1. The aircraft was being operated in weather conditions conducive to engine carburettor icing.
2. The aircraft maximum take-off weight limitation was exceeded for the flight.
3. The aircraft loaded moment envelope limitation was exceeded for the flight.
4. The aircraft departed controlled flight at a height above the ground from which aerodynamic stall recovery would have been unlikely.

Loss of engine power under normal loading conditions causes the aircraft nose to pitch downward for aerodynamic stall recovery because of the aircraft designed forward centre of gravity (C of G). Loading in a tail-heavy direction or rearward C of G condition has a most serious effect upon longitudinal stability, affecting the aircraft's ability to readily recover from stalls and spins. As the C of G moves rearward, a less stable condition occurs, which decreases the ability of the aircraft to right itself after manoeuvring or after disturbances by gusts. The aircraft as loaded had not exceeded the rearward C of G limitation. Although the rearmost C of G limit had not been exceeded, the location of the C of G just 25.4 mm ahead of that limit, meant that the aircraft exhibited a rearward C of G condition.

The aircraft exceeded both the aircraft manufacturer's Maximum Take-off Weight (MTOW) limitations and the aircraft loaded moment envelope. Exceeding the aircraft MTOW limitation may adversely affect flight characteristics. CAAP advisory number Number 235-1(1), advised pilots against using standard weights and recommended weighing occupants and baggage in order to prevent exceeding those limitations.

The examination of the engine and carburettor revealed no evidence of a preimpact failure or anomaly. The aircraft was most likely being operated with the carburettor heat set to the OFF position as indicated by the position of the heat lever at the air box on the engine. With the aircraft operating in weather conditions conducive to carburettor icing, it may have begun losing power. The onset of carburettor icing may have been insidious, as the pilot may not have noted the deterioration in engine RPM. As the aircraft was not equipped with a carburettor ice detection system, the pilot was not afforded any warning of the potential for carburettor icing. Without this warning, if the engine performance deteriorated, the pilot most likely would not have been able to apply carburettor heat in time for it to take effect sufficiently to regain full power. With the rearward C of G condition present, the pilot may not have been able to pitch the nose of the aircraft downward as required for aerodynamic stall recovery. However, in any case, the stall was most likely unrecoverable because of the low height above the ground. Weather conditions encountered on previous flights to and from the island may not have been sufficient to produce carburettor icing or may have been masked by the constant speed propeller.

The possibility also exists that wind shear and turbulence in the area, in combination with the adverse flight characteristics resulting from exceeding the aircraft's loaded moment envelope limitation, could have degraded the aircraft's controllability and resulted in the aircraft's departure from controlled flight.

While several possibilities exist as discussed in this analysis, the investigation could not conclusively determine the reason for the excessive nose-up pitch and departure from controlled flight.

The pilot of the Cessna 172G aircraft was conducting a series of charter flights between the Trefoil Island Aircraft Landing Area (ALA) and the Smithton, Tasmania aerodrome. Witnesses stated that the aircraft, with the pilot and three passengers on board, took off from the island ALA runway 28 on a west-south-westerly track at approximately 1745 hours EsuT, on the third return flight of the afternoon. Witnesses reported that the aircraft turned to the left on a southerly heading while climbing, followed by a left turn to the east. They reported that following the turn to the east, and after it had overflown the buildings on the island at approximately 200 feet above ground level, the nose of the aircraft pitched up abruptly to an angle of 30-40 degrees. According to the witnesses, following the nose-up pitch, the aircraft rolled abruptly to the left, lost altitude and descended from their line of sight. The witnesses heard the impact of the aircraft and ran to render assistance. The aircraft was destroyed by impact forces and all four occupants received fatal injuries.

Wreckage information

The wreckage of the aircraft was oriented on a heading of 191 degrees magnetic, indicating that it had rotated through about 270 degrees during the descent. The aircraft impacted the ground wings level, with a nose-down angle of approximately 39 degrees, on a downward sloping hill of approximately the same angle. There were no indications that the aircraft was in a spin at the time of impact. The cabin roof had separated at the rear attachment and both wing struts had separated. The forward cabin area had collapsed, with the tail section and the tail cone buckled and bent partially forward. Wreckage evidence indicated a high rate of vertical deceleration, in excess of 24 g (acceleration due to earth gravity, international standard value being 9.80665 metres per second squared, assumed at standard sea level), with indications of little forward airspeed.

The propeller/crankshaft assembly had separated behind the radius of the crankshaft flange. The fracture surface displayed evidence of a unidirectional bending overload failure, indicating low engine RPM at the time of the fracture. Examination of the propeller spinner and propeller blades confirmed low engine RPM at impact. The carburettor heat lever at the air box on the engine was noted to be in the OFF position. The aircraft was fitted with an elevator trim that allowed the pilot to minimise load forces on the elevator, depending on the position of the centre of gravity (C of G), airspeed and power settings. The aircraft's elevator trim system was found in the slightly nose down from the TAKE-OFF TRIM or neutral position. The seats and seat rails incurred substantial damage, but the pilot's seat end stop was still located intact on the seat rail. There were no indications of a bird strike on the aircraft.

Aircraft information

A 100 hourly inspection was completed on 17 February 2003 at 9,663.6 hours total time airframe (TTAF) with no major anomalies noted. A 50 hourly engine inspection was completed on 12 March 2003 at 9,713.6 hours TTAF and 1,085.2 hours engine time since overhaul (TSO), with no anomalies noted. At the time of the accident, the aircraft had accumulated 9,718.4 hours TTAF. The maintenance release listed no outstanding discrepancies for the aircraft and was current and valid.

Nothing was found during the investigation to suggest a mechanical failure of any part of the aircraft that could have contributed to the accident.

Engine information and examination

Supplemental Type Certificate number SA807CE was incorporated in 1977 with the installation of a 180 horsepower Lycoming model O-360-A1A engine and a Hartzell constant speed propeller, replacing the 145 horsepower Continental model O-300C engine and fixed pitch propeller. At the time of the accident, the engine, serial number L21971-36A, had accumulated 1,090 hours TSO. A technical disassembly and inspection of the engine and carburettor was completed at an independent maintenance facility under Australian Transport Safety Bureau (ATSB) supervision. The disassembly and examination did not reveal any evidence of pre-impact internal component failure or anomaly.

Pilot information

The pilot held a valid commercial pilot (aeroplane) licence and Class 1 medical certificate at the time of the accident. The pilot's last flight review was completed on 6 January 2003. Post-mortem and toxicological examination did not identify any factor that may have impaired the pilot's ability to operate the aircraft safely.

Meteorological Information

Documents recovered at the accident site included an Airservices Australia on-line weather forecast briefing for the area and for King Island and the Smithton aerodrome. The forecast was dated 14 March 2003 and the time noted was 1046 hours. Wind listed on the briefing for the 2,000 ft level was forecast as variable at 15 kts. The forecast also noted a south-westerly stream with a slow moving trough with drizzle and locally broken low cloud. King Island was located 107 km to the north of Trefoil Island. The King Island meteorological report noted light drizzle, south-south-westerly wind at 15 knots and a temperature/dewpoint spread of 5 degrees C.

A series of wind generators was located on the Tasmanian mainland at Cape Grim, approximately 5 km to the southwest of Trefoil Island. This facility periodically monitored and logged weather conditions. Documented information obtained from that facility indicated that the weather conditions at 1700 hours were: air temperature 15.4 degrees C; dew point 11.4 degrees C; relative humidity 77 percent; and wind from the southwest at 27 kts with gusts to 30 kts. Information documented at 1800 hours recorded: the air temperature 15.1 degrees C; dew point 10.6 degrees C; relative humidity 74 percent; and wind from the southwest at 28 kts, gusting to 29 kts. Relative humidity recorded from 1500 hours to 1700 hours (the estimated time of the first two return flights) was recorded as 77-79 percent. The wind recorded at that time was from the southwest and varied from 29 to 30 kts with the temperature/dew point spread from 4.0 to 5.6 degrees C. Witnesses stated that the accident occurred at 1750 hours. Last light on the day was about 2011 hours.

A pilot familiar with the area reported that a south-westerly wind often caused orographic lifting (when air is forced upwards by a barrier of mountains or hills) moving heavily laden moist air into the flight path of an aircraft departing from the island.

Carburettor and engine induction system icing

A search of the ATSB occurrence database indicated a total of eight carburettor or engine induction system icing related accidents since May 1994. These accidents resulted in two fatalities. One aircraft was severely damaged and three aircraft were destroyed. An article on the US Federal Aviation Administration (FAA) website, reprinted from Vintage Airplane Magazine and dated November 1994 stated: `According to the National Transportation Safety Board, carburettor ice was involved in over 360 accidents in the past five years. These figures do not include the unreported off-airport landings and incidents caused by icing. The results were 40 deaths, 160 injuries, 47 aircraft destroyed and 313 aircraft severely damaged.' Several of these accidents noted suspected carburettor icing at high power settings. Float-type carburettors, such at that used in the occurrence engine, are most susceptible to this event. Evidence of carburettor icing is highly perishable and dissipates rapidly.

When carburettor ice forms, it can obstruct the smooth flow of the air/fuel mixture, which results in a reduction of engine RPM, power, and an associated loss of airspeed and altitude. FAA Advisory Circular AC 20-113 provided information pertinent to aircraft engine induction system icing. It noted:

`c. Fuel Vaporization Ice- This icing condition usually occurs in conjunction with throttle icing. It is most prevalent with conventional float type carburettors, and to a lesser degree with pressure carburettors when the air/fuel mixture reaches a freezing temperature as a result of the cooling of the mixture during the expansion process that takes place between the carburettor and the engine manifold.'

The circular also noted that vaporisation icing may occur, when a relative humidity of 50 percent or higher is present, at temperatures from 0 degrees C to as high as 37.7 degrees C. It also stated that in general, when the temperature/dewpoint spread reaches 6.6 degrees C or less and a relative humidity of 50 percent or higher, there is a potential for icing.

The values from the weather observations for 1800 hours at Cape Grim were plotted on a carburettor icing probability chart. The temperature/dewpoint spread was 4.5 degrees C. The plot was located in the area of the chart labelled `serious icing- any power setting'.

Mitigating the effect of carburettor icing involves pilot action to apply full carburettor heat (the ON position), which initially causes a further loss of power (perhaps as much as 15 percent). The air heated by the exhaust is directed into the engine induction system, which results in a richer fuel/air mixture and additional power loss. A delay of 30 seconds up to several minutes may be expected until normal engine power returns. The circular recommended the use of carburettor heat briefly (particularly with float-type carburettors), immediately before take-off if the relative humidity was above 50 percent and the temperature below 21 degrees C, to remove any ice which may have accumulated during taxi and pre-flight engine checks.

The engine manufacturer recommended that carburettor heat should not be used for take-off as it was not necessary and it may cause detonation and possible engine damage. The aircraft manufacturer recommended a check of the system before take-off, and that carburettor heat be placed ON in the event of an engine failure, other than immediately following take-off. The Operations Manual noted that an unexplained loss in engine speed could be caused by carburettor icing or air intake filter ice and cautioned to watch for signs of icing and apply carburettor heat as required. The section titled `Engine Failure After Take-off (under 700 feet)' did not mention the use of carburettor heat.

One characteristic of the onset of carburettor, or induction system icing, on an engine fitted with a fixed pitch propeller, is the gradual deterioration of the engine RPM. Aircraft engines equipped with constant speed propellers, such as the accident aircraft, compensate for the gradual RPM deterioration by decreasing propeller pitch to maintain a given RPM.

The previous owner of the aircraft reported experiencing an engine power loss (with the 180 horsepower engine fitted), while on approach to land several years earlier. That event was believed to have been due to carburettor icing, as no mechanical anomaly was discovered. The previous owner further reported that the onset of the power loss was immediate, with little time to react.

Optional equipment for the Cessna 172 model aircraft included a Carburettor Ice Detector system. This system utilised an optical probe in the carburettor throat, which is so sensitive that it can detect `frost' up to five minutes before ice begins to form, giving the pilot time to take corrective action. Examination of the carburettor revealed that an optical probe was not fitted.

Aircraft performance

Reports from witnesses indicated that the aircraft took off from the island ALA runway 28 and that take-off performance was apparently acceptable. According to the Aircraft Flight Manual, the maximum permissible crosswind component for take-off and landing was 15 kts. The Operations Manual stated, `Pilots will not take-off or land a Company aircraft when the crosswind component exceeds that specified in the relevant Aircraft Flight Manual.' Using the weather information noted previously, the crosswind component during take-off was calculated by the ATSB to be in excess of 15 kts.

Aircraft fuel

The aircraft fuel selector was found in the BOTH position (both tanks feeding the supply line to the engine) as required for take-off. Damage to the aircraft fuel tanks precluded establishing the exact fuel state of the aircraft at the time of impact. There was a strong smell of fuel in the area of the crash site. A fuel sample was removed from the right wing tank and sent for analysis by a National Association of Testing Authority approved laboratory. The laboratory confirmed that the fuel sample was Avgas 100, which was the correct grade and specification for the engine and no anomalies were noted.

An examination of load sheets used on previous flights to the off-shore islands was completed. These sheets confirmed that the pilot had previously adhered to the Operations Manual policy of maintaining a maximum of 120 L total fuel for flights to off-shore islands of 40 minutes or less, to avoid aircraft structural stress during ground operations. The flight from Smithton to Trefoil Island was approximately 12 minutes. The investigation team generated a flight plan using this fuel information, weather data, witness statements and fuel consumption estimates for the aircraft. The ATSB calculated that the aircraft had approximately 90 L of fuel at the time of take-off on the accident flight.

Aircraft loading

The Operations Manual stated that `Pilots shall prepare a passenger list/manifest and leave it for retention at the aerodrome of departure on all Charter Flights. Pilots will also prepare and leave passenger list/manifest prior to departing the off-shore islands.' The manual specified that the passenger/manifest sheets were to be left `inside the tractor shed' on Trefoil Island. It also stated that if no scales for weighing were available, portable scales were to be carried for use by the company pilots. It further stated that company pilots were to ensure that the aircraft was loaded strictly within the weight and balance limitations.

No passenger/manifest sheet for the accident flight was recovered from the accident site, the island ALA area, or from the operator's Smithton facility. A witness stated that the pilot did not leave the immediate vicinity of the aircraft and did not leave any documents behind prior to the accident flight. No portable weighing scales were recovered from the accident site. When interviewed, passengers from previous flights to and from the off-shore islands reported that they were not weighed, and that the pilot had rarely asked for their body weights. These same passengers reported that the pilot personally loaded all baggage into the aircraft baggage compartment, but did not weigh it. Witnesses who observed the pilot loading the baggage compartment prior to the accident flight also reported that he personally loaded the baggage, but did not weigh it.

The FAA Type Certificate Data Sheet (TCDS) for the aircraft noted the maximum permissible baggage compartment load limitation was 120 pounds (54 kg). Numerous items, such as tools and personal equipment, were located in the immediate area of the wreckage. When weighed these items totalled 108.2 kg. Of that amount, 19.7 kg included items not normally kept in the baggage compartment, but in the main cabin area. That indicated a total baggage compartment load of 88.5 kg at the time of take-off, 34.5 kg in excess of the maximum weight limitation for the aircraft.

The TCDS further noted that the Maximum Take-off Weight (MTOW) for the aircraft was 2,300 pounds (1,043 kg). The MTOW is the maximum allowable weight at the start of the take-off run. The fuel estimated to be on board at the time of take-off was 90 L. ATSB and aircraft manufacturer's calculations indicated an aircraft take-off weight of 1,123 kg (2,475.7 pounds), signifying a take-off weight in excess of the maximum limitation by 79.6 kg. The calculations also indicated that the MTOW would have been exceeded even with all fuel removed.

Advisory material

The Australian Civil Aviation Safety Authority, Civil Aviation Advisory Publication (CAAP) Number 235-1(1), Standard Passenger and Baggage Weights, recommended the following:

`11. Because the probability of overloading a small aircraft is high if standard weights are used, the use of standard weights in aircraft with less than seven seats is inadvisable. Load calculations for these aircraft should be made using actual weights arrived at by weighing all occupants and baggage.'

Aircraft balance and centre of gravity

Aircraft balance refers to the location of the C of G, along the longitudinal and lateral axis. In order to assure predictable aircraft control, the aircraft manufacturer established limitations along the longitudinal axis at fuselage stations measured in inches, in relation to a reference point or datum (located at the forward face of the engine compartment firewall). The C of G limitation for operation of the aircraft in the normal category at maximum Take-off Weight of 2,300 pounds (1,043 kg) was a forward limit of 38.5 inches (977.9 mm) aft of the datum and rearward limit of 47.3 inches (1201 mm) aft of the datum. Aircraft longitudinal C of G was calculated by dividing the total moment of the empty aircraft and on-board items (weight multiplied by the fuselage station) by the total weight of the empty aircraft and items. The location of the C of G at the estimated take-off weight of 1,123 kg (2,475.7 pounds) was approximately 46.3 inches (1176 mm) aft of the datum.

US FAA AC 91-23A Pilot's Weight and Balance Handbook (superseded) stated, `C.G. limits may be expressed graphically in the aircraft weight and balance reports by means of an index envelope. The envelope defines the forward and aft limits and also the maximum weight limit in terms of index units.' The index envelope for the Cessna 172G was referred to as the loaded aircraft moment envelope. Flight with a total moment outside of this envelope was not recommended. When plotted, the calculated loaded aircraft moment at take-off was outside the C of G moment envelope, exceeding the aircraft manufacturer's recommendation.

The aircraft manufacturer advised that with an aircraft loading condition as plotted, if rising terrain and strong winds combined to create significant vertical shear, the risk of loss of control of the aircraft would be increased, even if anticipated by the pilot. The manufacturer further reported that the nose pitch-up and subsequent departure from controlled flight as witnessed were consistent with an aircraft that was flown exceeding the loaded aircraft moment envelope limitation. The estimated aerodynamic stall speed and other aircraft performance figures at the calculated aircraft take-off weight were not available from the manufacturer, as the aircraft was being operated outside the certified moment envelope.

The HS-748 aircraft was engaged in a night freight operation and was under charter to a freight carrier. The aircraft had been loaded with 4 LD type containers and an amount of freight. During the post loading walk-around, the first officer had difficulty removing the tail stand from the aircraft. The first officer asked the loading staff if the aircraft was loaded in accordance with the load sheet. The loading staff indicated that the aircraft had been loaded according to the sheet. The first officer consulted the pilot in command and ascertained that the loaded centre-of-gravity of the aircraft was towards the middle of the allowable centre-of-gravity envelope. The crew then removed the tail stand and completed the before start checks and started the engines in preparation for taxi.

Shortly after the engines had been started, the loading staff approached the aircraft and signalled to the pilot in command that they wished to speak to him. The pilot in command shut down one engine and dispatched the first officer to speak with the loading staff. After speaking with the loading staff, the first officer signalled to the pilot in command to shut down the other engine.

Subsequent inquiries by the loading staff revealed that an incorrect container had been loaded on board the aircraft. These inquiries revealed that the aircraft had been loaded with an empty LD container in the forward position. The correct LD container for the forward position was expected to weigh 1120 kg. The aircraft had been loaded incorrectly.

The freight carrier instituted an internal investigation that revealed a number of factors contributing to the incorrect loading. These factors included an absence of loading documentation for both the loading supervisor and the pallet loader operator. There was also an absence of loading documentation for flight crew that indicated the container identification and disposition.

The Robinson R44 helicopter departed from Shipwreck Bay, QLD at about 1245 Eastern Standard Time on the final leg of a charter flight to Mareeba aerodrome. The pilot intended to track directly to Mareeba aerodrome under the Visual Flight Rules, with an expected enroute flight time of about 45 minutes. The pilot did not submit a flight plan to Airservices Australia, nor was a flight plan required, but the company retained a flight notification form for search and rescue purposes.

At about 1 NM to the south of Mareeba township, at a cruise altitude of 1,000 ft and an indicated airspeed of about 95 kts, the pilot heard an unusual noise associated with the main rotor blades. The pilot reported that the noise sounded as though a potato chip packet had been caught in the blades and a violent vibration occurred in conjunction with the unusual noise.

The pilot was unable to identify or read the aircraft instruments because of the vibration and was unable to adequately control main rotor RPM after the onset of that vibration.

The violent vibration necessitated a landing. The pilot entered autorotation and selected a paddock for the forced landing. He reported that the helicopter was difficult to control during the descent. In particular, he was unable to turn the helicopter with left and right cyclic inputs. In particular, while cyclic inputs altered the attitude of the helicopter left and right it would not turn in the selected direction. The pilot also reported that it was difficult to control main rotor RPM during the autorotative descent and he relied upon the low rotor RPM horn and light as an indication of rotor RPM. The horn and light activated intermittently during the descent.

During the descent, the helicopter struck powerlines and was substantially damaged. The pilot reported that he was unable to avoid the powerlines because of the ineffective cyclic inputs. The pilot and passenger, the only occupants, suffered minor injuries.

A report from the accident site indicated significant main rotor blade skin disbonding 60 mm from the tip and extending 1070 mm inboard on one blade and the beginning of skin disbonding on the other blade. The ATSB conducted a preliminary investigation and it was apparent that the problem was a continuing airworthiness matter. ATSB released the main rotor blades to the Civil Aviation Safety Authority (CASA) for further examination as part of an airworthiness investigation.

Standard company procedures required a monitored approach in the weather conditions prevailing at Darwin for the aircraft's approach and landing. The approach was conducted at night and in conditions of rain and reduced visibility.

The aerodrome controller had selected the runway 29 HIAL and HIRL to a higher intensity than recommended in MATS for the initial setting given the prevailing weather conditions. Although this selection assisted the crew to acquire the HIAL and HIRL at an earlier stage of the approach, the apparent intensity of those lights increased significantly as the aircraft approached the runway. The crew did not realise that the lights were too bright until the aircraft was passing overhead the HIAL and consequently, did not request a lower intensity selection.

The aircraft commenced deviating from the extended runway centreline about 7 seconds after the autopilot was disengaged and as the HIAL was starting to disappear from the pilot's view. Associated with this deviation were control inputs by the pilot in command, including a left rudder pedal command and a right wing down control wheel input that resulted in the aircraft entering a cross-controlled sideslip to the right as the aircraft approached the runway. These control inputs were not conventional for the environmental conditions.

The control inputs made by the pilot in command during the final stages of flight altered the aircraft's flight path across the ground and directly contributed to the aircraft's deviation from the runway centreline. The control wheel inputs for right roll resulted in the deployment of the right wing's flight spoilers during the final 70 ft of the descent. This increased the drag on the right wing, reduced the lift produced by that part of the aerofoil section and consequently, increased the rate of descent. The investigation could not determine the reason for the rudder and aileron inputs made by the pilot in command.

The data derived by kinematic analysis by the aircraft manufacturer was consistent with other available sources of environmental wind data.

The pilot in command did not detect the aircraft's increasing lateral displacement from the runway centreline. He considered that he had sufficient visual reference to complete the landing. However, during the final seconds prior to touchdown, it was possible that he encountered an abnormal situation where few reliable visual cues were available for determining the aircraft's position relative to the centreline of the runway.

The nature of the available visual cues increased the difficulty for the pilot in command to detect the aircraft's increasing displacement from the runway centreline. This included a wet runway surface with a probable lack of surface definition, painted runway markings that were less conspicuous on a wet runway at night, a lack of touchdown zone lighting/centreline lighting on a runway that was wider than normal and the possibility that the HIRL was glaring on the wet windscreen. The investigation concluded that the presence of centreline lighting would have increased significantly the nature of the visual cues available and would have assisted the pilot to recognise the developing sideslip and lateral deviation from the runway centreline.

The lack of a positive flare, a marginal reduction in headwind component and deployment of the right wing's flight spoilers during the final stages of the approach contributed to the high rate of descent at touchdown. The wider runway would have provided an unfamiliar set of cues for judging the flare height. The lack of runway surface definition would have increased the difficulty for the pilot to estimate the height of the aircraft above the runway and possibly had contributed to the lack of a positive landing flare. This was also coupled with a different flap setting from that routinely used during landing.

The aircraft was sideslipping to the right at the point of touchdown. The excursion from the runway was not preventable due to the sideslip and the proximity of the aircraft to the edge of the runway. There was no evidence to indicate that standing water, or adhesion of tyres on the wet runway surface, were factors in the excursion from the runway.

During the final stages of the approach, the copilot was monitoring various parameters. He did not detect the increasing displacement of the localiser or make any call for correction prior to touchdown. The size of indicated deviation, together with the other instruments being monitored, made this an unlikely deviation to detect.

The investigation concluded that the presence of runway centreline lighting would have increased the visual cues available to the pilot and assisted with his recognition of the developing sideslip and lateral deviation from the centreline.

History of the flight

On 19 February 2003, a Boeing Company 737-376 (737) aircraft, registered VH-TJB, landed on runway 29 at Darwin. The aircraft touched down close to the right edge of the runway and ran off the sealed runway surface. The handling pilot returned the aircraft back to the runway during the landing roll. There were no reported injuries to either the passengers or crew. The aircraft sustained minor damage.

The aircraft was operating a scheduled public transport passenger service between Adelaide and Darwin, with six crew and 79 passengers. The approach was conducted at night and in conditions of reduced visibility due to rain. The automatic terminal information service¹ (ATIS) reported 6,000 m visibility at the aerodrome.

The runway was wet and the previous landing aircraft had reported that the braking action on the runway was good. The aerodrome controller had selected the high intensity approach lighting (HIAL) and high intensity runway lighting (HIRL) to Stage 6 (maximum intensity). The visual approach slope indicator system (T-VASIS)² was operating. Due to the weather conditions, the crew elected to perform a monitored approach³ and configured the aircraft with flap⁴ 40, and the autobrake⁵ set to 3.

Initial approach

Consistent with company procedures, the copilot was the handling pilot for the initial stage of the monitored approach and provided input to the aircraft's automatic flight management system. The pilot in command monitored the progress of the approach and attempted to establish visual reference with the runway. The aircraft was flown with both autopilots engaged and coupled to the instrument landing system (ILS) for runway 29. The threshold reference speed⁶ (Vref) for a flap 40 landing was 131 kts.

Analysis of data from the aircraft's flight data recorder (FDR) confirmed the aircraft was configured for landing prior to reaching the outer marker⁷ (OM) and that the approach parameters were stable. The pilot in command recalled that the aircraft crossed the OM and was at an altitude of approximately 1,000 ft when the HIAL became visible. At the decision altitude⁸ (DA) he could clearly see the approach lighting and runway lights and decided to continue the approach and land the aircraft. Consistent with company procedures for the monitored approach, the pilot in command became the handling pilot for the remainder of the approach and landing. The aircraft's landing lights were 'ON' during the final approach and landing.

Final approach

The FDR indicated that the autopilot was disengaged approximately 2 to 3 seconds after passing the DA and about 20 seconds prior to touchdown. At the time of autopilot disengagement, the aircraft was established on the localiser and glide slope, maintaining a heading of 283 degrees magnetic (deg M) and was about 200 ft above the height of the runway threshold (HAT). The aircraft's flight path deviated above glide path following disengagement of the autopilot.⁹

About 6 seconds after the autopilot was disengaged, the FDR recorded a control wheel input that resulted in a slightly right wing low bank attitude and then application of left rudder. The combined effect of those control inputs altered the aircraft's heading to the left, but introduced a sideslip to the right and a corresponding angle of drift. The localiser deviation recorded on the FDR indicated that the aircraft started to drift right of the extended runway centreline about 13 seconds before touchdown. Application of left rudder and roll attitude that was predominantly right wing low continued to the point of touchdown, by which time the aircraft's heading was about 7 degrees left of the aircraft's ground track and the localiser deviation indicated 0.57 dots 'fly left'.¹⁰ The control wheel inputs resulted in deployment of the right wing's flight spoilers¹¹ during the final 70 ft of descent.

The pilot in command recalled that during the final stages of the approach, he could see the runway lights along the full runway length and was satisfied that he had the required visual reference to continue the approach. He recalled that rain was streaming across the windscreen and that the wipers were operating. As the aircraft descended into the touchdown zone he observed that the runway surface was very dark and there was a lack of surface definition. He recalled shifting his gaze from the runway aim point to the cues available at the far end of the runway to judge the height for the landing flare. The pilot in command recalled hearing the synthesised calls of altitude from the aircraft's radar altimeter, which occurred at radar altitudes of 50 ft, 30 ft and 10 ft.

The pilot in command recalled that he did not detect any anomalies with the aircraft's approach path during the final stages of flight and was unaware that the aircraft had commenced to sideslip as it approached the runway. The copilot recalled that he did not detect any anomalies with the aircraft's flight path as he monitored the various flight instruments during the final stages of the approach.

The FDR recorded a backing of the wind direction and a gradual increase in wind speed during the 11 seconds prior to touchdown. Although this change in wind velocity represented an increasing left crosswind component as the aircraft approached the runway, this occurred after the aircraft had commenced the sideslipping manoeuvre.

Touchdown and landing roll

The pilot in command recalled that, due to the wet runway surface, he intended to make a firm touchdown on the runway. During the final stages of the flight, the aircraft's pitch attitude increased from about 2 to 3.2 degrees nose-up. Analysis of data from the FDR indicated that the aircraft touched down with a descent rate of approximately 600 ft/minute and a vertical deceleration of 2.3 g.¹² Information contained in the operator's manuals indicated that the normal descent rate for touchdown should be about 150 ft/minute, with a nose-up pitch attitude of between 4 and 6 degrees.

The pilot in command recalled that immediately after touchdown, he brought his view back down the runway and saw that the runway edge lights were tracking down the windscreen centre frame. He immediately realised that they were close to the right edge of the runway and heard the aircraft wheels strike runway lights. Corrective control inputs returned the aircraft to the runway centreline and the aircraft completed the rollout.

Marks on the runway and data from the FDR provided further information on the touchdown and landing roll. Touchdown occurred approximately 520 m from the threshold of runway 29, at a computed airspeed of 127 kts (a ground speed¹³ of about 128 kts). The right main landing gear was about 1.4 m inside the edge of the runway and the aircraft was sideslipping to the right. The right main landing gear departed the runway about 590 m from the threshold, at a ground speed of about 124 kts. The left main landing gear departed the runway about 760 m from the threshold, at a ground speed of about 111 kts. Corrective control inputs by the pilot had returned all wheels to the runway by about 1,130 m from the threshold. At the maximum point of excursion, the right main gear was about 7 m from the edge of the runway and the left main gear was about 2 m from the edge of the runway.

The pilot in command taxied the aircraft clear of the runway. Although there were no abnormal cockpit indications about the landing gear, he requested a precautionary inspection of the aircraft's landing gear by the airport's emergency rescue and fire fighting service before taxiing to the terminal.

Damage to the aircraft

During the landing roll, five runway edge lights were struck by the aircraft's landing gear. Examination of the aircraft revealed damage to the tyres and impact damage to the intakes and fan sections of each engine, mainly associated with the ingestion of runway light fragments. There was also minor damage to the surfaces of the wing flaps. Significant quantities of grass had accumulated in the vicinity of the main wheel brake packs and had also contaminated the wing flaps.

Runway 29 ILS

Runway 29 was equipped with a Category 1 ILS, which enabled pilots to make instrument approaches in conditions of low cloud and reduced visibility. The ILS consisted of a 3-degree glide slope and a localiser aligned on a track of 285 deg M. The relevant instrument approach procedure required the pilot to have visual reference with the runway threshold or approach lighting at the DA (290 ft pressure altitude, which was 209 ft HAT), with at least 800 m visibility. If the visibility subsequently reduced below landing minima, a missed approach was required. The aerodrome's ILS was operating normally at the time of the incident.

Runway 29 physical environment

Runway 29 was 3,354 m long and 60 m wide. The central 45 m of the runway was grooved to assist with wet-runway braking characteristics and tyre adhesion. The runway was not equipped with centreline lighting or touchdown zone lighting, nor was this required for runways equipped with a Category 1 ILS. However, the Manual of Standards (MOS) - Part 139 Aerodromes issued by the Civil Aviation Safety Authority (CASA) recommended provision of centreline lighting on runways where the width between runway edge lights was greater than 50 m.¹⁴ The runway touchdown zone and centreline were marked on the runway. These markings were relatively well defined and provided contrast against the dark runway surface during daylight conditions and on a dry runway surface.

At 60 m wide, Runway 29 was significantly wider than other Australian runways¹⁵ used by the operator's 737 fleet. As a consequence, the visual cues and runway perspective available to the pilot to complete an approach and landing on runway 29, were different from those normally available.

The average longitudinal slope of runway 29 was 0.2%, with two distinct crests along the runway. The runway crests obstructed portions of the runway and altered the pilot's view of the runway during the final stages of the approach and during the landing flare.

Runway 29 approach lighting

The intensity¹⁶ of the runway's HIAL and HIRL was selected by the aerodrome controller. These had each been set to intensity setting Stage 6 at some point before the 737 crew commenced their approach. The 737 had crossed the OM when the controller transmitted that the HIAL was selected to the maximum setting. The pilot in command of the 737 acknowledged this transmission.

The Manual of Air Traffic Services (MATS) indicated that the initial intensity of the HIAL/HIRL should be set according to the prevailing visibility and ambient light conditions. At night and for the ATIS reported visibility of 6,000 m, MATS indicated an initial setting of Stage 1. Variations to these settings could then be made at the pilot's request. Stage 2 was to be set at night with visibility greater than 4,000 m, but less than 5,000 m, Stage 3 with visibility greater than 2,000 m, but less than 4,000 m and Stage 4 when the visibility was not greater than 2,000 m.

Stage 6 was the maximum intensity for the HIAL/HIRL and MATS indicated that setting should be made during the day when visibility was less than 2,000 m.

Both the pilot in command and copilot recalled the intensity of the lights as they approached and overflew the HIAL. The copilot reported glancing outside during the final approach while the autopilot was still engaged, but after the DA and recalled seeing bright HIAL lights with a black empty area behind and thinking that the HIAL was too bright during the later stages of the approach. The pilot in command reported that, although HIAL set to Stage 6 helped with the early acquisition of the runway environment during the approach, the HIAL appeared brighter as they got closer to the field and the intensity was slightly uncomfortable as they flew overhead. By the time he realised that the lights were too bright, the aircraft was passing overhead the HIAL and it was too late to request the controller to select a lower intensity setting.

Meteorological information

A monsoonal squall line had recently moved through the Darwin area from the west-south-west, but was clear of the airport at the time of the occurrence. There were no thunderstorms in the vicinity of the aerodrome.

While the 737 was on descent and manoeuvring to intercept the final approach, a heavy shower of rain passed overhead the aerodrome and the controller broadcast that the visibility at the aerodrome had reduced to 4,000 m. The pilot in command of the 737 reported his position at the OM and requested an update on the visibility at the aerodrome. The controller indicated that the visibility from the tower was approximately 5,000 m. After landing, the pilot in command reported to the controller that visibility around 3,000 m was experienced during the approach.

Data from the Bureau of Meteorology's (BOM) Low-Level Windshear Alerting System (LLWAS) did not record any significant wind gusts in the period immediately preceding the aircraft's arrival at the runway threshold. The anemometer closest to the threshold of runway 29 recorded westerly wind between 5 and 9 kts at the time the aircraft landed.

Data from the BOM's automatic weather station recorded various parameters on a minute by minute basis. During the minute that the aircraft touched down, the recorded average wind was 250 deg T at 7 kts and 0.2 mm of rainfall was recorded. No rainfall was recorded during each minute either side of the minute during which the aircraft was landed.

Monitored approach procedure

The operator required that a monitored approach be performed when visibility was below 5,000 m and/or low cloud existed at the destination aerodrome. The procedure required the copilot to fly the aircraft (or provide input to the automatic flight control system) with reference to the flight instruments during the initial part of the approach. The pilot in command was required to monitor the progress of the approach and assess the visual reference available to complete the landing.

The pilot in command was responsible for deciding if sufficient visual reference was available to land the aircraft and became the handling pilot for the final phase of the approach and landing. The copilot was required to monitor the flight instruments and ensure that the parameters of the subsequent approach remained stable. He was also required to call out deviations outside approach tolerances until the point of touchdown. If visual reference with the landing runway was subsequently lost, the pilot in command was required to commence a missed approach.

The copilot reported that during the final stages of the approach, while monitoring the flight instruments, he did not detect any localiser deviation outside approach tolerances. He became aware of the aircraft's proximity to the runway's edge as he looked up from the instrument panel immediately following touchdown. The FDR indicated that the displacement of the localiser was about dot 'fly left' at the point of touchdown.

Use of autoland

The aircraft was equipped for autoland operations that permitted operations onto runways equipped with a Category 2 or Category 3 ILS. In addition, the operator permitted autoland operations on other approved runways when the weather conditions were above the minima for Category 1 operations. The operator had not approved autoland operations for runway 29 at Darwin.

Landing configuration

Due to the wet runway the crew had elected to perform a flap 40 landing. This reduced the threshold reference speed (Vref) and the possibility of the tyres dynamically hydroplaning¹⁷ on any standing water on the runway surface. A flap 40 landing required a slightly flatter attitude at touchdown when compared with the more routinely used flap 30 landing.

Technical crew information

The pilot in command was an experienced 737 check and training captain and had logged 17,906 hours aeronautical experience, which included 8,930 hours on the 737. The copilot had transferred from another aircraft type and had recently been endorsed on the 737. At the time of the incident, he was completing line training under the supervision of the check and training captain.

Both pilots held valid medical certificates. The pilot in command required reading glasses for near vision, but distance vision was reported to be normal. A post-incident ophthalmic examination revealed no other anomalies with the pilot in command's vision.

Both crew had signed on in Melbourne earlier on the day of the incident and had completed a sector to Adelaide. At the time of the incident, both had been on duty for about 6 hours 40 minutes and awake for approximately 16 hours. Both crew members had been free of duty for a period of 42 hours prior to signing on. They both reported being well rested prior to commencing duty.

Environmental conditions and relevant human factors

The investigation analysed the extent to which a number of environmental conditions could have altered the visual cues available for the pilot in command to complete the landing. This included an assessment of the available visual cues, the possible existence of visual illusions during the final stages of the approach and other factors such as rain on the windscreen, movement of the wipers and the effect of glare from the HIAL/HIRL.

The crew reported that they applied rain repellent to each windscreen during the approach. The copilot recalled that the pilot in command had requested the wipers to be set to 'high' during the final stages of the approach.

The analysis of the available visual cues during the final stages of the approach indicated:

• that the runway's HIAL lighting started to disappear from view (under the cockpit glare shield) about 23 seconds before touchdown
• the final bar of HIAL lights disappeared from view about 12 seconds prior to touchdown
• the runway threshold lights disappeared from the pilot's view about 10 seconds before touchdown and from this point of the approach, the pilot in command was required to judge the aircraft's lateral position over the runway surface using the runway edge lights
• a portion of runway lights would have disappeared behind the second crest of the runway about 4 seconds before touchdown and another portion of runway lights would have disappeared behind the first runway crest about 2 seconds before touchdown.¹⁸

The available visual cues were changing during the finals stages of flight and their salience depended on several factors. These included the extent to which individual runway lights may have glared on the wet windscreen and provided an indistinct reference set for judging the aircraft's increasing lateral displacement from the runway centreline. Other contributing factors may have included the action of the wiper blades across the windscreen and the effects of glare from the HIRL.

Data supplied by the operator

The investigation examined information from the operator's Flight Operations Quality Assurance (FOQA) database. This information was collected routinely from aircraft equipped with a quick access recorder and analysed for the purpose of monitoring the aircraft's performance during the landing approach. Around the time of the incident, this data was being expanded to monitor the accuracy of the aircraft touchdown.

A review of the available 737 data for approaches to runway 29 at Darwin from 2003 through to mid 2004, revealed that the mean average of daylight touchdowns was slightly further from the centreline when compared with night touchdowns, but that the standard deviation for these touchdowns was marginally larger at night.

FDR analysis by the aircraft manufacturer

At the request of the Australian Transport Safety Bureau, the aircraft manufacturer provided additional analysis of the data from the FDR. The manufacturer confirmed that the aircraft was cross-controlled¹⁹ during the manual portion of flight and that this had resulted in the development of a sideslip. Of particular significance, the manufacturer noted that the environmental wind data recorded by the FDR was not reliable during sideslipping flight. Kinematic²⁰ analysis of various FDR parameters indicated that the environmental crosswind component remained near zero during the final stages of flight and did not indicate any significant increase in crosswind component as recorded by the raw FDR data. The kinematic analysis also indicated a 5 kt reduction in headwind component during the final stages of flight.

¹ An automated transmission indicating the prevailing weather conditions at the aerodrome and other relevant operational information for arriving and departing aircraft.
² The T-VASIS consisted of high intensity lights on either side of the runway, in proximity to where the glide path for the instrument landing system intersects the runway. That provided visual approach slope guidance for pilots, and included a transverse bar of four lights on either side of the runway.
³ A monitored approach is a reduced visibility procedure where one pilot will fly the aircraft with reference to flight instruments, while the other pilot monitors the approach and assesses the visibility conditions. The procedure is fully explained later in this report.
⁴ Flap 40 is the maximum flap extension, which results in a lower approach/touchdown speed. This reduces the required landing distance and also the possibility of aircraft tyres dynamic hydroplaning on runways affected by standing water.
⁵ The autobrake system has four landing settings, 1, 2, 3 and Max, and selects the desired deceleration rate for landing.
⁶ The threshold reference speed is for a specific landing weight and flap configuration and is published by the aircraft manufacturer.
⁷ The outer marker is a navigation aid associated with the final approach fix for the runway 29 ILS approach procedure and is on the extended runway centreline, approximately 3.8 NM from the threshold.
⁸ The decision altitude for the runway 29 ILS procedure was 290 ft. This was 209 ft above the height of the threshold of runway 29.
⁹ This was probably a consequence of nose-up trim applied by the autopilot during the final stages of automatic flight.
¹⁰ Localiser deviation is indicated on a display marked with dots. An indication of dot 'fly left' represents the aircraft displaced to the right of the runway centreline.
¹¹ Roll control for the aircraft was provided by ailerons and flight spoilers on each wing. The flight spoilers would begin to deploy at about 10 degrees rotation of the control wheel.
¹² g - Acceleration due to Earth gravity, international standard value being 9.80665 m/s2, assumed at standard sea level.
¹³ Ground speed is the aircraft speed relative to the ground, whereas airspeed is a relative velocity between an aircraft and the surrounding air.
¹⁴ That was consistent with the recommended practices of the International Civil Aviation Organization (ICAO).
¹⁵ Most runways used by the operator's 737 fleet were 45 m wide.
¹⁶ Intensity of the HIAL/HIRL for 6-stage lighting, was 100% (Stage 6), 30% (Stage 5), 10% (Stage 4), 3% (Stage 3), 1% (Stage 2) and 0.3% (Stage 1) of the maximum lighting intensity. Changes of these magnitudes were required for the human eye to detect that a change in intensity had occurred.
¹⁷ Dynamic hydroplaning is a condition where standing water prevents tyre contact with the runway surface and results in a lack of traction between the tyre and the runway surface.
¹⁸ These lights normally provide visual cues to assist the pilot judge the landing flare.
¹⁹ Application of flight control movements in the opposite sense to those in normal turns or manoeuvres.
²⁰ A branch of physics that deals with the motion of a body without reference to force and mass.