to the left of the runway centreline.

The supernumerary pilot said he was not paying close attention to the instruments but heard the sound of the engine malfunctioning, and was aware the aileron control was held at about 90° from the neutral position. Both the copilot and the supernumerary pilot said that almost full right aileron had been used to control the aircraft but they could not recall the skid ball indication. The copilot also said he had almost full right rudder applied.

As the emergency progressed, the aircraft’s configuration changed as the undercarriage retracted, and the port propeller moved toward the feathered position. Although these actions substantially reduced aerodynamic drag, additional drag was induced when the copilot, having applied right rudder, also applied substantial right aileron control.

Back in 1947 and 1948, the RAAF carried out asymmetric handling and performance tests on C-47B Dakotas with the intention of producing a report for airline operators and the then Department of Civil Aviation. At a weight of 11,884 kg (26,200 lb), the tests showed that with the port engine failed at 86 kt (159 km/h), the undercarriage down, and the port propeller windmilling, a climb rate of 63 ft/min could be obtained. The rate of climb reduced to zero if the airspeed was increased to 94 kt (174 km/h) or reduced to 78 kt (145 km/h).

In 1953, further tests were carried out to investigate the possibility of raising the maximum all-up weight of the Dakota to above 12,700 kg. As part of the tests, one takeoff was made at 12,928 kg (28,500 lb) with a simulated port engine failure at 88 kt (163 km/h). In this instance the aircraft was able to maintain height only while flown in ground effect.

In 1954, TAA investigated the approved takeoff speeds for the DC-3 and the following was subsequently published in a TAA Supplement to Aircrew Bulletin:

Satisfactory asymmetric takeoffs were performed at 26,200 pounds (11,884 kg) with the engine failing at 81 knots (150 km/h), and at 27,000 pounds (12,247 kg) with the engine failing at 86 knots (159 km/h). Previous tests had shown that at 26,900 pounds (12,202 kg) with the engine failing at 80 knots (148 km/h) the aircraft lost airspeed as the climb was started and fell back onto the runway. It is felt that the extra five knots (9 km/h) is required at this weight to allow for the drop in airspeed when the aircraft attitude is changed on beginning the climb. It was suggested that the takeoff safety speed be 81 knots (150 km/h) for weights of 26,200 pounds (11,884 kg) and below, increasing linearly to 86 knots (159 km/h) at 26,900 pounds (12,202 kg). (Metrics added for reader convenience. Ed)

The TAA report showed that the success of an asymmetric takeoff was greatly dependent on the flying technique adopted after the engine failure. For example, sudden changes of attitude were accompanied by loss of airspeed and a consequent inability to climb away. It therefore recommended that, upon engine failure at the critical engine failure speed, the aircraft should be held at this speed while the undercarriage was being retracted, and the speed then increased to 90 kt (167 km/h) while the propeller was being feathered.

It was clear from this data that only minimal climb performance is available after engine failure at 81 kt (150 km/h) during takeoff at the MTOW of 11,884 kg (26,200 lb). At higher weights, the aircraft will not achieve any climb performance unless the takeoff safety speed is increased linearly with the increase in aircraft weight. At weights above 12,202 kg (26,900 lb) the DC-3 is unlikely to achieve any climb performance unless all parameters are within their optimum ranges.

In the case of the accident flight, the crew flew the DC-3 off at 81 kt (150 km/h) and subsequently reported that the aircraft accelerated to at least 100 kt (185 km/h) before they shut down the port engine. But despite instrument indications that full power was being obtained from the starboard engine, they were unable to prevent the speed reducing below the takeoff safety speed of 81 kt (150 km/h).

The following is an extract from the operations manual used by the operator:

Critical Engine And Minimum Speeds At Maximum Auw

The PORT engine is the critical engine. Minimum speeds vary from an absolute minimum of 68 knots (126 km/h) IAS with the port engine feathered and power settings of 42 inches manifold pressure and 2250 on the starboard engine, to 76 knots (141 km/h) IAS with the port airscrew windmilling in full fine pitch and full power of 48 inches manifold pressure and 2700 RPM on the starboard engine. The limiting factor VMCA of 180 pounds foot pressure is reached at approximately 73 knots (135 km/h).

It is preferable to keep straight by use of rudder alone, rather than using the aileron. The foot load is within the capabilities of all pilots, but there is a danger of the foot slipping up on the brake pedal unless the rudder pedals are adjusted before takeoff so that full rudder can be applied with the heel. For long-legged pilots, this makes for a rather uncomfortable seat position, with the control column fouling the knees.

The DC-3 at all-up weight has a marginal performance at V2 speed (81 knots/150 km/h) on one engine, and this requires concentration on the part of the pilot to see that the best performance is obtained. It is of vital importance that the climb performance of the DC-3 in the asymmetric condition is fully understood.

The operations manual thus contained instructions describing the actions needed to be taken to ensure that optimum single engine performance at a MTOW was achieved – accurate airspeed control, directional control using rudder alone, full power on the remaining engine, undercarriage and flaps retracted and, if necessary, the use of ground effect. Once the propeller had feathered, the speed was to be increased to 91 kt (189 km/h). Operations at weights above MTOW were not permitted and therefore were not addressed in the operations manual.

The captain had flown 29 hours on the DC-3 in the previous 90 days. He had completed his initial endorsement on DC-3 aircraft in 1979 and had recommenced flying the type in November 1992. Six months later he had been granted check and training approval for the DC-3.

The copilot, a part owner of the aircraft, had flown a total of 25 hours in the last 90 days, all on this DC-3. Granted a commercial pilot’s certificate in the USA in January 1992, he was issued with an Australian commercial licence in September 1993. He told the CAA he had completed DC-3 command endorsement training in the USA on April 5, 1992, and on the basis of this advice, his certificate of validation was annotated with a DC-3 rating. His most recent formal check was for the renewal of his command instrument rating on January 9, 1994. Subsequent investigation however determined that, at the time the licence was issued, the pilot had not completed the required DC-3 endorsement training.

Basi’s Summing Up:

The origin of the defect determined to have most likely caused the port engine malfunction could not be finally established. However, an appropriate response by the CAA to the port engine’s condition should have prevented further operation of the engine.

As the aircraft was heavily loaded, the captain had the option of conducting the takeoff himself to ensure that the aircraft’s performance was maximised. However, he not only permitted the inexperienced copilot to conduct the takeoff, but also chose not to follow through on the controls. Consequently, when the engine malfunction occurred, the captain did not immediately take over control of the aircraft. He assessed the engine should be shut down and carried out appropriate checklist actions. But following the engine shutdown, he did not respond to the inadequate aircraft handling by the copilot until the only option available was to ditch the aircraft.

Footnote: BASI makes the point that its investigation reports are provided to promote aviation safety. In no case are they intended to imply blame or liability.

Next Issue: The accident from the pilot’s perspective.