On 27 January 2023 a Boeing 777 aircraft, registered ZK-OKN, was conducting a scheduled passenger flight from Melbourne, Australia to Auckland, New Zealand. Shortly after touchdown at Auckland airport, the aircraft briefly lost directional control and the aircraft veered away from the runway centreline. There was damage reported to 6 runway edge lights and to the aircraft’s undercarriage assembly, including deflation of one tyre, there were no injuries to passengers or crew.

The Transport Accident Investigation Commission (TAIC) requested assistance from the Australian Transport Safety Bureau (ATSB) to download the aircraft’s cockpit voice recorder (CVR) and flight data recorder (FDR) to assist their investigation.

To facilitate this support and to provide the appropriate protections for the information, the ATSB appointed an accredited representative in accordance with paragraph 5.23 of ICAO Annex 13 and commenced an investigation under the Australian Transport Safety Investigation Act 2003.

The recorders were downloaded at the ATSB Canberra facilities on 9 February 2023 and the voice and flight data from the occurrence flight were retrieved and provided to the TAIC investigators in attendance. ATSB staff assisted with further analysis of the data as required. 

The TAIC was responsible for the investigation and released the final investigation report regarding this occurrence on 29 May 2025. The report is available here.

Investigation summary

What happened

On 13 January 2023, Pacific National intermodal service 1150, collided with Qube Logistics intermodal service T296 while conducting a propelling movement in Port Botany yard, New South Wales. 

Both trains were positioned on adjacent arrival roads when train 1150 passed a signal indicating stop and collided with the side of train T296, propelling on a proceed signal, on converging track at the end of the 2 arrival roads.

As a result of the collision, several wagons from both trains derailed and sustained significant damage. One derailed wagon crossed a vehicle access road and came to rest against a privately‑owned building. 

There were no injuries, but there was significant damage to rollingstock and rail infrastructure.

What the ATSB found

Propelling train 1150 passed signal BY31 at stop, colliding with train T296, which was propelling on a proceed signal on the converging adjacent track.

The qualified driver directing the propelling movement of train 1150 mistook the location of their train. As a result, when signal BY29 cleared to allow T296 to proceed, the qualified driver incorrectly authorised train 1150 to also move.

ARTC's line of sight shunting procedure was unique to Port Botany yard and was described by train crew as difficult and time consuming to conduct. This led to operators routinely deviating from the procedure, for a perceived simpler method of working.

Aspects of ARTC's line of sight shunting instructions were ambiguous, resulting in varying interpretations of the procedure by stakeholders.

ARTC did not proactively follow risk assessment principles specific to propelling movements to ensure safe rail operations within shunting yards under its control. ARTC conducted operations without the hazards/risks being fully or effectively assessed, examples of this included:

• Proposed treatments (risk controls) which may have reduced risk were not further evaluated
• risk reviews were not conducted in accordance with ARTC’s procedures 
• active train crew were not well represented in risk assessment processes
• issues raised by operators with line of sight shunting procedures were not risk assessed
• minimal compliance monitoring of risk controls.

The second person, who was not authorised to drive a train under instruction of the qualified driver, operated 1150 en route to Port Botany. On arrival at Port Botany, the qualified driver of 1150 performed the on-ground duties, leaving the unauthorised second person to operate the train unattended during the propelling movement.

What has been done as a result

ARTC has responded to safety concerns raised as part of this investigation and has:

• Updated and clarified local instructions for line of sight shunting in Port Botany yard, contained in Network Information Book OGW-30-25.
• Developed OPE-PR-072 Line of Sight Shunting Port Botany procedure with requirements consistent with Network Information Book OGW-30-25. The procedure includes a new requirement for ARTC to conduct assurance activities on rail operators conducting Line of sight shunting in Port Botany.
• Undertaken a structured review of risks associated with line of sight shunting at Port Botany in accordance with ARTC procedure RSK-PR-001 Risk Management. This process included formal risk workshops, review of existing controls, assessment of proposed treatments and establishment of ongoing review requirements. The workshop included operational subject matter experts and frontline rail safety workers involved in Port Botany line of sight shunting activities, including train crew, shunters and driver trainers. The following rail operators were involved:
• Pacific National Pty ltd
• Qube Logistics (Rail) Pty Ltd
• Holdco Holdings Pty Ltd (trading as Southern Shorthaul Railroad)
• Sydney Rail Services.
• Reviewed and updated the risk assessment in June 2025, October 2025 and January 2026. ARTC procedure RSK-PR-001 Risk Management requires this risk to be reviewed at a period no longer that every 12 months as the highest current risk level is medium. The risk assessment includes a review of all current controls and proposed treatments with justifications where treatments have been rejected.
• Noted, as an additional risk mitigation strategy, ongoing infrastructure developments at Port Botany that are expected to significantly reduce reliance on line of sight shunting operations over time. This includes installation of locomotive traversers at both DP World Stevedores (completion due Q4 2027) and Patrick Stevedores allowing for loco-leading arrival of trains. 

Safety message

Rail infrastructure managers should follow their accredited risk management procedures, to ensure risk controls remain relevant and effective.

An effective risk assessment¹ requires effective and meaningful engagement with subject matter experts and end users to identify risks and guide decision‑makers on the most appropriate way to manage risks; and if controls are administrative, also guide the thoughtful design and development, and practical implementation of procedures. 

Once procedures are developed and implemented, accredited operators should provide rail safety workers with sufficient guidance and means to access written procedures, to allow the end users to refresh their theoretical knowledge on practical protocols. 

End users must be given the opportunity to provide feedback on procedures. This feedback should be incorporated in risk reviews to ensure procedures remain fit for purpose and effective in managing risk. 

When unique procedures are developed, care should be taken to align protocols with existing overarching procedures to minimise the risk of ambiguity when implementing local controls. 

Rail operators must ensure that rail safety workers hold a verified competency for the task they are conducting.

The occurrence

Arrival of trains 1150 and T296 at Port Botany

On the morning of Friday 13 January 2023, Pacific National (PN) intermodal service 1150, consisting of 1 locomotive and 31 container flat wagons, was scheduled to travel from the Pacific National Intermodal Terminal in St Marys, to Port Botany yard, New South Wales (NSW), and return. 

The train crew of 1150 consisted of a qualified driver and second person employed by RailTrain (RT), a labour provider of train crew to PN. At the start of their shift on the day, the qualified driver and the second person swapped roles, with the second person driving under the supervision of the qualified driver on the trip from St Mary’s to Port Botany. 

At approximately 0430, train 1150 arrived onto arrival road number 2 at Port Botany yard. 

Positioned on the adjacent arrival road number 1 was Qube Logistics (Qube) train T296, an intermodal service that was scheduled to travel between Qube Intermodal terminal in Minto, NSW, to Port Botany and return. T296 consisted of 1 locomotive and 30 container flat wagons and was crewed by a driver and second person employed by Sydney Rail Services (SRS), a labour provider to Qube.

T296 had arrived at Port Botany at 0410, approximately 20 minutes prior to 1150, and had carried out a run around movement to place the locomotive on the rear end of their train in readiness for propelling towards the stevedore ports. Shortly after arrival, the crew of 1150 also went through the process of placing their locomotive on the rear end of their train, in preparation for a similar propelling movement.²

Trains 1150 and T296 in readiness for propelling movements

By approximately 0450, T296 and 1150 were both ready to carry out their respective propelling movement towards the stevedore ports, with the second person of 1150 continuing to act as the driver, and the qualified driver acting as the second person.

Both ‘second persons’ were in motor vehicles³ parked ahead of their respective trains, on the rail access road near signal numbers BY29 (for train T296 on arrival road 1) and the adjacent BY31 (for train 1150 on arrival road 2)⁴ (Figure 1). These signals protected the converging track at the stevedore end of the 2 arrival roads.

Figure 1: Trains 1150 and T296 prepare to propel to the Stevedore Ports

Events leading up to the SPAD and collision

At 0605:03 the network control officer (NCO) contacted the driver of T296 on arrival road 1 via the in-cab communications equipment (ICE) radio, to notify them that the stevedores at DP World were ready for their train and that the route and signal indications will be given shortly. T296’s driver relayed this message via handheld UHF two-way radio to the second person controlling their propelling movement at the rear of the train.

At 0605:32 signal BY29 located on arrival road 1 cleared for the passage of T296 towards DP World stevedores. Shortly afterwards, the second person of T296 communicated to their driver that the signal was now indicating proceed and to begin propelling their train. 

The lead wagon on T296 passes signal BY29 indicating proceed at 0606:17. Adjacent signal BY31 for 1150 located on arrival road 2 remained in the stop position (Figure 2).

Figure 2: T296 commencing its authorised propelling movement

Train 1150 SPAD event

The qualified driver of 1150, in their role as second person leading the propelling movement, also observes the proceed indication on signal BY29 on the adjacent arrival road and directs (via their UHF two-way radio) the second person acting as driver to also start propelling 1150. 

At 0606:21, train 1150 begins an unauthorised propelling movement towards signal BY31 at stop on arrival road 2, towards the converging track and T296 which had commenced its authorised propelling movement moments earlier. 

Both motor vehicles are then observed on CCTV footage to be moving simultaneously with the movement of T296, towards the stevedore ports, with both vehicles eventually well ahead of, and facing away from, their propelling trains (Figure 3).

Figure 3: Motor vehicles for 1150 and T296 moving away from the propelling trains

Collision of trains 1150 and T296

At 0607:16, with both trains propelling towards the converging track, the lead wagon on 1150 collided with the side of the fourth wagon of T296 (Figure 4). The lead wagon of 1150 then derailed and traversed the rail access road before coming to a stop against a nearby building. The following 3 wagons on 1150 also derailed, with the second wagon tilting laterally at approximately 30 degrees, causing its container to fall to the ground on the access road Figure 5).

On train T296, the collision caused 4 wagons to derail and the fourth wagon to collide with a yard light pole. This pole then fell to the ground across tracks, triggering track circuit alarms on the network controller’s panel, alerting the NCO to an issue in the yard. 

During the collision and derailment both motor vehicles continued to move in a southerly direction facing away from their propelling trains. 

The second person driving 1150 reported observing a loss of air on the locomotive gauges⁵ during the derailment. As they did not have visibility of the collision from the locomotive cab at the rear of the train, they radioed the qualified driver leading the movement from the motor vehicle to ask if everything was OK. 

The qualified driver reported observing a cloud of dust rising from the train in the rear‑view mirror of their motor vehicle on being contacted by the second person. 

The qualified driver is then seen on CCTV to reverse their motor vehicle rapidly back to the point of collision. They are then seen to exit the motor vehicle, crouch under the derailed wagon that was blocking the vehicle access road and run towards the locomotive on the rear of the train (Figure 6).

Figure 4: Train 1150 colliding with the side of T296

Figure 5: Train 1150 derailing with wagon crossing the access road

Figure 6: Motor vehicle for train 1150 reversed back to collision site

Events post-derailment

In the moments after the initial collision and derailment, the NCO’s Phoenix train control screen displayed numerous track circuit failures caused by the derailment. The NCO contacted the driver of T296 (via the ICE radio) and asked them to confirm their location, stop their train and standby. The driver of T296 confirmed these, and the call was then terminated.

Shortly afterward, the NCO called the driver of T296 back and after reconfirming their location and that they were at stop, advised them to continue movement towards DP World logistics. The driver of T296 replied, ‘no we can’t, apparently the PN train has run into us’. 

The NCO acknowledged the message from T296 and then made an initial attempt to contact 1150 using a standard controller call⁶ to 1150’s ICE radio. This was unsuccessful as the second person driving 1150 had left the locomotive cab to inspect the damage. 

Moments after the NCO’s initial call attempt, an emergency call⁷ was received by the NCO from the qualified driver of train 1150, who had run the length of the train to the locomotive from the motor vehicle to report the incident. The transmission included emergency protocols, such as announcing ‘emergency, emergency, emergency’, identifying the location, the significant number of derailed wagons (Figure 7), clarifying that nobody was injured, and that the rear wagon of 1150 was touching a nearby building. The NCO immediately contacted emergency services and requested their attendance.

Figure 7: Derailed wagons from trains 1150 and T296

Context

Weather conditions

On Friday 13 January 2023, the weather recorded at nearby Sydney Airport was high cloud, light winds, good visibility and a temperature low of 19.3°C through to a high of 27.8°C. Sunrise was at 0557. At the time of the incident at approximately 0600, the temperature was 21°C and the weather was clear. The investigation concluded that weather was not a factor involved in the incident.

Port Botany yard

Port Botany yard is located alongside the eastern shore of Botany Bay in New South Wales, approximately 10 km south of the Sydney central business district. It comprised of multiple stevedore sidings for containerised rail traffic. The sidings included Veolia, Hutchinsons, Patricks and 2 separate siding locations for DP World logistics. 

Port Botany yard was surrounded by the heavily populated and well-established industrial area of south-eastern Sydney (Figure 8).

Figure 8: Port Botany position in relation to the Sydney CBD

General

Since April 2011, the rail infrastructure manager (RIM) responsible for Port Botany was the Australian Rail Traffic Corporation (ARTC). Prior to April 2011, the RIM responsible for Port Botany was RailCorp.⁸

Yard signalling

At the time of the incident, Botany yard signalling predominantly comprised of shunting signals, with rail vehicle detection system (RVDS) in use. 

RVDS used continuous track-circuiting to detect the presence of rail traffic and prevent following rail traffic from entering an occupied section of track. 

Yard design

Port Botany yard comprised 2 arrival and 2 departure roads allowing access to a variety of stevedore ports and private sidings. The yard extended almost 5 kilometres, inclusive of the stevedore port sidings. 

The track layout negotiated around surrounding buildings and roadways utilising tight track curvature and at times narrow track clearances (Figure 9).

The yard had numerous motor vehicle access roads that followed beside rail tracks, allowing rail safety workers to access both track and rolling stock for operational and maintenance purposes. 

However, owing to constraints with established non-rail related infrastructure surrounding the rail corridor, some track clearances were too tight for pedestrian access and some motor vehicle access roads at Port Botany deviated away from rail sidings. 

Shunting operations at Botany therefore required a mix of traditional (walking beside the lead vehicle in the direction of travel) and adapted propelling procedures (line of sight shunting, where in some instances pedestrian access was limited).

On taking control of the yard in April 2011, ARTC completed an upgrade program which installed shunting signals with interlocked motor operated points throughout Botany yard. These replaced existing stop signs/boards and manual point levers. Additional tracks were also added by ARTC to the Port Botany site allowing for greater operational flexibility.

Figure 9: Port Botany yard track layout

Yard arrangements prior to April 2011

Prior to ARTC upgrading the yard, stop signs and manual point levers were predominantly used in lieu of signals and interlocked points (Figure 10 and Figure 11). Train movements were managed by a RailCorp/Rail Infrastructure Corporation NCO located on site. Movement authorities were given verbally via UHF two-way radio. The NCO instructed train crew to manually set routes, and verbally authorised movements past stop signs, manually managing track occupancy and authorising trains to occupy specific portions of track. 

A procedure known as ‘line of sight shunting’ was used to manage movements within the yard. The same procedure was used for both yard configurations, i.e. stop signs and RVDS. 

Figure 10: Port Botany yard track layout

Figure 11: Example of point levers that were replaced by interlocked point motors

Train crew information

Training and assessment for train crew

In addition to the foundational training and qualifications required for their respective roles (see Train crew roles), drivers and second persons are also qualified on specific routes. 

The qualified driver and the second person of 1150 were both assessed as competent for the route between St Mary’s and Port Botany in their respective substantive roles. They were also assessed as competent in the line of sight propelling shunting procedures local to Port Botany yard in their substantive roles. 

These were the only route qualifications held by the qualified driver and second person.

Qualified driver 1150

The qualified driver of 1150 had 3.5 years of rail experience. They had completed the requirements of Certificate IV in Train Driving in July 2022. 

The qualified driver had also been previously qualified as a second person on the route from St Mary’s to Port Botany, and in line of sight procedures at Botany, prior to obtaining their drivers qualification in 2022. 

On the day of the incident the qualified driver signed on for duty at 0300. It was their second day back from 4 consecutive days off. They reported they were feeling well rested. There was no evidence fatigue was a contributing factor to this incident.

Second person 1150

The second person of 1150 had started in the rail industry in June 2021 and received their Verification of Competency (VoC) as a second person on the route and at Port Botany in October 2021. 

On the day of the incident the second person signed on for duty at 0300, it was also their second day back from 4 consecutive days off. They reported feeling well rested. There is no evidence fatigue was a contributing factor to this incident.

Train crew roles

Driver role

While the position descriptions of train drivers may vary between operators, RailTrain’s key purpose, responsibilities and competencies for a driver’s role were as follows:

Key purpose: 

• to drive trains to the operational requirements of the client.

Key responsibilities and duties include but are not limited to:

• comply with all policies and procedures
• operate various locomotives on various freight networks including coal, iron ore, grain, general freight and work trains on mainlines, yards and construction site environments
• operate trains on various gradients utilising train handling techniques to avoid damage
• follow legislative and regulatory values on safeworking practices.

Qualifications and experience include but are not limited to:

• certificate IV in Train Driving TLI42615 or equivalent (including shunting and safeworking)
• dangerous goods and radio protocol training qualifications
• air or electric braking systems experience and/or competency
• shunting experience and/or competency
• train examination experience to respond and rectify minor faults en route to rollingstock.

Second person/driver’s assistant role

Similarly, for the second person the key purpose, responsibilities and competencies were described as: 

Key purpose: 

• to assist the locomotive driver in accordance with organisational policies and procedures and relevant safeworking requirements 
• the position includes preparing and conducting a visual inspection, shunting and marshalling, cross calling signals, roll-by inspections, motor vehicle driving and actioning train inspection results.

Key responsibilities and duties include but are not limited to:

• comply with all policies and procedures
• follow legislative and regulatory values on safeworking practices.

Qualifications and experience include but are not limited to:

• current safeworking accreditation equivalent or willing to obtain
• high level of rail safety awareness
• shunting experience and/or competency.

Train crew 1150 role swap

On the day of the incident, the train crew of 1150 swapped roles at the beginning of their shift at St Mary’s. The second person assumed the role of the driver controlling 1150 and the qualified driver assumed the position of second person en route.

The second person was not qualified to drive the train. The qualified driver supervised the second person driving 1150 on the journey from St Mary’s to Port Botany. However, during the propelling move, the second person continued to drive 1150 without supervision, while the qualified driver led the propelling movement as the second person from within a motor vehicle at the front of the train.

Train information

Train 1150

Train 1150 was a PN intermodal shuttle service scheduled from St Mary’s to Port Botany and return. It comprised of 31 container flat wagons, each 60 foot long, and a single 82 Class locomotive. Some of the wagons were loaded with containers, and some were empty.

Train 1150 was scheduled at the Hutchisons stevedores from 0600 through to 0930, totalling 3 and a half hours to service the 31-wagon train. Upon completion of servicing, the train was scheduled to depart Port Botany at 1202 returning to St Mary’s. It is understood that additional time in the schedule is allocated for trains requiring access to multiple stevedores at Port Botany.

1150 was driven from St Mary’s to Port Botany by the second person, within the speed and signal limitations and for the duration of the shunting activities, prior to and during the collision and derailment. 

Train T296

T296 was a Qube intermodal shuttle service scheduled from Minto to Port Botany and return. It comprised of 30 container flat wagons, a mix of both 40 and 60 foot long, and a single MZ class locomotive. Like 1150, some of the wagons were loaded with containers and some were empty. 

Train T296 was scheduled at the DP World stevedores from 0600 through to 0945 totalling 3 hours and 45 minutes to service 20 wagons of the 30-wagon train. Upon completion of servicing, the train was scheduled to depart Port Botany at 1015 returning to Minto.

Train braking system

General 

On propelling trains, movement is controlled by a driver operating power and brake controls in a locomotive pushing from the rear of the train, with the leading end directed by a qualified worker on the ground, using a two-way radio, instructing the driver on safe movement. 

There were 2 pneumatically operated, ‘airbrake’ systems fitted to the trains involved: 

• automatic brake, which controlled brakes on the entire train, including the wagons and the locomotive
• an independent brake which controlled brakes on the locomotive/s only.

Drivers must be qualified in air brake systems and apply the fundamentals of both systems in various operating environments, to avoid unsafe movements, or damage to the train. 

The use of each of these 2 braking systems depends on the circumstance during which braking effort is required. For both airbrake systems, braking is achieved by the application of brake blocks directly to the locomotive or wagon wheels.⁹

Automatic brake 

For shunting at Botany, the automatic brake is the primary applicable braking system.

Central to the automatic brake operation is the brake pipe, which runs along the length of the train. 

The driver regulates the pressure within the brake pipe through operation of the automatic brake handle in the driver’s cabin. To apply the brakes, the driver reduces pressure in the brake pipe. 

Conversely to release the brakes the driver recharges the brake pipe. Chokes on the wagons slow the release of the brakes to assist with brake pipe recharge. This delays wagon movement by up to 40 seconds each time the brakes release. 

In an environment where brakes are regularly applied and released, such as Port Botany yard, drivers must wait for the brakes to fully release before they can proceed. 

Automatic brake is termed ‘automatic’ because the brakes will automatically apply when brake pipe air pressure is lost, such as in a major derailment or train separation where the brake pipe is broken.

Train handling

Drivers are trained in train handling techniques while obtaining their driving qualifications. 

Train handling is best described as controlling a train in a safe and efficient manner by utilising route knowledge, train driving skills and taking into consideration the total environment. 

The method of control depends on environmental and operational factors such as gradients, allowable train speeds, signalling aspects, locomotive capabilities, train weights and lengths, and whether the train is on the main line or shunting in a yard.

On main lines generally the margin of error is broader than that of a yard environment. For example, drivers aim to stop their train some distance from a red signal, this distance creates a margin of error for drivers to avoid passing a stop signal.

Train movements in yards such as Port Botany require more precision, as coupling and uncoupling of vehicles may be required. 

While shunting crew are in close proximity to the train to carry out coupling and uncoupling tasks, it is critical that communication requests from the crew member on the ground are acted upon by the driver accurately and in a timely manner, to avoid asset damage and personal injury. As the driver may have limited visibility of the movement.

This precision is gained through appropriate knowledge, experience and training in driving tasks and should only be carried out by those rail safety workers qualified to drive trains.

Line of sight shunting at Port Botany

General

A procedure known as ‘line of sight shunting’ was used within Port Botany yard as described in ARTC operational document OGW-30-25 Network Information Book. This procedure informed operators how to perform propelling movements specifically within Port Botany yard. 

A propelling movement (Figure 12)is described as a locomotive pushing rail traffic in a direction away from the controlling locomotive and occurs after the locomotive performs a run around manoeuvre and is placed on the end of a train consist. 

Line of sight shunting was unique to the Port Botany environment in that there were site‑specific variations to standard propelling procedures, such as the use of motor vehicles to assist the shunter leading the propelling movement to navigate the yard, and communication requirements.

History of propelling procedures

ARTC advised that Port Botany yard was designed for train push pull operations¹⁰ where a locomotive is attached at either end of a train consist (Figure 13). This configuration eliminated the need for locomotives to perform a run around manoeuvre to place the locomotive(s) on the opposite end of the train to enable departure in the reverse direction. 

Despite the yard being designed for push pull operations, alternate propelling operations had become widely adopted by the operators. Several reasons were provided for this such as:

• operators had not moved to push pull operation as it was resource heavy, requiring an additional locomotive and crew. Instead, most operators had opted for single locomotive operations and shorter trains
• an incident in one of the stevedore ports in 2003 resulted in some stevedore’s banning locomotives remaining coupled to wagons in their terminals.

Standard practice had therefore become to propel trains rather than lead them into the stevedore terminal and run around for departure. In addition, not all stevedore sidings had rail infrastructure to support run around manoeuvres and propelling allowed locomotives to easily be uncoupled and moved clear of hardstand areas once wagons were placed for servicing. 

Figure 12: Propelling train configuration

Figure 13: Push pull train configuration

History of line of sight shunting procedure

Line of sight shunting procedures specific to Port Botany were inherited by ARTC in 2011 from the previous rail infrastructure manager, RailCorp/Rail Infrastructure Corporation (RIC), with the same procedures remaining in place after the ARTC yard upgrades.

These were introduced following a fatality in 2004 where a shunter fell from a moving wagon whilst propelling in the yard.¹¹ The rules at the time allowed shunters to ‘safely ride’ on rolling stock when directing propelling movements. This incident led to the rail regulator at the time prohibiting the practice. 

This prohibition necessitated an alternate method to be developed to lead a propelling movement. The traditional method involved shunters walking beside the leading end of their train in the direction of movement. At Port Botany the RIM identified specific site risks with this standard procedure including tight track clearances and the distance of the shunt. This led to the introduction of the unique line of sight shunting procedures at Port Botany, which included the use of a motor vehicle. 

OGW-30-25 Network Information Book – Line of Sight Procedure

The line of sight shunting procedure established unique yard protocols on how competent workers were to propel trains with the use of motor vehicles within the confines of Port Botany yard as follows:

• The Competent Worker controlling the shunting movement must gain authorisation from the ARTC Network Controller in Network Control Centre South (NCCS) before commencing any Line-of-Sight Shunting movements

• The Competent Worker must ensure all points are set for the intended route

• Prior to commencing the movement, the Competent Worker must proceed from the rear of the train to the line-of-sight location

• The Competent Worker must then instruct the driver to propel towards the line-of-sight location

• The Competent Worker must always maintain sight of the lead vehicle of the movement

• The movement must be stopped at the line-of-sight location

• Once the movement is stationary at the line-of-sight location, the Competent Worker must then proceed to the next line-of-sight location

• Repeat steps 3 to 7 for each successive line-of-sight location until the propelling movement is completed

• Any issues associated with the use of line-of-sight shunting must immediately be reported to the Network Controller in NCCS.

ARTC standard shunting procedures

General

ARTC’s procedural requirements for shunting in yards were in ARTC’s general network rules and procedures. These rules and procedures described the standard requirements for responding to signals, propelling trains, and communication protocols that must be followed by both train crew and NCOs.

In addition to the standard procedures, shunters and network controllers at Port Botany were also required to apply the local rules outlined in ARTC’s document OGW-30-25 Network Information Book (NIB). This NIB contained the local line of sight shunting procedures and the specific rules and procedures for propelling in Port Botany yard. 

ARTC rules and procedures

Signals

Signals were principally designed to ensure the separation and regulation of rail traffic and provide competent workers with information on the status of the line ahead and the route that is set. 

The shunting signals (Figure 14) used at the incident site in Port Botany yard were identical except for the signal numbering and were situated in line with one another on each adjacent arrival road. 

Figure 14: Shunting signal BY29 at stop in Port Botany post-incident

The general rules associated with responding to signals were contained within ARTC’s rule ANSG 606 Responding to signals and signs. 

This procedure described that a signalled proceed indication authorises the shunting movement and means:

• interlocked points protected by the signal are set in the correct position for the movement and no conflicting route has been set
• movement past the signal is now authorised.

In addition to the general rules above, the local line of sight shunting procedure for Port Botany stipulated that the competent worker controlling the shunting movement must gain authorisation from the ARTC NCO in NCCS before commencing any line of sight shunting movements. 

The line of sight procedure however did not provide guidance on how the authority from the NCO was to be used in conjunction with a signalled proceed authority. 

Signs

Prior to ARTC taking over as the rail infrastructure manager for Port Botany in April 2011, train movements were managed in the yard by stop signs and manually set points and routes.

In accordance with ARTC’s rule ANSG 606 Responding to signals and signs, a stop sign may only be passed if ‘verbal authority’ has been given to proceed.

In addition, where signage is used it is a requirement in ARTC’s procedure ANTR 420 Shunting and marshalling, that competent workers:

• ensure that routes are correctly set and safe for movements 
• make sure that it is safe to shunt 
• make sure that workers have been warned about the intended shunting
• tell signallers when shunting movements within the signaller’s area of control have been completed.

These requirements contrast with signalled proceed authorities which provide the competent worker with confirmation that the route has been set by the NCO, and it is safe to proceed.

Propelling movements

Propelling movements on the ARTC network were also subject to general rules and procedures and were described in ARTC’s procedure, ANTR 424 Propelling trains. 

This rule stated that propelling movements:

• must be directed from the leading end by a competent worker 
• and that the competent worker directing propelling must safely walk beside the leading vehicle.

Local line of sight shunting procedures at Port Botany differed from the standard procedures in that, rather than walking beside the leading vehicle as it propels, the procedure called for the competent worker to move ahead of the stationary train, to a line of sight location via the use of a motor vehicle.

The competent worker then authorised the driver to propel to their location whilst visually monitoring the rear of the train from a stationary position as it progressed. The procedure then stated that the train was to stop and wait until the competent worker moved to the next line of sight location. This was to be repeated until the train reached its destination. 

Unlike the standard propelling procedure, the line of sight procedure allows a motor vehicle to be used to transport the competent workers between the line of sight locations.

Responsibilities of crews

Rules and procedures also existed to outline the general responsibilities of train crew.

ARTC rule ANGE 232 responsibilities of Rail Traffic Crews states that it is the primary responsibilities of rail traffic crews to operate trains and track vehicles safely and efficiently through the ARTC network.

For drivers of rail traffic crew specifically, the relevant rules state that they must:

• be qualified to operate the rail traffic they drive in the ARTC Network, and

• as necessary, be qualified in the systems of Safeworking relevant to their area of operation, and

• delegate duties only to qualified workers.

The qualified driver of 1150 on the day of the incident delegated the driving duties to the second person, on the journey from St Mary’s to Port Botany, and whilst they were propelling in Botany yard. 

Responsibilities of rail safety workers

A rail safety worker is defined by Rail Safety National Law (RSNL) as an individual who carries out or is about to carry out rail safety work. Rail safety work includes driving, despatching, controlling or affecting the movement of rolling stock. 

Rail safety workers have general duties under Subdivision 2 Duties, s56 of the RSNL which includes:

(1) A rail safety worker must, when carrying out rail safety work—

(a) take reasonable care for his or her own safety; and

(b) take reasonable care that his or her acts or omissions do not adversely affect the safety of other persons; and

(c) comply, so far as the worker is reasonably able, with any reasonable instruction given by the rail transport operator to allow the operator to comply with this Law.

The ARTC ‘line of sight procedure’ formed part of ARTC’s accredited safety management system, and it was the responsibility of all rail safety workers involved in controlling movements in the yard to comply with these procedures under RSNL. 

Methods of communication at Port Botany

ICE (in-cab communications equipment) radio system

The ICE Radio communication system was developed by a Queensland company named ‘base2’ and is a modular system designed specifically for the rail industry. 

The base2 ICE radio system which is located within locomotive drivers compartments, provided the train crew with multiple communications options (Figure 15), including:

• establishing, broadcasting and receiving emergency calls across the network
• calling and receiving direct communications from the Network Control Officer
• establishing multiple channel UHF communications
• text message receival and acknowledgement sending
• receiving electronic authorities in the form of train orders and conditions affecting the network (CANs).

Figure 15: In cab communications equipment radio (ICE)

Operator closed/discreet channel UHF two-way radio system

Each rail operator had their own discrete UHF closed radio channel, that was not recorded or broadcast to other operators in the area. The UHF feature on the ICE radio allowed the driver in the locomotive cab to have contact with the competent worker leading shunting movements at Port Botany.

UHF was primarily used for communications between train crew of the same operator during critical coupling and uncoupling duties and other duties such as propelling. The discreet channel lessened the likelihood of disruption during these task critical communications.

Mobile phone usage

Rail operators in most instances also had work issued mobile phones for use in an operational environment such as Port Botany. These may have been a personally issued phone or mobile phones issued to crew at start of shift to assist in communications for the task required. 

Competent workers leading propelling movements at Port Botany did not have access to ICE radio communications when leading the movement from the ground, therefore relied on both their hand-held UHF radio and mobile phone for communications.

Each operator had guidelines on when and where mobile phone usage could occur.

Network control for Port Botany

General

Botany yard was remotely controlled from ARTC’s network control centre south (NCCS) located at Junee (485.500 km), in the Riverina region of south-western New South Wales. 

The Sydney 1 network control officer (NCO) at the NCCS, was responsible for managing rail traffic in the Port Botany yard. 

Phoenix train control system

The Phoenix train control system provided the NCO with a graphical user interface, through which they could control signalling equipment at remote locations. The functionality included setting signals, points and routes. 

The system used the track circuits from the rail vehicle detection systems (RVDS) in the yard, with each section of track showing on the controller’s screen as either occupied or free. 

The Phoenix system display for Port Botany below (Figure 16) showed the 2 adjacent trains on the arrival roads, with signals BY29 and BY31 both set to stop prior to the incident. 

Figure 16: Phoenix replay prior to route being set on signal BY29 for train T296

ARTC network control procedures

NCO’s for Port Botany were also required to apply a mix of both the standard rules and procedures and local procedures where applicable to Port Botany.

ARTC Network Information Book Sydney 1 board

Train crew and NCO’s had available for reference the ARTC Network Information Book (NIB) for Sydney 1 board, Botany yard to Enfield South (OGW-30-25). This document included the line of sight shunting procedure in section 2.1.1 (see below), and contained other relevant network information such as:

1. General Information describing infrastructure:

• board control boundaries of control 

• type of Safeworking system

• applicable Rules

• maximum Permanent Speeds and Permanent Speed Restrictions

• communications

• drawing Legend

2. Locations and Sections Information:

• Port Botany Yard, includes line of sight shunting

• Cooks River Loop

• Cooks River

• Marrickville Junction

ARTC network control officer training and assessment for Sydney 1 board

Similar to train crew, following general training and assessment, new NCO’s undergo additional training to become qualified for specific areas of control, otherwise known as panels or boards. This is a requirement as general network rules and procedures do not necessarily cover all the essential information needed to manage each specific area.

For Sydney 1 board at NCCS, the NCO would be required to prove competency in a variety of Port Botany specific processes and procedures such as:

• following local shunt procedures
• identifying and following local special instructions
• using local radio protocols

ARTC train control auditing

NCCS had processes in place to monitor NCOs daily for procedural compliance. Audits were conducted at random, on a structured list of network control procedures. However, there were no compliance activities within the standard audit program for Sydney 1 Board specifically related to Port Botany local line of sight shunt procedures.

Accessing rules and procedures

ARTC published all applicable rules and procedures for its network on its website (Figure 17).

Figure 17: Australian Rail Track Corporation procedures home page

ANGE 212 Network Information Publications prescribed the rules about access to Australian Rail Track Corporation (ARTC) NSW network information.

ANGE 212 Network Information Publications

This rule states that competent workers must have access to the following publications that relate to their area of work in the ARTC Network:

• network rules
• network procedures
• network forms
• Network Information Books (NIBs)
• electrical safety instructions
• Train Alteration Advice (TAA)
• Route Access Standard (RAS)
• Train Operating Conditions (TOC) manual
• TOC waivers
• SAFE notices
• speed restriction notices.

Competent workers must be provided with relevant information about changes to the above publications that affect their work duties and responsibilities. 

Organisations employing competent workers in the ARTC network must make sure that they provide workers with information and ARTC network publications relevant to their duties.

Competent workers must read and use the information in relevant publications to do their work.

The Network Information Book for Port Botany was also located on the ARTC website and was accessible to rail operators. 

The network control board controlling specific areas of track was displayed, along with the relevant NIB book, NIB diagrams and line diagrams (Figure 18).

Figure 18: Australian Rail Track Corporation network information books home page

Shunting rules and procedures at other ARTC locations

Leightonfield, Sydney, NSW

Leightonfield yard was located 26 km west of the Sydney CBD and was controlled by Network Control Centre South (NCCS) Sydney 2 board located in Junee, NSW. 

As line of sight shunting was exclusive to Botany yard, standard ARTC network rules applied to Leightonfield yard as described in Network Information Book OGW-30-26. 

Motor vehicles were not used for shunting movements at Leightonfield yard.

Kooragang Island, Newcastle, NSW

Kooragang Island yard was located approximately 6 km north of the city of Newcastle, NSW. This location was controlled by the Kooragang NCO located within Network Control Centre North (NCCN) situated in the Hunter Valley near Newcastle, in northern NSW. The Kooragang NCO was responsible for controlling the track in the immediate area of the Kooragang Island coal unloading facility through to Sandgate (exclusive). 

Like Port Botany, a specific procedure existed for propelling trains in the wrong running direction at the Kooragang departure sidings. This procedure was described in Network Information Book OGW-30-13 located in the Hunter section of the ARTC website. 

At this location the NCO activated warning bells and cleared the rail discharge shed of employees at Kooragang Island prior to the propelling movement being carried out by competent workers. Competent workers controlling the movement utilised the standard method of propelling a train as set out in ARTC procedure ANTR 424 Propelling Trains. 

Motor vehicles were not used for shunting movements at Kooragang Island.

Human error

Incidents that involve an element of human error, such as the collision at Port Botany, typically involve individuals, or a group of individuals, assessing a task and the environment in which the task is to be done. This assessment then leads to decision‑making and actions that can unintentionally contribute to an incident, or an incident outcome, either directly or indirectly. 

For investigators to understand and explain why individuals took certain action, and to help identify ways to prevent incidents from happening again, the science of human factors is frequently employed.

Error types

Human factors is a science that studies the capabilities and limitations of human performance. 

Professor James Reason was a renowned expert in human behaviour, human factors and organisational safety. His research into human error and its involvement in incidents, led to clear differences being distinguished between intended and unintended actions, which were then broken down into distinct error types as illustrated in Figure 19).

Figure 19: Human factors error types

Unintended actions

Reason’s model describes that unintended actions are typically skill-based errors. 

Generally, when these types of errors occur, the individual has the right knowledge, skills and experience to carry out the task properly, thus they are defined as skill-based errors. 

The crew were qualified for operations within Port Botany yard, with operations in the yard forming part of their daily routine. 

Skill-based errors typically occur during highly routine or familiar activities, when our attention is diverted from a task, either by our own thoughts or external factors. 

They can be divided into 2 main categories:

• Slips, which are errors made when we don’t pay attention, or where we have the correct intention or plan in place, but it is incorrectly carried out.
• Lapses are a slightly different error in executing a plan, in that we fail to carry out an intended action entirely, usually due to memory failure.

Intended actions 

Reason’s model describes that intended actions are typically skill-based failures, rather than errors, and that these can also be divided into 2 main types:

• Mistakes are distinct and “intentional”. They occur when we plan on doing something and carry out our plan correctly, however it does not produce the outcome we wanted. Often this is the result of inadequate knowledge or deciding that the rules applied were inappropriate.
• Violations are a different form of intended action as they involve deliberate deviations from the rules, procedures or values. Generally, some of the reasoning behind such acts can be described as:
• workers thinking rules don’t apply to them
• being under time pressure (both actual and perceived)
• perception they won’t get caught
• lack of understanding of why the rule is in place
• perceptions that rules are too strict or unnecessary
• taking the easy option
• peer pressure.

Violations

Violations can also be divided into 2 main types:

• Routine violations are commonly described as “no one follows that rule”. It occurs over time due to habitual practice, where individuals or groups of individuals believe the rules to be unnecessary leading to the violation being considered commonplace, and committed by most of the workplace 
• Situation violations can occur due to factors such as time pressures, inadequate equipment, lack of resources or poor planning. Generally, this occurs as a “one-off”, however if not addressed may become routine over time. This is indicative of a worker not following a rule because of a perceived lack of time to perform a specific task and attempting to complete something as quickly as possible.

Biases

Bias is another human factors concept. Biases are described as often unconscious errors in decision making. Humans will consider the immediate circumstances but then reach a decision based on, not only the immediate information at hand, but their own knowledge, beliefs and previous experiences. 

Expectation bias for example occurs when we interpret information in a way that confirms our pre-existing beliefs and/or expectations based on past experiences.

Information processing

Information processing is another important part of the decision-making process. A number of human factors concepts such as ‘automatic pilot’, and ‘assumptions’ can play a significant role in the way we process information and make decisions. 

Automatic pilot

Automatic programming or ‘automatic pilot’ is a key attribution to being human, in that we have the ability to transform complex tasks into skills we can apply automatically and at will when something is done regularly. However, this skill can also lead to errors when applied inappropriately in a situation that is not the same as usual. 

Prior experience

We may also over rely on our prior experiences and consequently may make incorrect assumptions when processing the situation and evidence in front of us. Prior experience may lead to us not always considering variability in circumstances and evidence.

Risk management

General

For the purposes of the Rail Safety National Law (RSNL), a Rail Transport Operator (RTO) was required to ensure the safety of its railway operations ‘so far as is reasonably practicable’.¹² The Office of the National Rail Safety Regulator’s (ONRSR’s) guideline Meaning of duty to ensure safety so far as is reasonably practicable, provided RTOs guidance on how to apply this concept to their operations. Within this guideline, ONRSR considered ISO 31000 (Risk management – guidelines) as ‘good practice’ in the management of risk, in addition to requirements of the RSNL. ISO 31000 described risk management as:

…the systematic application of policies, procedures and practices to the activities of communicating and consulting, establishing the context and assessing, treating, monitoring, reviewing, recording and reporting risk. 

RTOs were required to implement a safety management system (SMS) to comply with their risk management obligations. This was to include formal processes for the identification, assessment, control, monitoring and review of risk.¹³

Rail Safety National Law (RSNL) {NSW}, Section 50 states rail safety is the shared responsibility of:

• rail transport operators; 

• rail safety workers; 

• other persons who – 

• design, commission, construct, manufacture, supply, install, erect, maintain, repair, modify, or decommission rail infrastructure or rollingstock; or

• supply rail infrastructure operations or rollingstock operations to rail operators

• and ONRSR.

In relation to the above it was the responsibility of all involved in yard operations to effectively manage risk in Port Botany yard.

ARTC risk control monitoring and review

RSK-PR-001 Risk Management Procedure documented ARTC’s risk management framework and risk matrix. This document described risk levels and the associated review periods as shown in Figure 20:

Figure 20: ARTC risk level vs risk review period

This document also described how review of risk controls should take place soon after implementation, and the intervals appropriate for risk level, anticipated adequacy and effectiveness of the risk controls. 

For Port Botany, the line of sight procedure was highlighted as a specific risk control, not in use anywhere else on the ARTC network.

April 2020 ARTC risk assessment 

Following an almost identical incident that occurred on 5 March 2020 at the same location in Port Botany yard, ARTC facilitated a risk assessment on operations within the yard on 30 April 2020, referencing its risk management procedure. Representatives from 4 operators and ARTC were in attendance. 

The ARTC risk assessment took into consideration investigation reports on previous incidents at Port Botany and considered the constraints to operating in the yard.

The subsequent report identified the following risks and consequences and tabled risks to be either accepted, rejected or investigated. 

Identified risk:

• train to train collision
• train derailment
• train to person incidents.

Identified consequences:

• Signals Passed At Danger (SPAD’s) 
• damage to assets 
• delays in operation within Botany yard
• increased oversight by regulatory bodies
• damage to ARTC’s reputation. 

The risk assessment used a combined risk rating method, with the overall risk rating taking into consideration combined events. The risk at Port Botany assessed as ‘High 3B’, based on a ‘moderate’ consequence and a ‘likely’ likelihood. 

Proposed treatments were identified as shown in Figure 21:

Figure 21: Proposed Treatments from ARTC risk assessment 2020

The proposed treatment table provided by ARTC was observed to be incomplete, displaying blank fields under Accept/Reject/Investigate for management of trains in Botany Yard, and mandate push pull operations, although the introductory text suggests that the proposed treatments may require further investigation prior to acceptance. The document does not display any future review date.

OTSI requested evidence of all risk assessments conducted by ARTC from the date of takeover as the rail infrastructure manager on 25 April 2011 through to the incident date of 13 January 2023. 

ARTC’s safety interface agreement with Pacific National dated 23/05/2018 notes in Schedule 2 that a high-level risk assessment was carried out in May 2012 and reviewed in May 2018, however this assessment related to network‑wide operations and did not include line of sight operations at Port Botany.

At the time of the incident the only record of risk assessment for Port Botany was of the one conducted on 30 April 2020. And following enquiries it was also unclear what subsequent action had been taken arising from this assessment beyond the proposed duplication project.

Communication and consultation

A key component of risk management processes was communication and consultation with internal and external stakeholders. ISO 31000 described the aims of communication and consultation as bringing together different areas of expertise and views, and to gather sufficient information during the risk management process. It was an integral part of:

• establishing the scope and context within which the risk to be managed resided
• identification, analysis and evaluation of the risk
• identification and implementation of effective risk controls.

A companion handbook HB 327:2010 (Communicating and consulting about risk) applied to ISO 31000¹⁴ described the benefit of effective communication and consultation as ‘…a shared and better understanding of the risks faced and the range of treatment options.’ It further noted that consultation would help to comprehensively identify risks, increase acceptance of implemented controls and encourage feedback of effectiveness after implementation.

The importance of consultation was recognised within the RSNL, with an objective of the Act ‘to promote the effective involvement of relevant stakeholders, through consultation and cooperation, in the provision of safe railway operations.’ 

To achieve this, the RSNL advised that participation and consultation in establishing, reviewing or varying an RTO’s SMS (risk management), should include:

those that may be affected by the SMS, such as rail safety workers 

health and safety representatives (where applicable)

relevant unions

other RTOs (where interface agreements were required) 

the public (where appropriate).

It was a requirement of the Rail Safety National Law National Regulations 2012 that an RTO’s SMS contained systems and procedures to ensure this consultation occurred.¹⁵

A large variety of stakeholders were recorded to be in attendance at the risk assessment facilitated by ARTC in 2020. Of the train operator attendees, one member had previous experience from an end user cohort and one had current operational experience at Port Botany as a second person. The other attendees were described as ‘Train Operator’ and principally held management roles. 

Safety interface coordination

General 

Rail Safety National Law National Regulations 2012 required that an RTO’s SMS contained systems and procedures for safety interface coordination.

Clause 22 in Schedule 1 of Rail Safety National Law National Regulations 2012 prescribed that safety interface coordination between parties was to include:

1. Procedures for the identification of interface risks to the safety of railway operations and for the development and implementation of interface agreements in accordance with Part 3 Division 6 Subdivision 2 (Interface agreements) of the Law.

2. Procedures for monitoring the implementation and effectiveness of and compliance with interface agreements.

ONRSR’s guideline safety management system describes safety interface coordination as: 

(v) safety interface coordination – this includes systems to ensure that where risks occur at or arising from an interface, the responsibility for risk controls is appropriately assigned and understood by all those with a role in the implementation of the control as well as monitoring compliance to the implementation. 

In addition, the guideline advises each party had a responsibility to identify and assess risk. And interface coordination agreements should cover how parties would effectively communicate to manage their interface risks. 

ARTC and PN safety interface agreement

Document IA4019 – Interface Agreement between Pacific National and ARTC was the interface agreement between ARTC and PN for PN operations on the ARTC network, including Port Botany yard. 

Section 5 - Identification, assessment and management of risk, states: 

…the parties acknowledge and agree that they have identified and assessed risks to safety (or will endeavour to do so as soon as is practicable) that may arise in relation to the Interfaces and the results of this assessment are set out in Schedule 2. 

Section 5.6 clarifies that those persons set out in Schedule 2 under the heading ‘Party’ are responsible for implementing, maintaining and monitoring the performance of the applicable risk control measures, and evaluating, testing and revising the risk control measures as applicable. 

The extract from Schedule 2 in Figure 22 illustrates the nominated responsible parties for derailment risks. There was no risk of collision identified in the table. 

Figure 22: Extract from network safety interface agreement schedule 2

Section 6 of the interface agreement, covering Monitoring and review of the risk, stated that the nominated risk review representative will, from time to time as is reasonable, review and monitor the compliance by that party of its obligations under the agreement. And if following a party’s review and monitoring the safety risk is considered to be intolerable the parties will work collaboratively to resolve. 

Section 11 titled Compliance, stated that each party will report annually to the other party, or at such other times that the parties may agree, on its progress in implementing agreed safety management measures. 

ARTC compliance monitoring

ARTC was requested to supply evidence of compliance monitoring of both operators and network control officers (NCO’s) at Port Botany relating to line of sight shunting procedures. 

ARTC network control 

Internal to ARTC, NCOs were subject to random audits on processes and procedures to do with communication and adherence to procedures. 

ARTC had refresher training in place to reiterate procedures for NCOs who may take extended leave. This refresher training and examples of audits were provided to OTSI and showed areas for improvement identified by assessors and how network controllers were coached to rectify these process shortfalls. 

However, the only reference to specific ARTC control board procedures entailed general communication protocols and nothing specific to line of sight procedures for the Sydney 1 Board NCO for Port Botany. 

Operator compliance monitoring

ARTC advised that the operators were responsible for the compliance monitoring of the line of sight shunting procedure and that ARTC entrusted that adequate training of the procedure was being carried out by the individual operators that accessed Port Botany yard. 

ARTC advised that whilst it had no specific procedures for monitoring compliance, it had conducted monitoring through face-to-face conversations within the yard and random audits witnessing correct shunting movements being carried out. However, these were not documented. 

Post‑incident, ARTC conducted compliance audits of operators carrying out the line of sight shunting procedures. These audits were pre-arranged with train crew having knowledge of the audit being conducted.

Also post‑incident, ARTC requested that operators confirm that they both educate their rail safety workers and comply with the line of sight shunting procedure at Port Botany. 

PN responded as follows: 

1. PN can confirm that line of sight shunting requirements are included in the PN‑PRO-‑OPS Shunting Procedure and all current PN workers qualified for the Port Botany Yard have been taken through the OGW 30-25 procedure. The PN Training Standard outlines the requirement to complete Verification of Competency (VoC) assessments periodically and specifically RSW-VOC-0018 is the process used to verify knowledge and application of the PN-PRO-OPS Shunting procedure including the line-of-sight aspects. Records are kept in the PN Learning Hub system when completed. PN can confirm all PN workers currently used for Port Botany operations have current VoC’s.

2. PN can confirm that the Critical Risk Management (CRM) Framework used to monitor PN worker behaviours and compliance to required PN procedures is being applied at Port Botany. Part of the CRM process is the requirement to complete infield Critical Control Verifications (CCVs) with PN workers and in relation to line-of-sight shunting, this is specifically addressed by the completion of “Struck by rollingstock” CCVs at Port Botany. For example, PN can confirm that since the 20^th January 2023, 27 CCV’s have been completed on PN workers confirming that line-of-sight shunting is completed as per the PN Shunting Procedure and OGW 30-25.

It was unclear how these procedures were applied to labour hire crew.

Training of train crew in the line of sight procedure 

General 

Driver trainers or mentor drivers from each operator typically provide trainees with instruction on the Port Botany line of sight procedures. 

Initially the trainee learns the theory behind the procedure and then is placed with competent crew to learn the procedure in an operational environment. Once they are comfortable executing the procedure for themselves, whilst supervised, the trainee requests a practical verification of competency (VoC) to prove to an assessor that they can properly execute the procedure. The VoC typically involves an assessor assessing the trainee on components such as communication protocols, safely carrying out the work, etc. Once the trainee is deemed competent, they can then be rostered without supervision aboard Port Botany services.

VoC’s for shunting in the yards was also carried out by the operator’s assessors at scheduled intervals in accordance with their safety management system to verify that the competency of the crew operating at Port Botany remained adequate post initial assessment and verification. 

Compliance monitoring

During interview, competent crew from 3 different operators in Botany yard were asked if they had witnessed any compliance monitoring of the line of sight procedure from either their employer or ARTC. Of the crew interviewed none had witnessed any form of compliance monitoring by ARTC. 

The only compliance audits that crew witnessed at Port Botany were from an employer’s perspective in the form of what is known by some operators as a ‘check ride’. A check ride is best described as a driver trainer attending the shift of both train crew on board a train service and checking/auditing compliance or carrying out a VoC. The compliance check is carried out whilst the crew conduct their normal duties and results are fed back to each employee for any potential areas of improvement and/or recertification to continue normal duties.

Often the crew being assessed were forewarned of the impending check ride or VoC, giving time to prepare and conduct the line of sight procedure properly. 

Crew reported they carried out line of sight shunting correctly, as per procedure during these check rides and VoCs. One driver commented that they would also carry out the procedure correctly if the other crew member wanted to.

Reporting of issues with the line of sight procedure

Section 9 of the OGW-30-25 Network Information Book – Line of Sight Procedure states:

9. Any issues associated with the use of line of sight shunting must immediately be reported to the Network Controller in NCCS.

The procedure documents an opportunity for crew to notify the NCO of any issues identified with line of sight shunting. 

ARTC advised it had not received any reports of problems with the line of sight procedures. 

Previous collision and derailment March 2020

On Thursday 5 March 2020 at 1355, intermodal SSR Train T170 was positioned on arrival road 1 in Port Botany yard, with intermodal Qube Train T254 positioned on the adjacent arrival road 2. 

Both trains were awaiting proceed signals to conduct propelling movements to their respective stevedore ports. The crew controlling the propelling movements of T170 and T254 were positioned within motor vehicles alongside the lead wagons of both trains, on the rail access road near signals BY29 (arrival road 1) and BY31 (arrival road 2).

ARTC’s incident investigation report described Qube Train T254 propelling and passing signal BY29 in the stop position, simultaneously with SSR Train T170 as they began an authorised propelling movement past signal BY31. These 2 simultaneous propelling train movements resulted in the collision of the rear of both T170 and T254 and subsequent derailment of 3 SSR wagons and 4 Qube wagons. The leading wagon of Train T170, whilst derailed, came to rest across the rail access road close to a privately‑owned building adjacent to the rail corridor (Figure 23).

ARTC’s report identified contributory factors for this event were:

• Botany Yard is designed for train push pull operations, however these push pull operations are not in use. Stevedore enforced restrictions are currently in place on locomotives being attached to wagons being serviced within port hardstand areas. This restriction requires trains to perform propelling movements into the port hardstand areas. 
• The competent worker for Train T254 misidentified their location and the clearing of the incorrect signal for their intended movement from within a motor vehicle. 
• There were no formal communications protocols applied between the shunter of train T254 and the NCO. At no time were formal communication protocols applied such as stating full names, train service identification, train location, or the signal number. 
• Finally, the line of sight shunting procedure in place was not adhered to with the violations identified as train crew propelling trains from within motor vehicles and failing to bring propelling trains to a halt per line of sight shunting procedures. 

Regulatory action – March 2020 collision and derailment

Following the above incident that occurred on 5 March 2020, the Office of the National Rail Safety Regulator (ONRSR) requested and received the rail transport operator's internal investigation report¹⁶ conducted as part of its safety management system. This report was obtained through ONRSR's regulatory interactions with the operator.

In November and December of 2021, ONRSR conducted site visits at Yennora, Cooks River (near St Peters) and Minto, New South Wales. The site visits that took place were unannounced and focused on yard derailments and runaways. 

Figure 23: Image from March 2020 Port Botany derailment

ARTC investigation for collision and derailment January 2023

ARTC conducted and supplied its own investigation report into the later 13 January 2023 collision and derailment. This report included the following relevant factual information:

• the investigation review of the NCO voice recordings revealed there were no communications conducted between RailTrain Services train 1150 and the NCO prior to the incident.

• the individual who was operating train 1150 at the time of the collision was assigned as the 2nd person for the train and has admitted that they were not a qualified driver and had not completed training. They had accepted an offer by the qualified driver to drive the train of 1150. 

• the qualified driver of train 1150 had stated that due to the length of time waiting for the authority to propel the train into the terminal, they had forgotten which was their train and on which track, and had incorrectly identified the proceed signal as the one for their train and has given a proceed to the driver.

• the qualified driver has lost situational awareness due to time waiting for the authority to proceed and has also been subject to confirmation bias due to previous movements being from arrival road 1

• additionally, the CCTV footage does not show at any time the crew leading either service checking the route (points) prior to the movement of either train.

Regulatory action – January 2023 collision and derailment

Following the incident on 13 January 2023, the Office of the National Rail Safety Regulator (ONRSR) commenced a compliance investigation of involved rail transport operators. Upon conclusion of the ONRSR compliance investigation, findings were considered in the conduct of additional ONRSR regulatory activities¹⁷ at Port Botany focusing on observing shunting and stabling operations.

ONRSR conducted 3 compliance inspections at Port Botany Yard in 2023, followed by a further compliance inspection in February 2025. A site visit was conducted in October 2025 to observe line of sight shunting operations at Port Botany. Safety improvements from a previous inspection were verified in an additional compliance inspection conducted in November 2025.

SPAD management

A signal passed at danger (SPAD) is when rolling stock passes a stop indication or exceeds authorised location limits. 

The qualified driver of 1150, in their role as second person leading the propelling movement, observed the proceed indication on signal BY29 on the adjacent arrival road, and directed the second person of 1150 acting as driver to start propelling 1150 past the stop indication on their signal BY31, resulting in a SPAD. 

Rail Industry Safety Standards Board (RISSB) SPAD management guidelines provided guidance to operators on the minimum requirements for managing SPADs and proceed authority exceedance events, to reduce the likelihood of recurrence and minimise the potential consequences as follows.

Local conditions

Local conditions have a significant impact on human performance and therefore may be a major contributor to SPAD incidents.

For example, signal sighting decisions taken by an infrastructure manager can contribute to a more hazardous work environment for an operator. If the rail infrastructure manager does not engage the operators in decisions about signal sighting, they may inadvertently overlook risks. 

Design risk should be controlled through proactive risk management and system safety processes. Similarly, human factors should be systematically considered and integrated into design decisions.

Signal sighting standards are commonly based on the viewing of signals from locomotives. 

Reacting to the wrong signal

Some SPADs have been attributed to the driver reacting to the wrong signal or reading across through to a non-target signal. This error occurs when a target signal has been stopped at correctly, however due to other distractions such as shunting duties etc, the driver responds to a change in aspect on a non-target signal that is also in view. 

Organisational factors

Organisational factors are conditions associated with business and operational systems that may have an impact on working conditions and workforce performance. Impractical procedures, which lead to frustration, may contribute to violations. 

Organisational factors usually occur some time before the actual SPAD event and are usually only identified in systemic investigations. 

Organisational factors are typically controlled through the implementation of an effective SMS. 

One organisational factor that should be carefully considered is the effect key performance targets can have on staff performance and SPAD risk.

Figure 24: View of signals BY29 and BY31 from access roadway adjacent to arrival roads 1 and 2, Port Botany

Safety analysis

Introduction

On 13 January 2023, Pacific National train 1150 collided with Qube logistics train T296 in Port Botany yard, New South Wales. Train 1150 was a single locomotive, 31 wagon intermodal shuttle service from St Mary’s to Port Botany and return, crewed by RailTrain. Train T296 was also an intermodal service with a single locomotive and had 30 wagons, operating from Minto to Port Botany and return and crewed by Sydney Rail Services.

Both trains were performing a propelling movement at the time of collision and subsequent derailment. 

The following analysis will discuss the conditions prior, during and post‑collision and the risk management of the rolling stock operators and rail infrastructure manager in relation to operations at Port Botany.

Propelling train 1150 passes signal BY31 at stop

Both T296 and 1150 required access to the stevedore ports by means of a propelling manoeuvre. This was reportedly due to the stevedore ports banning locomotives from remaining coupled to the train after an incident in 2003 where the train moved whilst a reach stacker was loading it. Propelling into the stevedore sidings allowed the locomotive(s) at the exit end to uncouple and move clear of the hardstand area during loading.

CCTV gathered by OTSI shows both trains moving towards each other, however due to the footage angle it does not show the relevant signal aspects on BY29 and BY31. 

When the CCTV is analysed in parallel with the Phoenix train control replay, an accurate record of the events can be obtained.

With both trains ready to propel towards the stevedore sidings, the Phoenix train control system replay displayed a proceed indication on signal BY29 for T296, located on arrival road 1. 

This signal indicated to the second person leading T296, that the route was correctly set, and that they had been authorised to start their propelling movement towards the DP World stevedore port sidings. The Phoenix train control system replay also showed 1150 occupying arrival road 2 and a stop aspect on their signal BY31.

As T296 began its propelling movement, 1150 can also be seen to begin an unauthorised propelling movement passing BY31 in the stop position. As the 2 trains approached converging tracks, 1150 collided with the side of T296. This collision resulted in several wagons from both trains derailing.

The Phoenix replay shows the route set by the NCO for T296 in green, with signal BY29 set to proceed and BY31 remaining at stop (Figure 25). Additional Phoenix replays show track circuit alarms seen by the NCO, caused by derailed wagons and a light pole when it fell across an adjacent unoccupied track (Figure 26 and Figure 27).

Had 1150 not passed signal BY31 in the stop position, it is certain that the collision with T296 and subsequent derailment would not have occurred.

Figure 25: ARTC Phoenix replay screenshot displaying route on BY29 for T296

Figure 26: ARTC Phoenix replay screenshot during collision and derailment

Figure 27: ARTC Phoenix replay screenshot post‑collision and derailment

Figure 28: Collision at 163 points

Qualified driver directing train 1150 mistook the location of their train 

The competent worker responsible for directing the propelling movement for 1150 was the qualified driver for the train service. This resulted from a mutual Train crew 1150 role swap agreed to early in the shift, with the second person controlling the locomotive.

Trains T296 and 1150 were both ready for propelling movements towards the stevedore port sidings and located adjacent to one another on arrival roads 1 and 2 respectively (Figure 28). The crew leading the propelling movements were positioned inside separate motor vehicles, allowing sighting of the rear wagons on both trains and signals BY29 and BY31.

The previous day the qualified driver for 1150 was rostered for the same shift on the same train, but on arrival at Port Botany yard had been positioned on arrival road 1. On the day of the incident 1150 was positioned on arrival road 2.

At interview the qualified driver commented that the end of train markers on the rear wagons of both trains looked identical. However, when viewing CCTV of the incident, it is apparent that the rear few wagons on 1150 had a payload of containers, whereas the rear few wagons on T296 were empty. 

The qualified driver leading 1150 did not mention the payloads aboard the wagons, probably as these were expected to vary from shift to shift. Instead, it is likely that the qualified driver’s focus was inadvertently on signal BY29 on arrival road 1, based on the events of the previous shifts.

Expectation Bias occurs when an individual’s expectation influences their perception and interpretation of information. This in turn can impact decision‑making. James Reason describes further that the people prone to these Error types are those that possess the right knowledge, skills and experience. However, as tasks become more familiar and routine, they are performed with less conscious effort, making the worker more susceptible to inattention to the task at hand.

For both the qualified driver and second person of 1150, St Mary’s to Port Botany and return was the only route competency they held, and this train had become extremely familiar and routine. The qualified driver for 1150 in interview commented that arrival road 1, which on the day of the incident was occupied by T296, felt the most familiar and was a common location for train 1150 on previous occasions.

The qualified driver also describes in interview that they realised their error moments after the collision and derailment. Originally, they perceived the accident as their train derailing whilst it negotiated the points. However, on closer inspection they realised that the incident was a collision and subsequent derailment. The qualified driver leading 1150 had mistaken the location of their train.

RISSB SPAD mitigation guidelines state that signal sighting decisions taken by RIMs can contribute to a more hazardous work environment. Reacting to the wrong signal or reading across through to a non-target signal occurs when a target signal has been stopped at correctly, however due to other distractions such as shunting duties, train crew respond to a change in aspect on a non-target signal that is also in view.

OTSI observes also that shunting signals at Port Botany were being observed from within a motor vehicle. 

Train crew reported during interviews feeling safer staying within motor vehicles whilst conducting the line of sight procedure, likely contributing to train crew not exiting their vehicles to gain a better vantage point to view signals BY29 and BY31. Exiting the vehicle may have provided the qualified driver with another opportunity to locate their train correctly.

Also signal sighting standards are commonly based on the viewing of signals from within locomotives and not within motor vehicles. It was noted that the signal aspects were not fully visible from a motor vehicle, also potentially adding to the risk of read errors (see Figure 24).

Line of sight procedure non-compliance

The line of sight procedure is a unique stand-alone procedure only in use at Port Botany. The procedure allows motor vehicles to be used to assist train crew in performing propelling movements. Thus, replacing the standard requirement of crews to walk beside lead vehicles whilst propelling as described in the ANTR 424 Propelling Trains procedure.

Through the course of the incident investigation OTSI observed widespread non‑adherence to line of sight shunting procedures in Port Botany yard post‑incident. Specifically, train crew routinely leading line of sight propelling were not:

• stationary at a selected line of sight location
• sighting the lead vehicle moving towards the selected location
• stopping the propelling movement once the lead wagon reaches the selected location
• repeating the process until the destination had been reached as required by ARTC’s line of sight shunting procedure.

Instead, OTSI observed instructions were being given from moving motor vehicles, ahead of the lead wagon and whilst the trains were kept moving, in a similar manner to the day of the incident. 

Reasons for observed non-compliance were explored through interviews with crew from 3 different operators, analysis of human factors and available evidence. 

Crew from 3 different operators in Botany yard were observed during the site visit and on CCTV as specifically not selecting a line of sight location during propelling movements. OTSI interviewed a sample of crew from these observations. 

When questioned about their reasons for using an alternate shunting procedure crew reported that the line of sight procedure was complex, inefficient and impractical to carry out and they considered the alternative method to be significantly more efficient, although being aware of the line of sight shunting procedural requirements.

The reasons stated by train crew for not selecting a line-of sight location as per procedure were:

• brake recharge time following a train brake application (40‑second chokes on wagons)
• easier to keep trains moving (driving motor vehicle alongside the rear of train)
• avoiding stop/start nature of procedure
• impractical in that procedure takes too long to complete 
• hazardous to exit motor vehicle during propelling movements.

Yard topography

The second person contingent identified line of sight visibility issues arising from the surrounding buildings and topography of the rail corridor, which has created tight track curvature. The track layout around buildings and roadways restricted vision and as a result it was difficult to carry out the line of sight shunting procedures. 

To maintain line of sight around this topography crew could only move a short distance before calling the train onto their line of sight location, stopping the train, and then moving to the next location to repeat the process. This required that trains be stopped and started several times over a few hundred metres to fulfill line of sight location requirements.

Additionally, owing to narrow access roads, crew identified risks in stopping motor vehicles in some areas and deviated from procedure for personal safety reasons. This meant that the line of sight location was not selected, and propelling movement instructions were given from within moving motor vehicles while the trains also kept moving. Some second persons also reported in interview that they witnessed other operators, in their view, placing themselves in danger outside of the motor vehicles, to gain better vantage points in attempting to apply line of sight procedures. 

Train brake operation

The main reasons cited by drivers for not following the correct procedure is that line of sight shunting takes too long. Resoundingly, the issues related to the stop/start nature of the procedure while the alternate method used kept trains moving. 

The driver cohort in particular cited delays caused by the train braking systems. Once the train is stopped, the Automatic Brakerequires time to recharge the brake pipe, which in turn, allows train movements to recommence. This is due to the fitment of 40‑second chokes on wagons. When release and running is selected on the driver’s brake lever, the brake pipe begins to recharge. However, the train brakes are prevented from immediately releasing via the chokes fitted to wagons.

The stop/start requirements combined with delays to recharge the brake system meant that significant time was needed to comply with the procedures.

Compliance and training

Train crews were asked if they follow the line of sight procedure, to which all answered no. 

The interviewed train crew also stated that they were shown the correct procedure during their training, however then witnessed the alternate method being conducted when on site. 

Train crew described at the time of their Verification of Competency (VoC) assessment, that they carried out the procedure correctly to be deemed competent. 

However, of the train crew interviewed, all commented that in normal operations (not during a VoC), they had never witnessed the procedure being followed correctly. Instead, the alternate method was almost exclusively being used by all crew at Port Botany. 

Additionally, they advised compliance activities were not being undertaken by the rail infrastructure manager and only minimally from the various operators’ perspectives. Of the train crew interviewed, all reported they had never seen the rail infrastructure manager undertake compliance activities. 

Operator compliance was undertaken in the form of ‘check rides’, whereby driver trainers periodically assessed train crew on numerous activities which included line of sight procedural compliance. However, crew advised that on nearly every occasion, they were forewarned of the assessment taking place and therefore complied with the requirements of the line of sight procedure. As soon as practical assessments were completed, train crew reported returning to the alternate method of propelling trains, contrary to the procedure.

Time pressures

Most train crew interviewed reported a perceived time pressure, of not wanting to hold up other operators arriving and departing stevedore sidings. 

ARTC provided Train Information pathing times for the incident trains for both arriving and departing the stevedore sidings and Port Botany yard. The pathing allowed on average over 3 hours at the stevedore sidings to service up to around 30 wagons for each train. Typically, this would provide sufficient time for stevedores to service the train. 

When questioned, only one of the drivers interviewed reported experiencing delays or reportedly caused the delay of other trains at Port Botany. However, an operator representative commented that it was common to be delayed accessing stevedore ports because of other operator movements.

Summary

Crew interviewed from 3 separate operators in Port Botany yard cited line of sight procedures as being inadequate for the task. This inadequacy combined with minimal compliance monitoring was described as the most common reason for the observed routine violations. 

Situational violations, normally described as a ‘one off’ violation, are Intended actions/ violations. Breaking the rules to respond to an immediate situation, such as a time pressure, they are a choice rather than an accident and arise when people feel the violation is necessary to achieve a desired outcome, such as completing tasks quickly and as easily as possible. They do not arise due to a lack of knowledge, or malicious intent. 

Combined with poor or absent compliance activities, these situational violations may lead to routine violations if the conditions remain. 

Routine violations, described as ‘no one follows that rule’, become habitual practice, particularly where rules are considered unnecessary, time consuming or onerous. 

Crew operating in Port Botany yard cited numerous concerns with the practical application of, and time required to complete, line of sight procedures. They described low visibility around curves limiting the distance to the next line of sight location and creating the need for multiple stop/starts over a short distance. Brake pipe recharging times created further delays. Tight track clearances were also specifically called out as a safety concern.

Overwhelmingly crew considered the line of sight procedure, when carried out correctly, to be time consuming and onerous, and combined with perceived time pressure, crew had created an alternate means to complete propelling movements as quickly and efficiently as possible. 

OTSI found training in the procedure had been adequate. But once crew commenced working at Port Botany, they routinely observed line of sight procedures not being followed, and a faster and perceived ‘safer’ alternative method being in its place.

With minimal compliance monitoring train crew had formulated the alternate method of propelling over a significant period of time. Competent workers were routinely not following the rules, and the alterative propelling procedures had become a custom practice in the workplace.

Had both crew leading their respective propelling movements chosen a line of sight location, it is highly likely they would have seen the 2 trains converging towards one another and stopped their respective trains prior to any collision.

Ambiguity in the ARTC line of sight procedure

Through train crew interviews and reviewing recent ARTC investigation reports for the March 2020 and January 2023 incidents, the ATSB/OTSI observed some perceived ambiguity specific to steps 1 and 2 in the line of sight procedure, and inconsistencies with the requirements in the line of sight protocols and the general network rules and procedures. The line of sight procedure states that the competent worker controlling the shunting movement must:

    1) gain authorisation from the ARTC Network Controller in NCCS before commencing any line of sight shunting movements

    2) ensure all points are set for the intended route.

There is a likely link between these inconsistencies and some of the perceived non‑compliances observed.

Gaining authorisation to commence line of sight shunting

The line of sight procedure does not specify how train crew should gain authorisation to commence line of sight shunting. Gaining authority in the procedure could be interpreted as verbal authority, however the procedure does not explicitly state verbal. 

Movement authorities are described in the general network rules and procedures, specifically ANGE 606 Responding to signals and signs and ANTR 424 Propelling Trains as:

• a shunting signal authorises a movement at restrictive speed past that signal (ANSG 606)
• if possible, signallers must use shunting signals to authorise shunting movements (ANSG 606)
• unless the signaller instructs that a movement is to proceed for a shorter distance, a proceed indication by a shunting signal is an authority to proceed (ANSG 606)
• signals, if available, must be used to authorise shunting movements (ANTR 424)

Where there are no signals available for a shunting movement, or additional instructions are required, then verbal authority from the NCO must be sought. 

Verbal authorities at Port Botany were necessary when stop signs, rather than signals, were used throughout the yard. Historically, verbal authorisation was required from the NCO to pass these stop signs. When the yard was upgraded to shunting signals the requirement for verbal authorisation was effectively negated as train crew could then gain authority to proceed on the indications the shunting signals provided in accordance with the standard network rules and procedures.

During interviews, it was evident that the majority of Botany qualified train crew considered the clearing of a shunting signal as the authority to commence line of sight shunting. However, one operator with considerable experience at Port Botany, that pre‑dated ARTC taking over as the RIM in April 2011, considered authority to be verbal prior to any signal being cleared. 

The ARTC investigation reports for both the March 2020 and January 2023 incidents found that no line of sight verbal communication occurred between the incident train crews and the NCO in both incidents. Whilst this statement is correct, the line of sight procedure has no clear requirement for the competent worker leading the propelling movement to obtain verbal authority from the NCO.

Ensuring points are set for intended routes

The line of sight procedure also required that train crew ensure points are set for an intended route. This suggests that crew visually check the positioning of points prior to propelling movements commencing.

Again, prior to April 2011, points were protected by stop signs, not signals. The NCO at Port Botany directed train crew via UHF radio to set and check certain routes manually via hand thrown point levers throughout the yard. After ARTC took over as the RIM these stop signs and manually hand thrown points were replaced by shunting signals and corresponding interlocked motorised points.

The line of sight procedure stated that the competent worker must ensure all points are set for the intended route. However, the general network rules and procedures, specifically ANGE 606 responding to signals and signs, under the headings of ‘signal indications’ and ‘proceed’, state:

• a proceed signal indication shows that interlocked points protected by the signal are set in the correct position for the movement, and 

• no conflicting route has been set, and  

• if available the signal may show which route is set (via a route indicator).

Train crew advised at interview, that they utilise shunting signals and route indicators to ensure points are set correctly, negating the requirement to visually check the point switchblades. They further identified that exiting motor vehicles to conduct a visual point check during line of sight propelling was considered a risk owing to narrow clearances on motor vehicle access roads. 

ARTC’s investigation report for the January 2023 incident states that both crew members leading the propelling movements for both trains T296 and 1150 did not visually check points, as required by ARTC’s interpretation of the line of sight procedure. However, the standard network procedures state that the crew members leading the propelling movement for T296 were able to accept the route was set correctly by the indication on signal BY29. 

Train crew and ARTC appear to interpret the checking of points directive in the procedure differently in that ARTC believe it to be a physical visual inspection, whilst crew utilise the signal and route indications.

Summary

It is important to have clear and concise written procedures so that they are interpreted correctly and carried out uniformly by all stakeholders.

The line of sight shunting protocols when read in conjunction with existing network rules and procedures may explain why there is a perceived ambiguity to certain aspects of the procedure. 

Movement authorities and route requirements in the line of sight procedure may be interpreted as being adhered to, given the supporting evidence in the general network rules and procedures. 

For example, the requirement for authority to be given by the NCO for line of sight shunting to the competent worker may be understood as the clearing of a signal, rather than a direct verbal communication. 

Similarly, if the competent worker receives the intended proceed route on a shunting signal, this verifies that the points are set for the intended movement.

Competent workers may have fulfilled the requirement of ensuring points are set correctly via the route displayed on the shunting signal per standard rules and procedures rather than visually checking them per ARTC’s interpretation of line of sight procedures.

Crews may not be following some protocols set out in the line of sight procedure unintentionally. Crews are likely utilising their knowledge of the wider network rules and procedures and interpret this as compliance with protocols required in the standalone line of sight shunting procedure.

Train crew also report that exiting a motor vehicle to check points is not considered by them to be safe if they were to apply ARTC’s interpretation of the procedure. 

Therefore, alignment of the line of sight procedures with standard network rules may reduce risk, and any ambiguity and confusion when propelling at Port Botany in comparison to other yards where similar shunting movements take place.

ARTC risk assessment process

Risk controls which may have reduced risk were not further evaluated

A formal risk assessment of propelling movements in Port Botany yard was conducted by ARTC on 30 April 2020. This risk assessment was instigated in response to the earlier almost identical collision and derailment that occurred in March 2020.

During this risk assessment it was documented that train operators referred to ‘the difficulty in compliance to this procedure’ due to the ‘access and vision restrictions’ arising from the yard layout and infrastructure.

Section 2 of the ARTC risk assessment report captured the identified risks from the workshop, including train to train collision, with the overall current risk rating assessed as High 3B.

Section 3 documented the proposed treatments for these risks. One of these proposed treatments was ‘compliance with line of sight shunting’. Although the table documented ownership of managing compliance with line of sight shunting in Port Botany yard, it is not clear from the assessment what was required for the control, as the accept/reject or investigate column was not complete. It was noted in the document section preamble that some of the treatments may have ‘further implications for customers and operators’ and ‘requires further investigation’. However, ARTC was unable to provide evidence that further evaluation of line of sight protocols took place, or that risks associated with ‘difficulty with compliance’ were documented or addressed.

As the documented table for the Proposed Treatments from ARTC risk assessment 2020 (see Figure 21) was incomplete, it did not record any decision i.e. adopt, reject or investigate compliance with line of sight shunting procedure.

ARTC has stated that it considers the line of sight shunting procedure to be an adequate control in the prevention of collisions/accidents at Port Botany, however this was not documented. There was no evidence the risks identified by operators were reviewed prior to the second derailment and collision that occurred on 13 January 2023.

Risk reviews were not conducted in accordance with ARTC procedures

The April 2020 risk assessment was ARTC’s only risk assessment of operations at Port Botany prior to 2023. This was initiated because of the collision and derailment that occurred in March 2020. 

ARTC’s risk management process document describes how overall risk ratings trigger risk review periods. For Port Botany, the overall risk rating was determined to be a High 3B at the March 2020 workshop, which called for a review period no longer than every 6 months, unless a review is triggered earlier (i.e as a result of incident investigation, audit findings, etc). 

This risk review period should have led to a risk review being done around October 2020. Instead, it wasn’t until the January 2023 incident, at the same location as the March 2020 collision and derailment, that another risk assessment was convened. The time elapsed between risk assessments was over 2 years.

Active train crew were not well represented in risk assessment processes

Operators labelled as present at the ARTC Port Botany risk assessment held on 30 April 2020 primarily held management positions. Only 2 of the 7 operators’ representatives in attendance were identified to be from an end user cohort. One was a qualified second person and the other was identified by their employer as a Subject Matter Expert (SME). The SME had been previously qualified as train crew at Port Botany. 

International standard ISO 31000 Risk Management – Guidelines stated best practice was consulting the end user cohort who carry out the procedures, as they are the most effective at identifying risk relating to their use. 

Having additional Port Botany qualified train crew present at the 30 April 2020 risk assessment may have reinforced some of the challenges identified with compliance.

Issues raised by operators with line of sight shunting procedure were not risk assessed 

In the April 2020 risk assessment workshop operators described difficulty in compliance with the procedure due to safe access to line of sight locations and vision restrictions, arising from the access roadways deviating away from the rail track, and some tight clearances at Port Botany. Whilst this was mentioned in the 2020 risk assessment background, there is no record of the specific concerns raised, or evidence that the concerns/risks were assessed or addressed.

The proposed treatments section of the 2020 risk assessment highlights, ‘compliance with line of sight shunting’, under the heading of ‘improved management of trains in Botany yard’. However, the document did not state how ARTC planned to address the risks and difficulties with carrying out the procedure identified by the operators.

Minimal compliance monitoring of risk controls

On the proposed treatment table in the 2020 risk assessment, compliance activities were not labelled as accept, reject nor investigate, although ownership sat with a designated ARTC general manager.

ARTC was requested to provide evidence of compliance monitoring of line of sight procedures in the yard, prior to the incident date of 13 January 2023. No evidence was provided.

Compliance activities by ARTC did not begin until after the January 2023 incident date. ARTC stated that in relation to compliance monitoring, the interface manager at Port Botany observed shunting movements and conversed with operators in March 2024. There was no formal documentation to support this occurring, and it is observed that this was more than 12 months after the January 2023 incident.

It is important to note also that OTSI observed ongoing routine non-compliance with the procedures during 2 random site visits post the 2023 incident. 

Exemplar risk assessment

OTSI requested an exemplar risk assessment for reference to Kooragang Island, Newcastle NSW as a propelling movement procedure (not line of sight) also exists for this location. ARTC advised that a risk assessment for Kooragang Island could not be supplied, as this activity had not been conducted.

Second person operated train 1150

On the day of the incident, the train crew of 1150 swapped roles at the beginning of their shift at St Mary’s. The second person assumed the role of the driver controlling 1150 and the qualified driver assumed the position of second person.

The second person drove the train supervised by the qualified driver from St Mary’s to Port Botany, including during the required run around movement shortly after arrival. 

During the propelling movement the second person was left unsupervised to drive the train whilst the qualified driver performed the role of second person.

During interviews, the crew of 1150 described this swap as common practice within the industry. The qualified driver perceived the second person to be knowledgeable, although this second person had not completed training that would have likely allowed them to fully understand the locomotive and airbrake systems. 

Train handling is a key skill that combines practical train driving experience with the knowledge of locomotive systems. Shunting in a yard environment such as Port Botany requires competency in train driving for the purposes of coupling and uncoupling wagons in such a way as to not injure crew and avoid damage to rolling stock.

The operator’s driver position description outlined key responsibilities, which did not include any mentoring role. The key responsibilities of the qualified drivers role stated the requirements to follow both occupational health and safety procedures. These procedures required rail traffic crews (specific to drivers) to be qualified to operate the rail traffic they drive in the ARTC network and delegate duties only to competent workers.

The operator’s second person position description does not state any requirement to operate locomotives or rolling stock. Additionally, the role description also prescribes compliance with network policies and procedures, specifically ARTC network procedure ANGE 232 Responsibilities of rail traffic crews. These responsibilities specifically state that employees of the operator must have competency to perform specific work functions and must only undertake duties and tasks for which they have been deemed competent for, based on knowledge, training and experience.

Whilst OTSI found that the role swap occurred, it did not directly contribute to the occurrence event and is described as a factor that increased risk. Operators and train crew, however, are reminded to only conduct work that they are qualified for. 

Findings

From the evidence available, the following findings are made with respect to the collision and derailment of trains 1150 and T296, in Port Botany New South Wales, on 13 January 2023. 

Contributing factors

• Propelling train 1150 passed signal BY31 at stop, converging into and colliding with train T296, which was propelling on the adjacent track. 
• The qualified driver directing the propelling movement of train 1150 mistook the location of their train. As a result, when signal BY29 cleared to allow T296 to proceed the qualified driver incorrectly authorised train 1150 to move. 
• Operators at Port Botany were routinely deviating from the ARTC line of sight shunting procedure for a perceived simpler method of working. The procedure was described by train crew as difficult, time consuming and at times unsafe to conduct. 

Other factors that increased risk

• Aspects of ARTC's line of sight shunting instructions were ambiguous, resulting in a varying interpretation of the procedure by stakeholders. (Safety Issue)
• ARTC did not proactively follow risk assessment principles specific to propelling movements to ensure safe rail operations within shunting yards under their control. ARTC conducted operations without the hazards/risks being fully or effectively assessed. (Safety Issue)
• The second person who was not authorised to drive a train under instruction of the qualified driver, operated 1150 en route to Port Botany. On arrival at Port Botany, the qualified driver of 1150 performed the on-ground duties during the shunt, leaving the unauthorised second person to operate the train unattended.

Safety issues and actions

Ambiguity in the ARTC line of sight procedure

Safety issue number: RO-2023-002-SI-02

Safety issue description: Aspects of ARTC's line of sight shunting instructions were ambiguous, resulting in a varying interpretation of the procedure by stakeholders.

ARTC risk assessment process

Safety issue number: RO-2023-002-SI-03

Safety issue description: ARTC did not proactively follow risk assessment principles specific to propelling movements to ensure safe rail operations within shunting yards under their control.

ARTC conducted operations without the hazards/risks being fully or effectively assessed.

Additional safety action by Pacific National

Pacific National advised that after reviewing the draft report that it has:

• Reviewed the ARTC OGW-30-25 Network Information Book – Line of Sight Procedure (Section 2.2.1) and has confirmed that the PN shunting procedure meets or exceeds the requirements outlined in the ARTC OGW-30-25 document. 
• Confirmed that frequent (quarterly) combined Port Botany shunting inspections are occurring with ARTC and PN on site verifying protection of propelling movements (line of sight shunting) and this activity is being completed by PN qualified workers not in a moving or stationary vehicle.
• Confirmed that in addition to the scheduled ARTC on site activity, critical control verifications are carried out by Enfield/Clyde leaders on PN crews performing shunting at Port Botany frequently and have a scheduled process in place to prompt this Port Botany focus.

ATSB comment

ATSB welcomes the additional safety action taken by Pacific National, in response to findings within the report.

Glossary

Sources and submissions

Sources of information

The sources of information during the investigation included:

• ARTC rules and procedures
• Australian Rail Track Corporation
• crew of Train 1150
• digital voice recordings
• documentation that is publicly available
• Informa corporate learning
• Office of Transport Safety Investigations
• Pacific National
• photographs taken on the day of the incident
• QR Train Handling guide
• RailTrain
• recorded data from locomotive on train 1150
• RISSB (now known as ARISO)
• Southern Shorthaul Railroad
• Sydney Rail Services
• Transport for New South Wales.

References

• James Reason 1991
• Rail Safety National Law (RSNL).

Submissions

Under section 26 of the Transport Safety Investigation Act 2003, the ATSB may provide a draft report, on a confidential basis, to any person whom the ATSB considers appropriate. That section allows a person receiving a draft report to make submissions to the ATSB about the draft report. 

A draft of this report was provided to the following directly involved parties:

• Crew of train 1150
• RailTrain
• ARTC
• Sydney Rail Services (SRS)
• Southern Shorthaul Railroad (SSR)
• Pacific National
• Qube Logistics
• Office of the National Rail Safety Regulator (ONRSR).

Any submissions from those parties will be reviewed and, where considered appropriate, the text of the draft report will be amended accordingly.

Submissions were received from:

• Sydney Rail Services (SRS)
• Office of the National Rail Safety Regulator (ONRSR)
• ARTC
• Pacific National.

The submissions were reviewed and, where considered appropriate, the text of the report was amended accordingly.
 

Rail safety investigations in New South Wales Most transport safety investigations into rail accidents and incidents in New South Wales (NSW) and Victoria are conducted in accordance with the Collaboration Agreement for Rail Safety Investigations and Other Matters between the Commonwealth Government of Australia, the State Government of NSW and the State Government of Victoria. Under the Collaboration Agreement, rail safety investigations are conducted and resourced in NSW by the Office of Transport Safety Investigations (OTSI) and in Victoria by the Chief Investigator, Transport Safety (OCI), on behalf of the ATSB, under the provisions of the Transport Safety Investigation Act 2003. The Office of Transport Safety Investigations (OTSI) is an independent statutory body which contributes to improvements in the safety of bus, ferry and rail passenger and rail freight services in NSW by investigating safety incidents and accidents, identifying system-wide safety issues and sharing lessons with transport operators, regulators and other key stakeholders. Visit www.otsi.nsw.gov.au for more information. Purpose of safety investigations The objective of an ATSB safety investigation is to improve transport safety through: identifying safety issues for action by organisations with the responsibility for managing that safety risk influencing safety action through engaging with stakeholders, communicating findings, and fostering awareness of safety issues and concerns.  In accordance with the TSI Act, the ATSB does not investigate for the purpose of taking administrative, regulatory or criminal action, and cannot apportion blame, assist in determining liability, or, as a general rule, assist in court proceedings.  About ATSB reports ATSB safety investigation reports are developed in accordance with ATSB procedures and guidelines, and with regard to applicable international standards and instruments. Reports must include factual material of sufficient weight to support the investigation’s analysis and findings. At all times the ATSB endeavours to balance the use of material that could imply adverse comment with the need to properly explain what happened, and why, in a fair and unbiased manner. An explanation of terminology used in ATSB investigation reports is available here. Publishing information Released in accordance with section 25 of the Transport Safety Investigation Act 2003 Published by: Australian Transport Safety Bureau © Commonwealth of Australia 2026 Ownership of intellectual property rights in this publication Unless otherwise noted, copyright (and any other intellectual property rights, if any) in this report publication is owned by the Commonwealth of Australia. Creative Commons licence With the exception of the Commonwealth Coat of Arms, ATSB logo, and photos and graphics in which a third party holds copyright, this report is licensed under a Creative Commons Attribution 4.0 International licence. The CC BY 4.0 licence enables you to distribute, remix, adapt, and build upon our material in any medium or format, so long as attribution is given to the Australian Transport Safety Bureau.  Copyright in material obtained from other agencies, private individuals or organisations, belongs to those agencies, individuals or organisations. Where you wish to use their material, you will need to contact them directly.

1. ^    The Office of the National Rail Safety Regulator (ONRSR) provides useful guidance on effective risk assessment to operators in its Safety Message dated 27 October 2022.
2. ^    A propelling movement requires one crew member to be situated on the ground, typically walking beside the front of their reversing train. Their role is to observe signals and monitor the path of the train as it propels, communicating instructions via handheld two-way radio to the driver in a locomotive attached at the rear of the train. Controlling the movement was the role of the second person.
3. ^    At Port Botany, rather than walking, motor vehicles were used to assist the second person on the ground leading the propelling movement.
4. ^    Signals BY29 and BY31 were shunting signals that protected 163 points which converged arrival roads 1 and 2 into one track at the stevedore end (southern end) of Port Botany yard.
5. ^    Loss of air was a result of the train separating during the derailment.
6. ^    Standard calls are direct point to point calls only between the driver and the NCO.
7. ^    Emergency calls are broadcast to all rail traffic in the area via the ICE radio emergency call function.
8. ^    RailCorp was responsible for the infrastructure maintenance, train control and safeworking rules that applied to the metropolitan freight network that Port Botany yard was a part of.
9. ^    For a more detailed explanation of train airbrake systems, refer to OTSI/ATSB investigation: RO-2020-022, Runaway and derailment of loaded grain train 3966, Dombarton, NSW, on 15 December 2020.
10. ^   Push pull operations described as locomotive placed on either end of a train consist.
11. ^   Further information is contained in the 1 July 2004 OTSI Rail Safety Investigation Report, Shunting Fatality Port Botany Rail Yard (OTSI file reference 02325).
12. ^   Commonly referred to as ‘SFAIRP,’ the Rail Safety National Law (NSW), s. 47, defined this as ‘…that which is (or was at a particular time) reasonably able to be done in relation to ensuring safety.’
13. ^   Rail Safety National Law (NSW), s. 99.
14. ^   While this referred to an earlier version of the standard (ISO 31000:2009), at the time of the accident it remained current.
15. ^   Rail Safety National Law National Regulations 2012, Sch. 1, item 13.
16. ^   For more information relating to ONRSR investigations, visit Investigations | ONRSR
17. ^   For more information relating to ONRSR regulatory activities, visit Audits, inspections, compliance measures | ONRSR

Executive summary

What happened

On 6 February 2023 at about 1532 local time, a Coulson Aviation Boeing 737-3H4 large air tanker (LAT), callsign Bomber 139 and registered N619SW, departed from Busselton Airport, Western Australia (WA) on a firefighting task to Fitzgerald River National Park, WA. There were 2 pilots on board, the aircraft captain in the left seat as the pilot flying and a co‑pilot in the right seat as the pilot monitoring. At about 1614, during the go-around from a second partial retardant drop, the aircraft impacted a ridgeline at an elevation of about 222 ft. The pilots suffered minor injuries and the aircraft was destroyed.

What the ATSB found

The ATSB found that the accident drop was conducted at a low height and airspeed downhill, which required the use of idle thrust and a high descent rate. The delay in the engines reaching go-around thrust at the end of the drop resulted in the aircraft’s height and airspeed (energy state) decaying as it approached rising terrain, which was not expected or detected by the pilot flying. Consequently, the aircraft’s airspeed and thrust were insufficient to climb above a ridgeline in the exit path, which resulted in a controlled flight into terrain. The operator’s practice of recalculating, and lowering, their target drop speed after a partial load drop also contributed to the low energy state of the aircraft leading up to the collision with terrain.

The ATSB also found that the operator and tasking agency had not published a minimum drop height, which resulted in the co-pilot, who did not believe there was a minimum drop height, not making any announcements about the low energy state prior to the collision. The ATSB found the operator’s pilot monitoring duties were reactive to the development of a low energy state and did not include call-outs either before or at the minimum target parameters to reduce the risk of a low energy state developing.

The ATSB benchmarked the WA, New South Wales and National Aerial Firefighting Centre standards against the United States Forest Service and United States National Wildfire Coordinating Group standards and found inconsistencies between the Australian agencies’ standards but not among the United States agencies’ standards. This was likely a result of each Australian state participating in the LAT program independently producing their own standards.

What has been done as a result

As a result of this accident, the operator developed a master corrective action plan to address the following topics:

• minimum drop speed and height
• recalibration of airspeed during partial load drops
• pre-drop planning and familiarisation
• pilot situational awareness
• flight training exercises
• exit flight path
• communication and coordination with Birddog
• adherence to standard operating procedures
• crew resource management
• human factors
• continuous training and proficiency assessment.

The Birddog operator reported that as part of their continuous improvement they removed their Lead Plane procedures from an annex in their Company Operations Manual and introduced a Lead Plane Standard Operating Procedures manual with expanded procedures in October 2023.

In November 2023, the WA Department of Fire and Emergency Services, and Department of Biodiversity, Conservation and Attractions released the Western Australian Aerial Fire Suppression Procedures 2023-2024. The draft version of this document was reviewed by their contracted operators and included a chapter for Large Air Tanker (LAT) Operating Procedures. The document included standard drop heights for all the aircraft types employed, procedures for task rejection, standdown and reinstatement of aviation resources after a safety incident, and a section on tactical flight profiles with the addition of the Lead and Chase-Position profiles. 

In response to the draft report, the WA Department of Biodiversity, Conservation and Attractions reported they would review the 2024‍–‍25 published drop heights with the intent of prescribing a ‘minimum’ safe drop height, noting the actual drop height to safely achieve the objective is the pilot-in-command’s decision.

In response to the draft report, the Australasian Fire and Emergency Services Authorities Council committed to develop a national standard for LAT Operating Procedures. This was delegated to the Aviation Safety Group through the National Aerial Firefighting Centre Strategic Committee within the purview of the Consistent Doctrine working group.

Finally, the ATSB has issued a safety recommendation to Coulson Aviation to take safety action to address their crew resource management procedures for retardant drops to reduce the risk of the aircraft entering an unrecoverable state before the pilot monitoring alerts the pilot flying.

Safety message

The history of aerial firefighting includes collisions with terrain accidents during the descent, retardant (or water) drop and go-around that have primarily manifested as controlled flight into terrain and loss of control accidents. The history of accidents needs to be understood within the context of the effect of drop height and speed on the firefighting effort, and the environmental conditions at the fireground, which may lead to an elevated level of risk tolerance. 

The lessons from this accident indicate that standard operating procedures and crew resource management should be implemented with the intent to prevent an unsafe situation from developing. While they should also include recovery actions, there may be insufficient safety margin available for a successful recovery and therefore the safety standards should not be solely dependent on the performance of the pilot flying and recovery call-outs. 

At a large fireground, it is likely there will be personnel and assets from multiple organisations and jurisdictions interacting. In this scenario, non-standard procedures and practices may result in unforeseen risks emerging. It is therefore important for tasking agencies to take the lead, with the support of stakeholders, in developing the quality and safety standards they require for the firefighting effort to mitigate operational risks.

The occurrence

On 6 February 2023 at about 1532 local time, a Coulson Aviation Boeing 737-3H4 large air tanker (LAT), callsign Bomber 139 and registered N619SW, departed from Busselton Airport, Western Australia (WA) on a firefighting task to Fitzgerald River National Park, WA (Figure 1). There were 2 pilots on board, the aircraft captain in the left seat as the pilot flying and a co‑pilot in the right seat as the pilot monitoring.[1]

Figure 1: Bomber 139 flightpath from Busselton to Fitzgerald River National Park

Source: FlightAware, annotated by ATSB

Earlier in the day, at 1015, and then again at 1215 and 1310, the WA Department of Biodiversity, Conservation and Attractions (DBCA) submitted aerial fire suppression requests to the State Operations Air Desk (SOAD). The request was for fixed-wing assets to the location of a fire 24 km west-north‑west of Hopetoun, WA under the criteria of ‘known high fuel loads and likelihood of excessive ROS [rate of spread] and/or extreme fire danger’. In response, the SOAD:

• identified the aerial assets available
• established a fire common traffic advisory frequency (F-CTAF)
• spoke to the pilots
• sent tasking messages at 1012, 1127, 1407 and 1505.

Bomber 139 was included in the 1127, 1407 and 1505 taskings. The SOAD LAT SARWATCH log[2] recorded that Bomber 139 departed from Busselton on these taskings at 1208 and 1330 and returned to Busselton each time before departing on the accident flight at 1530.

At 1519 the Birddog[3] (BD) responsibility for the LATs at the fireground transferred from Birddog 125 (BD125) to Birddog 123 (BD123) due to BD125 refuel requirements, with Birddog 682 (BD682) stationed above as the primary air attack aircraft.[4] Before BD125’s departure, a familiarisation flight of the fire zone was conducted in company with BD123 to discuss the layout of the fire, tactics, retardant drops, and objectives for the drops. The pilot of BD123 reported that no hazards were discussed. Bomber 139’s captain contacted BD123’s crew when they were 15 minutes from the fireground, and the Birddog pilot acknowledged their call and advised an altimeter QNH[5] setting of 1003 hPa. The captain acknowledged the altimeter setting and informed BD123’s pilot they would contact BD682 when they were 5 minutes from the fireground.

The pilot of BD682 subsequently cleared Bomber 139 to enter the F-CTAF not above an altitude of 2,500 ft and the captain notified BD682 they would be working with BD123 on an altimeter setting of 1003. The crew of BD123 queried BD682 if the line was clear and were advised that no ground personnel had been sighted and that they were not expected for about another hour. At this stage, Bomber 139’s co-pilot advised the captain of their drop speeds, which included a target drop speed (V_DROP) of 133 kt (1.25 Vs[6]).[7] When the captain reported BD123 in sight, the Birddog pilot asked Bomber 139 if they wanted a Show-Me[8] run or to follow BD123 straight to the drop, which would be a Lead profile.[9] The captain responded that they would follow BD123 straight to the drop. BD123’s pilot then obtained confirmation from the captain that right-hand circuits would be acceptable to avoid reduced visibility in smoke. 

BD123’s pilot briefed Bomber 139 that the plan was to use a road (Figure 2) as the start point and tag and extend the existing line of retardant, while keeping the smoke off to the left side, and that the line for the drop was clear of ground personnel. Bomber 139’s captain subsequently reported to BD123 that they were in position and the captain called for the co-pilot to complete the pre-drop checklist, which included:

• confirming coverage level[10] 3 was set
• disconnecting the autothrottles
• arming the retardant aerial delivery system.

The captain notified BD123 that their target speed would be 135 kt for final and then set the flaps to flap-40 (full flap). BD123’s crew then briefed Bomber 139 on the drop, which included advice of:

• a straight exit
• no hazards
• a downhill approach
• target altitude initially of 500 ft descending to 400 ft. 

Bomber 139’s co-pilot subsequently reported to the captain that the checklist was complete. The captain then queried the co-pilot about the flightpath BD123 was leading them on to the target and commented that their approach appeared to be high.

BD123’s pilot reported turning onto final for the drop, which was on a heading of 155º, and that visibility would be good after they cleared the smoke. About 15 seconds later, the captain queried the co-pilot where the road was for the start of the drop and the co-pilot responded that it was just ahead of them. BD123’s pilot then briefed Bomber 139 to ’start at the road and keep all smoke to the left, 3, 2, 1, start, your target altitude is 500 descending to 400’. The captain then confirmed sighting the road.

Bomber 139 descended to a minimum height[11] of 78 ft and completed a 9-second partial drop[12] of three-quarters of their tank at about 70% N1[13] with a minimum computed airspeed[14] of 124 kt before the captain stopped the drop because their retardant line was starting to enter area that was already burnt. The captain reported the partial drop to BD123, and another Lead profile was set up to tag and extend the first drop with the remaining retardant, with the captain remarking to the crew of BD123 ‘…and head down off the hill’.

The captain instructed the co-pilot to conduct the pre-drop checklist for the second drop and the co‑pilot recalculated and reported that 118 kt would be their revised drop speed. On the second circuit the captain again queried the co-pilot about the flightpath they were taking to the final approach. The captain requested BD123 slow to 120 kt for the next drop, which was acknowledged. The Birddog pilot then briefed Bomber 139 to ‘tag and extend all existing retardant, it is start at the hill as it pushes down, target altitude 500 descending 400’. 

The co-pilot reported that flap-40 was set, and the pre-drop checklist was completed. The captain again reported uncertainty to the co-pilot about the approach path BD123 was taking. The Birddog pilot then provided the following commentary, ‘this is final, fully retardant drop out here in a second, standby, [pause], retardant’s right at our 12 o’clock, [pause], 3, 2, 1, tag and extend existing retardant’. The captain reported at interview that the second approach was conducted through drift smoke from the fire, and they were unable to identify the target until they exited from the smoke, which was at the start of the drop. The co-pilot reported not being able to see the target on final until BD123 deployed smoke over the location.

During the second drop, Bomber 139 descended to a minimum height of 57 ft at 110 kt and about 30% N1 (engines at high idle)[15] as the retardant line was extended downhill. The aircraft recovered to a height of 81 ft at 107 kt at the end of the drop at which stage the captain had started to advance the thrust levers as the rate of descent peaked at about 1,800 ft/min. The co‑pilot had their hand on the flap lever in anticipation of the flap retract call from the captain as soon as they started to climb for the go-around.

The engines did not immediately respond to the movement of the thrust levers and the captain started to pitch the nose up, which resulted in a reversal of the rate of descent and a decay of the airspeed. The captain then announced ’fly airplane’ followed immediately, at about 1614,[16] by the activation of the stick shaker[17] and an abrupt vertical acceleration associated with the aircraft impacting a ridgeline at an elevation of about 222 ft at 104 kt with the engines at 85‍–‍89% N1. 

The co-pilot did not announce any deviations during the drop and accident sequence and later reported their focus of attention was likely on the airspeed indicator and radio altimeter, monitoring for any adverse trends. After the impact with the ridgeline, the aircraft cleared a small line of foliage before impacting the ground a second time and then sliding to rest (Figure 2). In response to the collision, BD123’s pilot broadcast an all-stations MAYDAY[18] call.

Figure 2: Bomber 139 retardant drops and accident site

Source: Department of Fire and Emergency Services, annotated by ATSB

After motion had ceased, the co-pilot started the evacuation checklist. Both pilots were unable to open the cabin door as it had buckled and the co-pilot was unable to open the right-side window. The captain observed out of the left side window that a post‑impact fire had started and managed to open that window on their second attempt. Both pilots then exited out of the left window and moved clear of the wreckage and fire. They were subsequently rescued by a helicopter involved in the fire control activities after 2 single-engine air tankers had dropped water with foam on the aircraft fire, believing the crew were still inside. The crew suffered minor injuries and the aircraft was destroyed. Figure 3 depicts Bomber 139’s recorded accident flightpath.

Figure 3: Bomber 139 recorded accident flightpath at Fitzgerald River National Park

Source: Google Earth, annotated by ATSB

Context

Personnel information

Aircraft captain

The aircraft captain joined the operator, Coulson Aviation, in 2016 with a United States Department of Agriculture Forest Service (USFS) Air Tanker Training Pilot qualification. The captain became their director of fixed-wing operations in 2017, having flown several multi-engine airtanker types over about 20 years. The captain held an airline transport pilot certificate, reissued by the US Federal Aviation Administration (FAA) on 6 October 2020 with a rating for Boeing 737 (B737) and a current first class medical certificate with no limitations. The captain had accumulated 8,233 hours of flying experience, which included 1,399 hours on the B737 and about 5,500 hours aerial firefighting experience. In the previous 90 days the captain had accumulated 84 hours, all on the B737.

The captain’s last pilot-in-command proficiency check on the B737 (in accordance with the FAA regulations) was on 30 November 2022, with a satisfactory result. The captain conducted 2 company check flights on 11 May 2022, which were simulated initial attack mission training flights for a new co-pilot. These flights were part of the operator’s annual northern hemisphere spring season training program. Both reports indicated no assistance was required and no deficiencies were recorded. 

Co-pilot

The co-pilot joined Coulson Aviation on 18 April 2022 with prior aerial firefighting experience flying the air tactical group supervisor, but no previous low flying or large air tanker (LAT) experience. The co-pilot attended the operator’s 2022 spring season ground training and occupied the jump seat as an observer in the B737 for the flying training. Between 1‍–‍7 June 2022, the co-pilot completed a B737 type rating course with a satisfactory result recorded for the check on 7 June 2022. The co-pilot was issued with their USFS qualification on completion of their B737 type rating, which was an administrative process for co-pilots and did not require an evaluation flight.

The co-pilot held an airline transport pilot certificate, reissued by the FAA on 7 June 2022, with a rating for B737 and a current first class medical certificate with no limitations. On 12 September 2022, the co-pilot participated in an initial attack evaluation flight as the co-pilot on the B737. The co-pilot had accumulated 5,852 hours of flying experience, which included 128 hours on the B737, about 500 hours aerial firefighting experience and 36.9 hours on the B737 in the previous 90 days. 

Crew work-rest history

The captain reported they had 6 and 14 hours sleep in the previous 24 and 48 hours respectively. They self‑assessed that their level of mental fatigue at the time of occurrence was 2 - Very lively. Responsive, but not at peak.[19] The co-pilot reported they had 8 and 18 hours sleep in the previous 24 and 48 hours respectively and a self‑assessed mental fatigue level at the time of occurrence of 2 - Very lively. Responsive, but not at peak. 

The crew’s tour of duty started on 20 January 2023. They had flown 7.3 hours and conducted 13 drops (including partial drops) on the current tour prior to the day of the accident. The accident occurred on their third flight and fourth retardant drop of the last day of their tour after flying about 3.3 hours that day. The pilots’ self-assessments and work-rest history did not indicate that mental and physical fatigue were likely to have been contributing factors to the accident.

Aircraft information

General information

The aircraft was a Boeing 737-3H4, serial number 28035, registration N619SW, issued with a certificate of airworthiness in the transport category^[20] by the FAA on 9 November 1995 and fitted with 2 CFM56-3 turbofan engines. It entered service with a US airline on 12 November 1995 and accumulated 69,016 hours before transfer of ownership to Coulson Aviation on 8 August 2017. The latest certificate of registration was issued on 14 November 2017 and was valid until 30 November 2023.

Tanker modification

On 10 November 2018, Coulson Aviation were issued with a supplemental type certificate (ST04050NY) by the FAA for the installation of the Coulson Aerial Firefighter Tanker Modification (retardant aerial delivery system – RADS) to Boeing model 737-300 series aircraft, which limited passengers to persons related to firefighting mission‑essential activities. The RADS modification was completed on 30 May 2022. At the time of the accident, the airframe had accumulated 69,187.6 hours.

The RADS incorporated 2 tanks on either side of the centre fuel tank with a total capacity of 36,000 pounds (lb).[21] The cockpit had a primary and secondary user interface installed and push button drop switches on both the pilot and co-pilot controls. According to the flight manual supplement:

When the flight crew activates either switch, the primary user interface software will produce a door position command based on the predefined Coverage Level (CL) requirement, drop quantity, tank levels, ground speed and aircraft height above the ground.

The RADS design included an emergency dump switch that provided full system hydraulic pressure to dump the full load in less than 2 seconds. To avoid nuisance audio alerts during a retardant drop, the RADS modification incorporated an audio inhibit switch, which inhibited the landing gear configuration, ground proximity warning system and traffic collision avoidance system for 5 minutes.

Flight instruments

The captain and co-pilot instrument panels were fitted with radio altimeter indicators to display the aircraft height above terrain when below 2,500 ft above ground level. A decision height (DH) could be set on each of the radio altimeters, which activated a DH light on the respective attitude indicator when that height was reached, but did not include an audio alert.

The altimeters and airspeed indicators incorporated bugs that could be set to reference altitudes and airspeeds. The operator reported that the orange airspeed bug was used for the target drop speed. The altimeter bug was not used because the retardant drop was conducted at a height above terrain and not at an altitude. Figure 4 illustrates the co-pilot’s instrument panel with the radio altimeter DH set at 50 ft.

Figure 4: Co-pilot’s partial instrument panel

Source: Operator, annotated by ATSB.

Head-up guidance system

Under previous ownership, the aircraft was fitted with a Rockwell Collins flight dynamics model 2300 head-up guidance system (HGS) for the left seat pilot (captain’s seat), which was approved for use as a supplemental display. The HGS permitted the flying pilot to scan the airspeed and radio height, while maintaining an external visual scan of the environment during the retardant drop. Consequently, the captain reported that they considered the DH light on the instrument panel to be of no benefit while flying the retardant drop with the HGS.

The HGS had 4 modes of operation. All 4 modes could provide an airspeed error symbol, to indicate the difference between the aircraft’s actual airspeed and the reference airspeed set by the pilot on the mode control panel. A DH annunciation was also available in the mode used by the captain. The captain set the target drop speed on the mode control panel to provide a reference airspeed during the retardant drop but did not use the DH setting.

Boeing 737-300 engine acceleration requirements

According to the engine manufacturer, CFM (GE and Safran partnership):

An acceleration check is defined in Boeing 737-300/400/500 AMM Task 71-00-00-715-015-C00. This demonstration test requires the engine to accelerate from high idle to the fan speed corresponding to approximately 95% of static take-off thrust within 7.4 seconds.[22] The target fan speed is corrected for pressure altitude and outside air temperature. However, there are multiple factors that can impact the expected engine acceleration time during flight, including but not limited to:

1) The 7.4-second requirement is for a rapid throttle movement (1 second or less) from idle to take-off power. Slower throttle movements may result in a longer acceleration time.

2) The 7.4-second requirement is with engine bleeds and electrical loads OFF. Engine bleeds ON can extend the acceleration time.

3) Mach number can impact the acceleration time, but an airspeed of 110 kts at low altitude would likely have an insignificant impact.

Meteorological information 

On the day of the accident there was a low-pressure trough crossing southern WA, resulting in thunderstorms and wind direction changes from the north-east to the south-east to the south-west. The graphical area forecast for the accident site included moderate turbulence from the surface to 10,000 ft with thermals. A SIGMET^[23] was issued at 1400 for frequent thunderstorms associated with the trough (Appendix A – SIGMET thunderstorm activity). While the north-western boundary of the SIGMET was close to the accident site, satellite and radar imagery indicated there were no thunderstorms in the vicinity of the accident site at the time of occurrence. 

The Hopetoun North weather station, located about 24 km east‑north‑east of the site, did not record any rainfall and the 10-minute weather data from the station is provided in Table 1. With reference to Figure 2, the smoke from the fire and accident site indicated the retardant drops were conducted into wind, which the Bomber 139 captain reported was about 10‍–‍15 kt.

Table 1: Hopetoun North weather station recordings

Pre-flight risk assessment

The aircraft captain completed the operator’s flight risk assessment tool (FRAT) at 1015. The FRAT score was 14, which was an acceptable level of risk with no mitigation or escalation required. If the FRAT had been reviewed between each flight, the SIGMET issued at 1400 would have added 8 points, resulting in a FRAT score of 22 for the accident flight that departed from Busselton at about 1530.

A score greater than 21 required mitigation or escalation, noting the operator’s procedures prohibited an elevated FRAT from being approved by the aircraft captain. While the weather conditions, including the thunderstorm activity advised in the SIGMET, were not considered to be a factor in this accident, this has been included in the report to raise awareness of how a risk management process could miss a developing flight hazard.

Wreckage and impact information

The aircraft’s approach to the ridgeline left 2 distinct jet-blast lines, evidenced by trees broken in the opposite direction of travel (Figure 5). Closer to the ridgeline, trees were broken in the direction of travel, likely from contact with the aircraft’s engines and airframe with evidence of retardant transfer from the airframe to the foliage just below the ridgeline. The length of the jet‑blast lines were about 74 m and 62 m respectively for the left and right engines. After contacting the ridgeline, the aircraft became airborne for about 69 m, shedding engine, wing, and fuselage debris before impacting a second time in a slight nose down attitude on a heading of about 140º.

Figure 5: Jet-blast corridors and ridgeline impact

Bulldozer fire tracks constructed after the accident. Source: ATSB

The aircraft came to rest about 176 m from the ridgeline and yawed left to the direction of travel onto a heading of about 080º. The fuselage had a main fracture near the tail and the left engine had separated from the left pylon and was resting adjacent to the forward fuselage (Figure 6). The left engine pylon exhibited a 70º upward bend, which was likely from the impact with the ridgeline, noting the left engine debris field started from the ridgeline and there was no evidence of left engine drag marks. The aircraft was consumed by fire, but there was no evidence of:

• fire trail^[24] or fuel spill before the aircraft came to rest
• any debris separating from the aircraft before it impacted the ridgeline. 

Figure 6: Second impact and main wreckage

Bulldozer fire tracks constructed after the accident. Source: ATSB

The flight data recorder (FDR) and cockpit voice recorder (CVR) were recovered on the first day of the site and wreckage inspection and retained by the ATSB for examination and download. Due to the extensive fire damage, only a limited inspection of the aircraft was achievable. This included establishing the positions of the leading-edge flaps, trailing-edge flaps, and horizontal stabiliser, with no anomalies found. The trailing-edge flap ballscrews were in the fully extended position, consistent with a flap-40 setting. 

The left main landing gear was found adjacent to the aircraft and part of its support structure was in the debris field indicating it was torn from the aircraft prior to it coming to rest. The nose and right main landing gear were retracted in place. The extent of the fire damage precluded an inspection of the cockpit and flight instruments. Figure 7 depicts the horizontal distances and elevations (measured with a differential global positioning system), and the angles presented by the foliage damage associated with the accident sequence.

Figure 7: Accident sequence distances and elevations

Source: ATSB

Recorded information

The CVR (Honeywell SSCVR P/N 980-6022-001) and FDR (Honeywell SSFDR P/N 980‑4700‑001) were transported by a recorders specialist from the accident site to the ATSB’s Canberra technical facility. The condition of the CVR prior to disassembly is depicted in Figure 8. The condition of the FDR prior to disassembly is depicted in Figure 9.

Figure 8: CVR prior to disassembly

Source: ATSB

Figure 9: FDR prior to disassembly

Source: ATSB

As a result of fire damage to the recorders the data recovery process required disassembly, inspection, and repair of the memory boards inside the crash survivable memory unit. The FDR memory board exhibited discoloration, melting and flow of the conformal coating with multiple controller pins desoldered. The CVR exhibited discoloration of the heat indicator, several controller pins had separated from the memory board and multiple short-circuits were identified. 

Following repairs to both memory boards, successful downloads of data were achieved. The FDR provided 25.5 hours of flight data and the CVR 30 minutes of audio on 4 channels (pilot, co-pilot, public address, and cockpit area microphone). The FDR plot of the final 30 seconds before the impact with the ridgeline, with the approximate position of the accident retardant drop inserted, is shown at Figure 10. The time parameter in the figure (UTC)^[25] is an approximate calculation based on the aircraft’s auxiliary telemetry unit global positioning system recording of the retardant drops. Appendix B contains the FDR plot of the last 5 minutes of flight with the 2 drops conducted on the accident flight.

During the final low-level retardant drop the thrust levers were positioned at high idle (0°) while at a radio height above the ground of less than 100 ft. As the engine N1 speed decreased to about 30%, the rate of descent increased to 1,800 ft/min before the thrust levers were advanced while the aircraft continued to descend. About 2 seconds prior to impact, at a radio altitude of about 28 ft with the flaps at 40°, an increasing aircraft nose-up pitch attitude resulted in both angle of attack vanes exceeding 20° (maximum 23.6°), which triggered the stick shaker. The aircraft attained a positive rate of climb just before impact with terrain.

Boeing provided the ATSB with a copy of their preliminary flight data analysis report of the accident, which was used to validate ATSB flight data output. Noting the aircraft attained a positive rate of climb with high engine N1 speed just prior to impact, Boeing reported in their preliminary findings that ‘Had there been sufficient altitude remaining, the airplane would have likely climbed away without impacting the ground.’

Figure 10: FDR plot of final 30 seconds of accident drop

Final 30 seconds of flight data showing corrected altitude, radio height, pitch angle, angle of attack (same scale), thrust lever angle, engine N1 speed, computed airspeed, groundspeed, vertical speed, stick shaker and approximate position of the retardant drop. The vertical acceleration spike indicates the initial impact. Source: ATSB

Engine acceleration performance and drop times

According to the FDR data, the captain started to advance the thrust levers about 7 seconds before the collision and took about 5 seconds to advance them from 0° to 30° thrust lever angle. The engines started to accelerate from 30% N1 (high-idle) about 1 second after the captain started advancing the thrust levers and reached about 85‍–‍89% N1 in a 6-second acceleration period that coincided with the first impact. The acceleration performance depicted by the recorded data was consistent with the Boeing 737-300 engine acceleration check requirements, discussed previously.

According to the operator’s flight crew operations manual (FCOM) take-off procedure, after the pilot flying has aligned the aircraft with the runway, the thrust levers are initially advanced to 40% N1 and the engines allowed to stabilise, then TO/GA [take-off / go-around] is selected and thrust levers advanced to take-off N1. The accident FDR plots depicted a distinct change in the engine acceleration curve from about 40% N1 (at 0813:46 in Figure 10). The engines took about 4 seconds to accelerate to 41‍–‍43% N1 and another 3 seconds to accelerate to 85‍–‍89% N1.

Noting the FCOM take-off procedure and the distinct engine acceleration profile, the ATSB asked the operator if the take-off procedure was related to engine acceleration time and how this was managed during a retardant drop. The operator clarified that the FCOM take-off procedure to allow the engines to momentarily stabilise at 40% N1 before advancing to take-off thrust was in accordance with the Boeing 737 flight crew training manual, which provided the following explanation:

Allowing the engines to stabilize provides uniform engine acceleration to take-off thrust and minimizes directional control problems. This is particularly important if crosswinds exist or the runway surface is slippery. The exact initial setting is not as important as setting symmetrical thrust.

For the management of engine acceleration during retardant drops, the operator reported that if the terrain at a drop site was level, then the engines would likely be set at 60‍–‍65% N1 and they would quickly accelerate to the go-around thrust at the end of the drop. However, to control the airspeed during a downhill drop, they could not prescribe a minimum N1 setting, and the pilot flying may need to use flight idle thrust to control the drop speed.

In this situation pilots were trained to advance the thrust levers about halfway through the drop so that the engines could quickly accelerate to go-around thrust at the end of the drop. The operator acknowledged that if the pilot was late advancing the thrust levers from idle, it would result in a delay of about 3 seconds before the engines could be accelerated up to the go-around thrust setting.

On the accident drop, the FDR recorded that the captain advanced the thrust levers about halfway through the drop. However, the accident drop was a quarter load, which took 5 seconds to release. In comparison, the first partial drop of three-quarters of the load took 9 seconds to release. The 2 previous drops on the day of the accident were both full loads at coverage level 3 and took 12 and 15 seconds, respectively. Two days earlier, 7 full or partial drops were conducted at coverage levels 6 and 8. The full drops at coverage level 6 were released in 6‍–‍7 seconds and the full drops at coverage level 8 were released in 5 seconds. This indicated that, given the typical engine spool‑up time, either a partial drop or a drop at a high coverage level with the engines at idle could result in a post-drop delay in engine acceleration to go-around thrust.

Tactical flight profile

First and second flight taskings

The first retardant drop of the day for Bomber 139 at the fire ground was with BD125 and started with a left-hand circuit Show-Me profile about 500 ft above the subsequent drop height in a northerly direction on the western flank of the fire. The overfly of the target was followed by a close left-hand circuit with BD125 in the Chase-Position profile[26] and a full load release at coverage level 3 over 12 seconds. 

On the second drop, on the eastern flank of the fire, another Chase-Position profile left-hand circuit was flown with a tag and extend full load release in a southerly direction at coverage level 3 over 15 seconds but without a Show-Me profile. Figure 11 depicts the flight profiles and Figure 12 shows imagery of the first drop, which started at the same road as the accident flight, and the second drop tag and extend an existing line of retardant.

Figure 11: First and second drop flight profiles on 6 February 2023

Source: Google Earth, annotated by ATSB

Figure 12: First drop (left) starting at road and second drop (right) tag and extend

Source: Department of Fire and Emergency Services, annotated by ATSB

Accident flight tasking

When Bomber 139 returned to the fire ground for the third drop of the day, BD123 had replaced BD125. The BD123 pilot reported having prior experience working in terrain with more extensive vertical relief in Canada and considered the accident flight target and exit to be relatively flat. They further advised being aware of a 40‍–‍50 ft change in the elevation and, as a matter of practice, would not conduct uphill drops.

The Birddog pilot assessed the altitudes for the drop were about 450 ft at the start and 400 ft at the end, which they estimated would provide 100‍–‍150 ft height above terrain. However, the pilot provided the altitudes in hundreds of feet and therefore briefed 500 ft as the start altitude. At follow-up, the pilot confirmed they briefed a safe altitude(s) for Bomber 139 to transit through the run and that they levelled their aircraft at their briefed altitude over the target as a visual reference for Bomber 139’s captain. 

The Birddog pilot reported that it was not uncommon for the Show-Me profile not to be flown if the LAT had conducted previous drops in the area. A right-hand circuit was not their preferred option but they did not consider it would be a problem because they were providing a Lead profile to the target. However, they reported a ‘slightly longer final’ was conducted to ensure that they could line up on their visual cues and see the target. 

After the first partial drop, BD123 conducted a sharp right turn to assess the drop while Bomber 139 conducted the go-around. When Bomber 139’s captain reported the need for a second drop the Birddog pilot offered the captain the choice of either another Lead or Chase-Position profile. The captain had no preference and the Birddog pilot elected to conduct another Lead profile, which required Bomber 139 extend their circuit until BD123 could regain the Lead position. The Birddog pilot reported that this resulted in another ‘slightly longer final’.

The Coulson Aviation fixed-wing flight operations manual did not include detail of the setup and conduct of retardant drops and the Bomber 139 captain referred the ATSB to the US National Wildfire Coordinating Group (NWCG) aerial supervision standards for how they expected a Lead Plane (Birddog) would conduct operations in the US. The Birddog operator’s standard operating procedures (SOPs) for Lead Plane Operations (2009 and 2023) provided the same tactical flight profiles and descriptions as those published by the NWCG.[27] Their SOPs also stated that drops ‘should be downhill’, that they were to provide ‘the final heading for the run and the altitude for the drop start point’, the drop height ‘minimum is 150 ft above the top of the vegetation for heavy tankers’ and that ‘The airtanker pilot is responsible for maintaining safe drop heights.’

Bomber 139’s captain’s preference for the circuit, and what they reported teaching, was to fly a close circuit that kept the target in sight. To achieve this, the captain would initiate a descending base turn abeam the drop start point and maintain visual contact with the target through the first half of the base turn. About halfway through the base turn, the captain would extend their visual scan from the target to include the exit and on roll-out to wings level, the aircraft would be positioned on short final close to the start of the drop. 

On the first drop of the accident flight the captain reported the large circuit was due to the manoeuvring required to join up with BD123. The captain recalled that this resulted in a very wide base but final was flown clear of the smoke on the right side of the fire-line and at a normal angle of approach (Figure 13).

However, in the low ground of the depression the fire-line had extended further west, which resulted in a climbing right turn away from the fire before the downhill section of the drop and the captain did not see the exit before starting the second circuit. Subsequently, BD123 led them further east into the drift smoke for the next approach to turn the line in a more westerly direction. This resulted in Bomber 139 approaching the target through the smoke and the captain did not identify the target until they were at the start of the drop and had not seen the exit.

Figure 13: Accident flight first and second drops

Source: Google Earth, annotated by ATSB

Energy management

United States Federal Aviation Administration handbook

The FAA provided a description of aircraft energy management in their Airplane Flying Handbook (FAA-H-8083-3C) Chapter 4: Energy Management: Mastering Altitude and Energy Control. The introduction to the chapter described it as ‘about managing the airplane’s altitude and airspeed using an energy-centered approach.’ While this chapter was not written for a LAT rapid inflight weight reduction, it included 4 concepts of relevance to the accident. They were: Irreversible Deceleration and/or Sink Rate, Total Mechanical Energy, Energy Height or Total Specific Energy and Specific Excess Power, which were described as follows:

Irreversible Deceleration and/or Sink Rate

Unrecoverable depletion of mechanical energy as a result of continuous loss of airspeed and/or altitude coupled with insufficient excess power available under a given flight condition. Failure to recover above a certain critical AGL [above ground level] altitude results in the airplane hitting the ground regardless of what the pilot does.

Total Mechanical Energy

Sum of the energy in altitude (potential energy) and the energy in airspeed (kinetic energy).

Energy Height or Total Specific Energy

Measured in units of height (e.g., feet), it represents the airplane’s total energy per unit weight. It is found by dividing the sum of potential energy and kinetic energy by the airplane’s weight. It also represents the maximum height that an airplane would reach from its current altitude, if it were to trade all its speed for altitude.

Specific Excess Power

Measured in feet per minute or feet per second, it represents rate of energy change—the ability of an airplane to climb or accelerate from a given flight condition. Available specific excess power is found by dividing the difference between power available and power required by the airplane’s weight.

The FAA’s equation for specific excess power was Ps = (T-D) * V / W, where Ps = specific excess power, T = engine thrust, D = total drag, V = airspeed, and W = aircraft weight.

Flight data review

The ATSB reviewed the 4 retardant drops conducted by Bomber 139 on the day of the accident (including the third partial drop followed by the fourth accident drop), and analysed the height, vertical speed, airspeed, angle of attack and N1 parameters. The 2 states compared for each run were at the minimum drop height and at the go-around thrust lever setting on completion of each drop (based on the thrust levers reaching a stable setting after being advanced at the end of each drop). The data for the 4 drops was as follows:

• On the first drop, Bomber 139 reached a minimum height of 46 ft with 120 ft/min rate of descent at 113 kt and 61‍–‍66% N1. Minimum airspeed was 109 kt at 50 ft and maximum angle of attack was 19.7° at 64 ft.[28] On reaching the go-around thrust lever setting, the height was 64 ft with 360 ft/min rate of climb at 128 kt and 77‍–‍86% N1. Idle thrust was employed intermittently near the start of the drop, which resulted in a V-shaped airspeed trace during the drop (Figure 14) and in BD125 closing on Bomber 139 during the drop (Figure 15).
• On the second drop, Bomber 139 reached a minimum height of 69 ft with 120 ft/min rate of descent at 132 kt and 67‍–‍70% N1. Minimum airspeed was 129 kt at 77 ft. On reaching the go‑around thrust lever setting, the height was 140 ft with 480 ft/min rate of climb at 136 kt and 85‍–‍88% N1.
• On the third drop, Bomber 139 reached a minimum height of 78 ft with 600 ft/min rate of descent at 126 kt and 70‍–‍71% N1. Minimum airspeed was 124 kt at 82 ft. On reaching the go‑around thrust lever setting, the height was 153 ft with zero vertical speed at 129 kt and 87‍–‍88% N1.
• On the accident drop, Bomber 139 reached a minimum height (before initiating the go‑around) of 57 ft with 1,800 ft/min rate of descent at 110 kt and 30‍–‍31% N1. Minimum airspeed was 106 kt at 59 ft. On reaching the go-around thrust lever setting, the height was about 30 ft with 480 ft/min rate of descent at 105 kt and 55‍–‍60% N1.

Figure 14: FDR plot of first drop with minimum airspeed reference

Source: ATSB

Figure 15: Aspect change of Bomber 139 from start of first drop (left) to go-around (right)

Birddog 125 closing on Bomber 139 during the first drop. Source: Department of Fire and Emergency Services, annotated by ATSB 

The effects of the FAA‑specific excess power equation were evident in the 4 retardant drops on the day of the accident. The high engine thrust (N1) and large weight reduction on the first 3 drops resulted in a positive reversal of the vertical speed and airspeed at the end of those drops. However, the idle thrust and smaller weight reduction on the accident drop resulted in the airspeed and height decaying at the end of the drop as the captain arrested the aircraft’s significantly higher rate of descent.

This was consistent with a negative specific excess power (drag greater than thrust) and the FAA’s description of an Irreversible Deceleration and/or Sink Rate. Table 2 summarises the parameters from the flight data review of the 4 drops.

Table 2: Summary of parameters

After the accident, the operator informed the ATSB that they had increased their minimum drop height from 150 ft to 200 ft and their target drop speed to 1.3 Vs for downhill drops with flaps-40 and 1.35 Vs for flat terrain drops with flaps-30. They later reported to the ATSB that they believed the recalculation and lowering of drop speeds between partial drops was a ‘significant contributing factor to the accident.’ 

The Bomber 139 captain reported that reducing the target drop speed (1.25 Vs) following a partial drop provided them with the same stall speed safety margin at the start of each drop as the stall speed also reduced following a partial drop.

If a performance problem occurred in the go-around from the first drop of a split load, the crew could jettison the remaining load to gain an airspeed safety margin equivalent to the loss of the full load. However, if a performance problem occurred on the final drop of a split load, then the additional safety margin was reduced as it was limited to the weight of the remaining partial load. 

The release of the entire load on the first drop of the accident flight would have reduced the aircraft weight by about 28% (124,640 lb to 89,440 lb), whereas jettisoning the second drop reduced the aircraft weight by about 9% (97,850 lb to 89,440 lb).[29] In consideration of the implications of recalculating the drop speed between partial drops for a LAT, the ATSB used the FAA’s mechanical energy formulas to assess the potential gain in energy-height from higher drop speeds.

The drop speed at the end of the accident drop was used as the datum and compared with the target drop speed lower deviation limit (refer Figure 17), target drop speed, the previous end of drop speed, which was also the minimum airspeed during that drop, and previous target drop speed. The results are presented in the following table.

Table 3: Drop speed energy-height differences

From Table 3, the actual airspeed at the end of the accident drop resulted in a loss of about 90 ft in energy-height compared to the target drop speed. However, the recalculation of the target drop speed resulted in a loss of about 167 ft in energy-height from the full load target drop speed. Therefore, a comparison between the full load target drop speed and the accident drop speed found that about 65% of the energy-height loss was from recalculating and lowering the drop speed and about 35% of the energy-height loss was from flying slower than the target drop airspeed. 

The divergence in the go-around performance of Bomber 139 between the end of the third drop and the final drop is presented in the following FDR plot (Figure 16), in which the parameters are aligned with the end of each drop. The ATSB noted the following from the plot:

• the third drop (RETARDANT DROP 1) exits with a higher N1, airspeed, vertical speed, and lower angle of attack (AOA) than the final drop (RETARDANT DROP 2)
• the pitch attitudes and radio heights diverge at the exit, noting the third drop overflew the depression
• the airspeeds diverge as the vertical speeds and N1 converge.

Figure 16: FDR plot of the third (left labels) and final (right labels) drops

Source: ATSB

Crew resource management

Introduction

Crew resource management (CRM) and controlled flight into terrain (CFIT) prevention training were LAT contract requirements in Australia and the US. It was the responsibility of the contractor and not the contracting agency to deliver the training and provide evidence of completion. The operator, Coulson Aviation, delivered these training programs to their personnel during their northern hemisphere spring training in April 2022. The relationship between CRM and SOPs was described in FAA advisory circular AC 120‑71B (2017): Standard Operating Procedures and Pilot Monitoring Duties for Flight Deck Crewmembers, as follows:

Effective crew coordination and crew performance depend on the crew’s having a shared mental model of each task. That mental model, in turn, is founded on SOPs. 

Monitoring performance can be significantly improved by: (1) developing and implementing effective SOPs to support monitoring and cross-checking functions, and appropriate interventions, (2) by training crews on monitoring strategies, and (3) by pilots following those SOPs and strategies.

Coulson Aviation standard operating procedures

Target drop height

The Coulson Aviation B737 flight crew operations manual (FCOM Volume 1, July 2019) RADS flight pre-drop checklist, included a step to ‘Set the DH REF [decision height reference]/radio altimeter minimums to target drop height.’ However, their B737 Fireliner quick reference handbook normal checklist (QRH - FCOM Volume 3, August 2019) abbreviated this checklist item to ‘Altimeters set’, as per the FAA approved flight manual supplement for the Fireliner modification. The CVR revealed several instances of the pilots cross-checking their altimeter pressure settings with each other and other aircraft, and confirming their drop speeds to bug, as per their QRH pre‑drop checklist. However, there was no reference to a target drop height or decision height setting.

The aircraft captain confirmed that the decision height setting on the radio altimeter was not used because it did not provide an aural alert and because the pilot flying is predominantly ‘eyes-out’ during the drop and therefore not monitoring the decision height light on the attitude indicator. The co-pilot did not use the decision height on the radio altimeter and did not believe it was a SOP to use it because it could cycle on-and-off during a retardant drop over undulating terrain.

The operator confirmed that, despite the documented FCOM requirement, the decision height setting was not used. They further stated that it could be a distraction if it activated intermittently during a retardant drop.

According to the aircraft captain, 150 ft was the standard target drop height, with the caveat that ‘drifting down’ to lower drop heights was not uncommon at lighter coverage levels. The operator confirmed that while they used the USFS LAT minimum drop height of 150 ft as their standard target drop height at coverage level 3 (accident drop setting), it was not published in their fixed‑wing flight operations manual or their B737 and C-130 FCOMs. The operator also reported that pilots may be tempted to drop below the standard target drop height at light coverage levels as the retardant will be subjected to less wind interference and this will improve the accuracy of the drop. 

At their initial interview, the co-pilot stated they did not believe there was a minimum drop height, but also reported having limited industry knowledge.

At a follow-up interview, the co-pilot confirmed that none of the captains had briefed a target drop height and that they had flown with all the operator’s B737 captains. The co-pilot reported understanding that 150 ft was the standard drop height figure, but also stated that ‘150 ft is the highest you want to be for that coverage level [3] depending on wind drift.’ The co‑pilot also reported that the altitudes briefed by BD123 were for terrain clearance on the approach and exit and indicated it would be a downhill drop, and that they were not target drop heights. Finally, the co‑pilot advised that they felt confident about speaking up if something felt unsafe and, on that day, ‘there was nothing that seemed unsafe’. 

Monitoring duties

The aircraft captain reported the co-pilot’s duty was to scan inside and out and call deviations, such as airspeed, if different from what was briefed, and confirmed the 150 ft drop height published by the USFS was a minimum height. The CVR indicated that the target drop speed was the only parameter briefed and the last announcement from the co-pilot was about 70 seconds before impact to inform the captain that the pre-drop checklist was complete. The co-pilot reported their focus of attention was inside on the airspeed, which they recalled was on their target drop speed of 118 kt, and on the radio altimeter, which they could not recollect but believed ‘would have been about 100 ft’. The co-pilot was listening to the engine noise but was not monitoring engine thrust and couldn’t recall the thrust lever settings. 

A colleague of the co-pilot confirmed the operator’s CRM philosophy was not to overcommunicate – ‘brief first, then only communicate if there is a deviation from the brief’. Figure 17 depicts the operator’s B737 Fireliner FCOM Volume 1 Retardant drop procedure. With reference to the tolerances in Figure 17, the FDR data indicated that the aircraft descended below their standard target drop height deviation limit of -25 ft about 13 seconds before impact, and remained below it until impact. The airspeed was below their target speed deviation limit for about 11 seconds before impact and 4‍–‍5 seconds before the captain started to advance the thrust levers.

Figure 17: Retardant drop procedure

Source: Coulson Aviation

Crew resource management training

The operator’s CRM training package, delivered in April 2022, included material on the following topics:

• training objectives
• introduction to CRM
• current SOPs
• discussion
• roles and responsibilities.

The training objectives topic highlighted one of the challenges in their industry as follows:

Given the dynamic nature of fire, SOPs allow a relatively wide range of individual preferences among captains, this can and does lead to a level of confusion within the cockpit, especially among new/inexperienced crew members…

The introduction to CRM topic indicated that human error is a contributing factor to most accidents, and highlighted the following factors:

Problems are associated with poor group decision making, ineffective communication, inadequate leadership and poor task or resource management. 

The ‘current SOPs’ topic within the CRM training package included detail of the FCOM Volume 1 Retardant drop procedure (Figure 17) and the QRH Pre-drop checklist. The FCOM Volume 1 Pre‑drop checklist procedure, which included setting the radio altimeter decision height to the target drop height was not included. 

The ‘discussion’ topic listed the following items:

• callouts
• calling a go-around
• when and how to communicate discomfort
• identifying hazards
• when should FO/FE [first officer/flight engineer] hit the button [emergency dump]?

The ‘roles and responsibilities’ topic included, but was not limited to, the following:

• captain – Position aircraft on Show-Me to observe line and start/stop point
• co-pilot – Backup captain on configuration/altitude/airspeed/checklist/drop/escape. 

Controlled flight into terrain prevention training

The operator’s CFIT prevention training package, also delivered in April 2022, included material on the following topics:

• What is CFIT?
• Why does it happen?
• How can I avoid CFIT?
• What are some of the contributing factors to CFIT?
• CFIT accident reviews [which included a general aviation accident, an airliner accident and 3 airtanker accidents].

The first 3 topics were based on material from the FAA’s general aviation joint steering committee (GAJSC) that defined CFIT as follows:

CFIT is defined as an unintentional collision with terrain (the ground, a mountain, a body of water, or an obstacle) while an aircraft is under positive control.

The operator had highlighted several sections of the FAA’s material, which included the following:

…the General Aviation Joint Steering Committee (GAJSC) observed that a clear majority of the CFIT accidents in a typical year occur in daylight, and with visual conditions.

Other top causes of CFIT are…unrealistic aircraft performance expectations (e.g., high density altitude, tailwinds on approach).

While the operator’s accident case studies included multi-crew operations, their introductory material from the GAJSC was directed at single-pilot general aviation operations. Consequently, the avoidance strategies focused on assisting single-pilot operations, such as a FRAT and technological devices such as moving maps with terrain overlays. The slides in their CFIT training package did not include CRM and SOPs as prevention strategies in the retardant drop. 

The role of multi-crew CRM to prevent CFIT was described in FAA advisory circular AC 120-51E (2004): Crew resource management training, as follows:

Effective monitoring and cross-checking can be the last line of defense that prevents an accident because detecting an error or unsafe situation may break the chain of events leading to an accident. This monitoring function is always essential, and particularly so during approach and landing when controlled flight into terrain (CFIT) accidents are most common.

The FAA advisory circulars (AC 120-51E and AC 120-71B) were directed towards the commercial transport sectors to assist operators with the development and implementation of their CRM training, SOPs, and pilot monitoring duties. Consequently, they addressed the multi-crew environment and followed recommendations from a 1994 US National Transportation Safety Board safety study (NTSB/SS‑94/01): A review of flightcrew-involved, major accidents of U.S. air carriers, 1978 through 1990. The NTSB’s review found that 23% of the contributing errors were monitoring and/or challenging errors, which occurred in 84% of the accident sequences. They reported that ‘Most of the errors that were not monitored or challenged played very important roles in the accidents.’

Tasking agency standard operating procedures

National Aerial Firefighting Centre

The National Aerial Firefighting Centre (NAFC) was formed by the Australasian Fire and Emergency Services Authorities Council (AFAC) in 2003 through a partnership between the federal and state/territory governments to deliver cooperative national arrangements for the provision of aerial firefighting resources. The states and territories recognised that if they had one contract for operators to follow then they could easily be moved around the country as necessary to facilitate the sharing of the resource and associated cost. To support this process, the NAFC produced Operational Standards, which provided the minimum standards for a NAFC contract. These were procurement standards, with the respective state/territory emergency services having responsibility for operational policies and procedures. 

The NAFC contracted about 150 firefighting aircraft, including LATs, for use by state and territory emergency service and land management agencies across Australia. At the time of the accident, the NAFC Members participating in the LAT program were New South Wales (NSW), WA, Victoria, and Queensland.

The NAFC-contracted aircraft were assigned a nominated operational base where they were tasked by the NAFC Member’s emergency services. Each of the NAFC Members participating in the LAT program confirmed to the ATSB that they had developed their own SOPs for aerial firefighting, which included LAT SOPs.[30] A LAT service provider (such as Coulson Aviation) therefore operated under a NAFC contract, which required them to comply with the NAFC Standards and the Member state’s SOPs associated with their base.[31] However, their base could change throughout the season in response to the fire threat, which was acknowledged in the National Aerial Firefighting Strategy 2021‍–‍26 section on risk management and resource sharing, as follows:

During any single fire season, it is highly likely that risk factors will lead to Commissioners and Chief Officers wanting to redeploy assets within jurisdictions and potentially redeploy assets across jurisdictional borders. Aerial assets offer the greatest flexibility to redeploy both within and across jurisdictions. LATs and strategic mapping aircraft can operate in more than one jurisdiction on the same day.

To date all aircraft have been assigned to a state or territory in line with the co-contribution commitment which has been part of the NAFC Resource Management Agreement since its inception. The 2020 Royal Commission into National Natural Disaster Arrangements introduced the notion that national assets, particularly LATs and Type 1 Rotary Wing, be tasked according to ‘greatest need’ and their deployment and application would be considered on a national rather than jurisdictional basis.

The NSW Rural Fire Service (RFS), which had their own LAT and could have multiple additional LATs on contract in any season, reported that if one of their LATs was re-deployed to a non‑participating Member state or territory, then it would be operated in accordance with the NSW RFS LAT SOPs, but not inconsistent with local requirements. However, the RFS acknowledged that the non-participating Member state or territory would need to brief the LAT and Birddog operators on any differences, which would require that Member to be familiar with the NSW LAT SOPs.

The federal government contributed funding to the NAFC national fleet of firefighting aircraft but the primary responsibility for aerial firefighting remained with the states and territories and therefore, Member states produced their own SOPs. However, AFAC stated that their Member states and territories would be undertaking greater resource sharing in the future. Therefore, they concluded that the industry should develop national standards and that the NAFC business unit, under the AFAC banner, is a mechanism for this as they had existing areas of Member collaboration in contracts and innovation for aerial firefighting. 

Following the C-130 LAT accident in NSW on 23 January 2020 (AO-2020-007), a statement of intent for aviation safety by the NAFC led to the formation of the Aviation Safety Group (ASG). The ASG was supported by the NAFC and had Members from agencies in all jurisdictions, including New Zealand and a representative from the Civil Aviation Safety Authority. During the B737 LAT accident investigation, the NAFC advised the ATSB that the ASG had developed a list of priorities that included:

• airspace management, including cross-border operations
• human factors and CRM training – for agency personnel, especially those who interface with aviation
• consistent doctrine – review and national standardisation (where practical) of standards and operational procedures
• safety information framework – reporting and sharing of safety information and messaging
• common cross-border radio operations, including hardware innovations and frequency management
• a national LAT Lead Plane pilot training standard.

At the time of writing, the ASG had not identified a national LAT SOP as a priority item.

United States Forest Service standards

The USFS has managed wildland fires on national forests and grasslands for over 100 years, working with federal, state, tribal and local partners. In 1976 the National Wildfire Coordinating Group (NWCG) was established to provide national leadership for these partners. The NWCG had an Executive Board comprised of 12 member entities and their role was described as follows:

The Executive Board works in a collective, collaborative and consensus-seeking manner to accomplish the NWCG mission. NWCG develops the interagency standards that are core to implementing its mission, which the member agencies then choose to implement through their own directive systems.

The USFS reported that federally‑owned lands made up most of the western US and some states had adopted the USFS SOPs as their de facto standard where they had a small area of state responsibility. Local requirements were addressed with a briefing package at each airtanker base. In illustrating the multi‑jurisdictional nature of aerial firefighting, the USFS provided an example operational scenario where a LAT could be employed on a fire in Montana in the morning, in Southern California in the afternoon and end the day in Washington state on another fire.

They reported that these movements across geographical boundaries required national standards to be in place, and that most of the aviation standards were lessons learned from accidents. The national standards were accepted as a better way of operating and they considered the first 30 days after an operator moved between jurisdictions to be a high-risk period due to the change in rules. Therefore, given the movement of aerial firefighting aircraft and operating personnel between the US and Australia, they considered that closer harmonisation of standards would likely bring safety benefits across the industry.   

Benchmarking of tasking agency standards and procedures

Following the C-130 LAT accident in 2020, the ATSB obtained a copy of the NSW RFS LAT SOPs (Operating guidelines for air tankers operations 2018) for review. That review included benchmarking the NSW SOPs against the USFS SOPs, noting the accident aircraft was a US registered LAT and the operator was also contracted by the USFS during their fire season.

The ATSB found several procedures in the USFS SOPs, relevant to the C-130 accident, which were omitted from the NSW SOPs. They were a task rejection policy and procedure, known as a ‘turn down’ in the USFS SOPs under their policy of ‘Risk Refusal’, and procedures for deploying a LAT without aerial supervision. These omissions were raised as safety issues by the ATSB in August 2022 and directed to the NSW RFS for action. The safety issues were closed on 2 May 2023 as adequately addressed.

Following the B737 LAT accident, the ATSB reviewed the WA LAT SOPs (Large air tankers operational procedures 2021-22 version 3.3), current at the time of the accident. This was a joint publication between the WA Department of Fire and Emergency Services and the WA Department of Biodiversity, Conservation and Attractions. The ATSB compared procedures relevant to the C‑130 and B737 accidents from the WA SOPs with the 2018 NSW RFS LAT SOPs and the USFS airtanker SOPs (Standards for airtanker operations 2019) and NWCG aerial supervision SOPs (Interagency aerial supervision guide 2017).[32]^,[33] While the USFS airtanker and NWCG aerial supervision SOPs were separate documents, the WA and NSW LAT SOPs also incorporated the LAT aerial supervision procedures.

The WA SOPs included 3 target identification techniques (tactical flight profiles), but they did not include the Lead or Chase-Position profiles, which were employed on the day of the accident. They also did not include a drop height and there was no task rejection policy or procedure. However, they did include the following requirements:

The BRDG [Birddog] AAS [air attack supervisor] is to confirm the drop zone clear with the Primary AAS or Ground Controller prior to the LAT commencing the final drop run.

The LAT must not drop until DROP ZONE CLEAR has been confirmed.

The NSW SOPs included the same target identification techniques as the WA SOPs and therefore they also did not include the Lead or Chase-Position profiles. However, the NSW SOPs did include reference to a LAT drop height and stated the following:

The Birddog aircraft will assess the Air Tanker for drop height to ensure a height of approximately 150 feet above the canopy or bare ground is attained. Low drop heights should be avoided to minimise the risk during low level flight operations and ground crews in the vicinity of the drop.

The NWCG SOP tactical flight profiles consisted of the Show-Me profile, Chase-Position profile, and Lead profile. All 3 of these profiles were flown on the day of the accident. The USFS and NWCG SOPs both stated the drop height for a LAT was a ‘minimum’ of 150 ft above the top of the vegetation. Below the minimum drop height reference in the NWCG SOPs were the following notes:

It is important for the retardant to “rain” vertically with little or no forward movement. The airtanker pilot is responsible for maintaining safe drop heights. 

The preference for downhill drops was captured in the NWCG and WA SOPs. The NWCG SOPs stated that the approach should be ‘downhill, down canyon, down sun’. The WA SOPs stated, ‘implement where possible that all exits are downhill’. The NSW SOPs did not include an explicit statement to this effect but their flight safety considerations included ‘Terrain may limit firebombing run directions and the availability of exits.’ 

While not a factor in either the C-130 or B737 accidents, the NWCG SOPs included emergency overrun procedures (as did the Birddog operator). In the Lead profile (accident flight profile), the Birddog pilot cannot see the LAT and therefore there is a risk of the LAT overrunning the Birddog. The WA and NSW SOPs did not include procedures for how to mitigate the risks associated with an overrun, which include wake turbulence-induced loss of control for the smaller aircraft, mid-air collision and terrain collision associated with an avoidance manoeuvre. 

Development of the target drop height and speed

Background

The NAFC contract for the operator’s multi-engine airtankers required their aircraft to have US Interagency Airtanker Board (IAB)[34] approval in addition to any local NAFC Member approval requirements. Multi-engine airtankers were approved for use in the US by the IAB, which was a contract requirement of US government agencies at the federal, state, and local level. The current Interagency Airtanker Board procedures and criteria for evaluating aircraft and dispensing systems used in aerial firefighting were published in 2013. According to the USFS, the theory of aerial application of retardant was as follows:

For proper aerial application of retardant to the fuel (vegetation), all forward momentum of the retardant cloud should stop, and the liquid should rain down vertically. This allows an even coating and better vertical penetration into the tree or brush canopy.

Section 7 (B) of the IAB procedures on tank system criteria, described the retardant coverage levels as the gallons per 100 square feet (gpc) and that the flow rate from an aerial delivery system is ‘the most significant controllable factor’ in determining the level of coverage obtained. The USFS considered drop tests to be the most accurate means of measuring ground pattern performance. The IAB procedures defined a drop test as a ‘dynamic flying test of the aircraft retardant delivery system over a cup/grid matrix, which is used to determine the coverage level production of the system for each drop type.’ According to the USFS:

A balance of factors is required for airtankers to meet the coverage level requirements. This balance is established during the IAB testing and is done in a controlled environment with little to no wind on flat terrain. Multiple passes are made over a test grid at different tank controller settings, airspeeds, and heights. This establishes the flight profile and tank controller settings that produce acceptable ground pattern performance under test conditions.

Section 3 (H) of the IAB procedures on multi-engine airtanker requirements included a descent profile requirement as follows:

Aircraft shall be capable of descending at Board-approved maximum operating weight along a 13 percent (7.4°) slope for 30 seconds to 5,000 ft pressure altitude in the drop configuration without exceeding maximum drop speed. At the 25 second mark, a full load of water shall be dispensed at coverage level 6 and again at coverage level 8 and demonstrate no more than 7% difference in average flow rate from grid results.

Drop height requirements

Section 3 of the IAB procedures on multi-engine airtanker requirements did not include a minimum drop height. However, section 7 (B) of the IAB procedures on tank system criteria stated that the requirement was that drops ‘shall be made at a 200-foot drop height or the minimum safe drop height if greater than 200 feet.’ The operator had conducted IAB drop testing with their B737 Fireliner and described the process as follows:

The IAB grid test evaluates the performance of the drop tank in negligible wind conditions at various coverage levels, aircraft configurations, ground speeds, and drop heights. As part of that process, high speed video was evaluated to determine the optimal drop height which is the lowest height that allows the forward trajectory of the retardant to cease, thus minimizing shadowing effect so that the fuels are evenly coated.

As part of the grid test evaluation, the IAB provided the operator with a table of the recommended minimum grid test drop heights under the heading of ‘minimum altitude above the height of the fuel to eliminate the forward momentum, i.e. eliminate shadowing[35] (feet).’ Each drop height corresponded with a controller setting for the coverage level to be achieved. For the operator’s B737 Fireliner, the minimum drop heights were over the range of 150‍–‍200 ft for the controller settings 1‍–‍8 (150‍–‍185 ft for the C-130). For all the LATs listed in the table, the recommended minimum grid test drop height was 150 ft for controller settings 1‍–‍3 (setting 3 was used on the day of the accident). The operator reported that they adopted the heights in the table as their standard target drop heights for training and operations.

The USFS clarified that the table drop heights were for determining ground pattern performance during testing only and were not operational requirements. The operational drop heights were identified in the contracts, which may reference interagency operational handbooks. They reported that environmental variables, such as wind, terrain, fire behaviour and suppression tactics may require a drop height higher than was demonstrated during testing. 

The USFS Airtanker Services contract standards[36] stated a minimum drop height of 150 ft, which was consistent with their airtanker SOPs and the NWCG aerial supervision SOPs. Their contract standards also stated that ‘Retardant shall be dropped as accurately as possible on the designated target areas of the fire.’ The USFS confirmed that if an operator conducted drops below 150 ft, they would be non-compliant with the contract. However, they acknowledged that operators have dropped lower to improve their accuracy or with the assumption that it will assist the ground firefighters by providing a thicker line of retardant. 

In March 2000, the USFS published the results of a research project into safe drop heights for fixed-wing airtankers (Lovellette). The purpose of the project was to examine the cessation of the forward momentum of the retardant cloud, which served 2 purposes:

• prevent shadowing of the fuel (as discussed above)
• reduce the risk to ground personnel from flying debris struck by the retardant cloud.

Video analysis of drop tests from the project led to the development of a relationship between safe drop height, load size, and the peak flow rate (Lovellette, 2000). The danger to ground firefighters was minimised if the retardant lost its forward momentum and fell vertically (Lovellette, 2000). This provided a safe drop height that increased with higher controller settings that use a higher flow rate to achieve the required coverage level. 

Another review of drop patterns by the USFS found that lower coverage level patterns showed distortion under higher wind speeds while the higher coverage levels were relatively unaffected (Suter, 2005). Suter (2005) noted that higher drops increased the retardant cloud susceptibility to wind erosion and the maximum coverage levels tended to decrease as the drop height increased. While the analysis of drop patterns showed disadvantages to increasing drop height, Suter (2005) also noted that increasing drop height is ‘safer for flight crew and ground personnel.’  

The USFS confirmed that Lovellette’s (2000) safe drop height project was for the safety of ground personnel and not the airtankers. The focus on the safety of ground personnel was reinforced with their interagency aviation safety alert in 2019 (IASA 19-02)[37] on the subject: Retardant Safe Drop Height, area of concern: Safety of Ground Personnel. From the safety alert:

The force of the retardant dropped from too low of an altitude can topple trees up to 90 feet in height and a trunk a foot in diameter.

Pilots must remember that lower is not always better. Drops that are too low fail to provide retardant in an efficient manner with the desired coverage level. This is not only dangerous but fails to provide the support ground crews require.

Drop speed requirements

Section 3 (E*)[38] of the IAB procedures on multi-engine airtanker requirements included the following points about the effect on flight conditions from the release of retardant:

Release of retardant in all normal drop configurations and at all normal drop speeds does not result in dangerous or seriously objectionable flight conditions.  

1. The minimum drop speed is not less than the Vmc (minimum control speed), nor 1.25 Vs (stall speed), both speeds being evaluated in the drop configuration.

2. The maximum drop speed does not exceed Va (design maneuvering speed).

According to the USFS, environmental variables may require an increase in the drop speed of 10‍–‍20 kt, but excessive drop speeds could adversely affect the firefighting effort as per the following description:

At higher airspeeds, the retardant can shear at the tank opening and not drop in a column, thus reducing coverage levels. If the retardant cloud has too much forward momentum the retardant will coat only one side and create a shadowing effect, allowing fire spread. Forward momentum creates a hazard to ground crews if they are inadvertently in the path of the drop. Too slow of a release speed is a safety of flight issue.

The USFS reported the drop test airspeed requirements dated back to 1976 and the rationale for the prescribed parameters could not be found. However, they reported that the speed is not limited to the maximum load the applicant is seeking approval for and the relevant condition was ‘Release of retardant in all normal drop configurations and at all normal drop speeds does not result in a dangerous or seriously objectionable flight condition.’ They further stated that: 

Release can be one drop of the entire volume or split drops up to 4 drops so all normal drop configurations could be between the full volume down to a quarter volume. To produce the desired coverage level the speed may change but not go below the minimum drop Vmc nor 1.25Vs or above the maximum of Va.

The operator incorporated tables for weights and speeds in their B737 Fireliner QRH normal checklist. The weights column was based on 5,000 lb increments in the aircraft operating weight from 85,000 lb to 135,000 lb. There were 4 columns for airspeeds, which included a column for 1.25 Vs that was also annotated as V-drop (target drop speed). This table provided a quick reference for the crew to update their drop speed between partial drops and was consistent with the IAB minimum drop speed requirements for evaluating multi-engine airtankers. 

Emergency dump

Section 7 of the IAB procedures on tank system criteria required ‘features that enhance safety in the event of an in-flight emergency.’ These features included an independently‑controlled emergency dump system. The operator’s flight manual supplement for the B737 RADS indicated the emergency dump system could jettison a full load in 2 seconds.

Previous controlled flight into terrain accidents

The ATSB reviewed several LAT and multi-engine water-bomber CFIT accidents during the investigation. The following 3 LAT accidents from Canada and the US were discussed with various personnel during the investigation.

Canada

Cranbrook, British Columbia – 2003

On 16 July 2003, a Lockheed L-188 Electra (Tanker 86) collided with terrain, fatally injuring the 2 crewmembers, during the go-around from a first partial retardant drop (A03P0194). In the final seconds before impact there were 2 separate retardant drops into the trees and ground witnesses heard the engine noise level increase significantly. The Transport Safety Board (TSB) of Canada noted that the ‘ridge that the aircraft struck blended into the rising terrain and was not obviously separate from its surroundings.’ The TSB published the following findings as to causes and contributing factors:

For undetermined reasons, the Electra did not climb sufficiently to avoid striking the rising terrain.

Given the flight path and the rate of climb chosen, a collision with the terrain was unavoidable.

The characteristics of the terrain were deceptive, making it difficult for the pilots to perceive their proximity and rate of closure to the rising ground in sufficient time to avoid it.

Lytton, British Columbia – 2010 

On 31 July 2010, a Convair 580 (Tanker 448) collided with terrain, fatally injuring the 2 crewmembers, while crossing the edge of a ravine in the side of the Fraser River canyon before descending on the fire located in the ravine (A10P0244). The TSB findings as to causes and contributing factors included the following:

Visual illusion may have precluded recognition, or an accurate assessment, of the flight path profile in sufficient time to avoid the trees on rising terrain.

Visual illusion may have contributed to the development of a low energy condition which impaired the aircraft performance when overshoot action was initiated.

United States

Modena, Utah – 2012

On 3 June 2012, a Lockheed P2V-7 (Tanker 11) collided with terrain while following the Lead Plane onto final for its second retardant drop of the day at the same location (WPR12GA243), fatally injuring the 2 crewmembers. From their analysis of the 2 drops using the flight recorders, the NTSB made the following observation:

The most significant difference between the first drop approach and the accident drop approach was that Tanker 11's first approach into the drop zone was initiated from a higher altitude, about 1,000 feet above the ridge line surrounding the west side of the fire, and had a relatively steady rate of descent on to the final drop course. In comparison, the accident flight approached the final drop course from a lower altitude that approximated the elevation of the ridge lines surrounding the west side of the drop zone in level flight while it turned onto the final drop course.

The CVR revealed an ‘airspeed’ call from the first officer about 2 seconds before impact but no altitude or height calls. The NTSB determined the probable cause(s) of the accident to be:

The flight crew’s misjudgement of terrain clearance while maneuvering for an aerial application run, which resulted in controlled flight into terrain. Contributing to the accident was the flight crew’s failure to follow the lead airplane’s track and to effectively compensate for the tailwind condition while maneuvering.

The hidden hill visual illusion

The ATSB discussed previous LAT CFIT accidents and the topic of visual illusions in the low-level environment with the Bomber 139 captain. The TSB of Canada explored the problem of visual illusions during low-level flight in their reports of CFIT LAT accidents at Cranbrook, British Columbia (2003) and Lytton, British Columbia (2010). The Cranbrook accident was during the go‑around from a partial drop and the Lytton accident was during the approach to the drop zone. They concluded that visual illusion may have contributed to the accidents by delaying the crew’s recognition of the terrain. 

The Bomber 139 captain presented lessons learned from the NTSB’s report into the LAT accident at Modena, Utah (2012) at the US National Aerial Firefighting Academy for a period of 3‍–‍4 years. During this period, the captain developed a belief that the crew of Tanker 11 experienced the ‘hidden hill’ illusion, which occurs when the terrain in the foreground blends into the terrain in the background, consistent with the TSB reports. The captain reported that ‘hidden hill’ was a common industry term they used in their annual CFIT prevention training and the ATSB noted the Tanker 11 accident was one of the case studies in the operator’s CFIT prevention training package.

While the ATSB was at the B737 LAT accident site in WA, the Bomber 139 captain visited the site with senior management personnel from the operator. At the start of the visit, the ATSB escorted them to the location of the start of the accident drop, which provided them with a view of the accident site from across the depression. The captain later reported the following observation:

If you looked away from where the airplane hit to where the vegetation still existed. It was hard to see any depth or any descent throughout that at all or the fact that there was a rise on the other side.

Safety analysis

Introduction

On 6 February 2023, a Coulson Aviation Boeing 737 large air tanker (LAT), callsign Bomber 139, was tasked to conduct fire retardant drops in the Fitzgerald River National Park, Western Australia (WA). The aircraft was tasked by the WA Department of Fire and Emergency Services (DFES) at the request of the WA Department of Biodiversity, Conservation and Attractions (DBCA). The captain was the pilot flying, in the left seat, and the co-pilot was the pilot monitoring, in the right seat.

Bomber 139 conducted 3 flights on that day, which was the last day of the flight crew’s tour of duty. The first 2 flights each comprised a single full load drop at the fire ground with a return to Busselton. On the third flight, following an initial partial drop, the crew conducted a second drop, which released all the remaining retardant. The aircraft collided with terrain during the go-around from the second drop. The wreckage and site examination, meteorological information, flight data recorder (FDR), cockpit voice recorder (CVR) and flight crew interviews found no technical fault with the aircraft, evidence of flight crew fatigue or hazardous weather condition that contributed to the accident.

This analysis will discuss the operational and human factors at the fire ground that influenced the accident, in addition to the operator and tasking agency standard operating procedures (SOPs). It will also discuss the need for national LAT SOPs following this accident and a previous collision with terrain of a Lockheed C-130 LAT during the go-around from a partial load drop in New South Wales on 23 January 2020 (ATSB investigation AO-2020-007).

Accident sequence

On Bomber 139’s arrival at the fire ground on the first flight of the day, Birddog 125 (BD125) provided a Show-Me profile for Bomber 139 before Bomber 139 conducted a close left-hand circuit to the target with BD125 in the Chase-Position. The fire was burning on the right side of the drop line, which kept the circuit clear of smoke.

On arrival for the second drop, BD125 once again set Bomber 139 up with a left-hand circuit to the target for a drop flown by BD125 in the Chase‑Position, but this time without a Show-Me profile. The left-hand patterns clear of the smoke provided Bomber 139’s captain with a clear view of the drop zone and exit during the approach to the target. 

When Bomber 139 arrived at the fire ground on the third (accident) flight, BD123 had replaced BD125. In preparation for the drop, BD123 briefed Bomber 139 that the drop zone was clear of ground crew, there were no hazards, and that it was a downhill drop. The captain declined BD123’s offer of a Show-Me profile, consistent with the second flight. However, the captain accepted a Lead profile right-hand circuit, which was inconsistent with the earlier flights.

The right‑hand pattern was chosen to keep them clear of smoke, as the plan was to tag and extend an existing line of retardant downhill into a depression towards the south with the fire burning to the east of the line. It also meant that the captain, in the left seat, was sitting on the outside of the circuit, which made it more difficult to view the drop zone on the downwind and base legs. 

Bomber 139 then followed BD123 on a ‘wide’ right-hand circuit pattern with a final approach and descent to the target down the right side of the smoke. The captain, who preferred a closer circuit pattern to maintain visual contact with the drop zone and exit during the approach, had trouble identifying the start of the drop (a road crossing) on the first approach. The captain’s difficulty identifying the road was likely due to the wide right-hand circuit pattern and resulted in the captain’s attention becoming focussed on sighting it for the start of the drop.

During the drop the captain detected that they were not on the correct line to keep the retardant drop clear of the burnt area. Therefore, they stopped the drop and conducted a go-around from the high ground with the fire in the depression burning across their path. The elevation of the terrain beneath at the time of that go-around was higher than the elevation of the terrain at the accident site, which was on the far side of the depression. Therefore, the accident ridgeline remained below the horizon for the crew on the first drop, potentially obscured by smoke from the fire in the depression and likely indistinguishable from the surrounding terrain due to the relatively consistent vegetation coverage.

The captain’s decline of the Show-Me profile, wide right-hand circuit, Birddog brief of a downhill drop and subsequent go-around from the high ground, meant that the captain did not expect, or detect, rising terrain in the exit path prior to attempting the second drop.  

The operator had adopted 150 ft as their standard target drop height. However, a low drop height and speed would improve the accuracy of a drop at a low coverage level. The crew was using a low coverage level [3] throughout the day and on the 2 previous flights that day, Bomber 139 recorded minimum drop height parameters of 46 ft and 69 ft with the engines at 60‍–‍70% N1. When the thrust levers were advanced at the end of each drop, there was an immediate positive response in N1, airspeed and vertical speed. The minimum parameters on the first (partial) drop of the accident flight were 78 ft and 124 kt at about 70% N1. Following the partial drop, when the captain advanced the thrust levers for the go-around there was once again an immediate positive response in N1, airspeed and vertical speed. 

The performance response from the aircraft on the previous drops, expectation of a downhill drop with no hazards and no personnel on the ground, likely contributed to the captain’s confidence that the aircraft could be safely flown and recovered from a low drop height and speed. The minimum parameters recorded on the accident drop were 57 ft and 106 kt. However, this time, the captain used the idle thrust setting of 30% N1 at the start of the drop to control the airspeed for the downhill run, which required a relatively high rate of descent that peaked at about 1,800 ft/min. At this setting, the engines required 7‍–‍8 seconds after the thrust levers were advanced before they would reach go-around thrust.

The captain was aware of the need to advance the thrust levers before the end of a drop if idle thrust was used. However, the accident drop was a partial load that only took 5 seconds and encompassed a downhill section with rising terrain in the exit that the captain was not aware of. Consequently, when the captain advanced the thrust levers about mid-way through the drop, there was a delay in the engines accelerating to go-around thrust which resulted in a low and decreasing aircraft energy state at the end of the drop and the collision with rising terrain in the exit path.

Energy management

A comparison of the drops conducted by Bomber 139 on the day of the accident indicated that N1 was a key parameter for a successful go-around and no performance problems were evident when N1 was in the region of 60‍–‍70% at the end of the drop. Variation in the N1 acceleration profile data at about 40% on the accident drop indicated that a slower engine acceleration occurred below this point, which was consistent with an idle thrust setting. However, the industry preference was for drops to be conducted downhill and consequently, the operator reported it was not practical to prescribe a minimum N1. Instead, they trained their pilots to advance the thrust levers during the drop if idle thrust was required at the start of the drop.

During the accident drop, Bomber 139’s captain advanced the thrust levers about mid-way through the drop, which provided insufficient time for the engines to respond. The ATSB found this would likely be the case for partial drops and drops conducted at higher coverage levels, due to the shorter drop time. If idle thrust was used in these scenarios, a safe go-around would depend on the LAT having sufficient energy, in terms of height and airspeed, to maintain terrain clearance until the engines accelerated to go-around thrust.  

When considering the drop height parameter, the ATSB found that the US Interagency Airtanker Board (IAB) had investigated and determined that 150 ft was a LAT minimum safe drop height for their airtanker evaluation process. This was the reference drop height in the US airtanker and aerial supervision standards for a LAT.

Minimum drop heights were developed to reduce the forward momentum of the retardant cloud such that it would rain down vertically over the vegetation to prevent shadowing and minimise the hazard to ground personnel from falling debris. However, the retardant cloud can be subject to wind‑induced drift during the drop, which makes higher drop heights unsuitable due to the loss in coverage and incentivised lower drop heights over light fuel loads to improve accuracy.

Given the above factors, and noting that Bomber 139 was conducting drops significantly below 150 ft, there was no evident justification for changing the use of 150 ft as a LAT minimum drop height standard. Nevertheless, any increase in drop height would improve the aircraft energy state and therefore the safety of the flight crew.

The operator used a drop speed of 1.25 Vs for their drop configuration, which was based on the IAB minimum drop speed for their testing. The drop configuration was based on the flap setting and weight of the aircraft at the start of each drop. If a series of partial drops were conducted, then the flight crew would lower the target drop speed between each drop in accordance with the table of aircraft weights and speeds on their checklist.

While the IAB drop testing permitted drop speeds between 1.25 Vs and Va, the US Forest Service (USFS) explained that higher drop speeds can shear the retardant column at the tank opening and reduce the coverage level. Therefore, while an increase in the drop speed would improve the energy state of the aircraft, it would also require testing to ensure the required coverage level for effective fire suppression was maintained.

While the target drop speed was 1.25 Vs and the operator had published a lower deviation limit of ‑5 kt, the recorded data from the 4 drops on the day of the accident revealed that it was only on the second drop that Bomber 139 maintained the target drop speed within the tolerance. During the first drop, the airspeed decayed about 25 kt below the estimated target drop speed before recovering to within the tolerance at the end of the drop. On the third and fourth drops, the airspeed decayed about 9 kt below the target drop speed during the drops. 

The retardant drop is a manually flown contour flight manoeuvre in potentially rough air and over undulating terrain across a drop zone that has not been formally surveyed for obstacle clearance and gradient. Therefore, airspeed and height deviations could be expected. If a performance problem developed before the entire load was released, then the crew had the option to conduct an emergency dump to reduce the weight and increase the airspeed safety margin. However, if the drop speed was recalculated and lowered between partial drops, the safety margin for the go‑around was also reduced.

During the investigation, the operator reported to the ATSB that they believed the recalculation and lowering of the target drop speed was a ‘significant contributing factor to the accident.’ It was also noted by Boeing, from their analysis of the recorded data, that had Bomber 139 avoided the ridgeline, there was nothing to preclude the aircraft climbing away from the ground. 

The ATSB’s analysis of energy-height revealed Bomber 139 would have gained about 39‍–‍90 ft if the recalculated target drop speed lower tolerance of 113‍–‍118 kt was maintained during the final run. As such, even considering the low operating height of the final drop, if the aircraft had been flown at or above the lower deviation limit (113 kt) of the reduced target speed of 118 kt, the ground collision would probably have been avoided.

The energy analysis also identified that when the accident end-of-drop speed was compared with the previous drop, the recalculation of the drop speed was a greater proportion of the energy‑height loss (65%) than flying slower than the recalculated target speed (35%). The full load target drop speed would have provided them with +167 ft of energy-height to their target accident drop speed and +257 ft of energy height to their accident drop speed. Therefore, while the ATSB concluded that lowering of target drop speed between partial drops was not a contributing factor to the accident, it significantly reduced the terrain clearance energy‑height safety margin. 

Crew resource management

Analysis of the cockpit voice recorder (CVR) for the accident flight revealed the Bomber 139 captain and co-pilot were working cooperatively with a division of duties during the set-up for the first drop. Between the first and second drop, they were again working cooperatively to track the position of BD123, update their checks and set their aircraft up for the second drop. However, the CVR also revealed that the co-pilot was silent in the cockpit for about the last 70 seconds of the flight, during which Bomber 139 was flown below the operator’s standard target drop height and speed deviation limits published in their flight crew operations manual (FCOM) Retardant drop procedure and with the engines at idle. In accordance with the FCOM procedure, deviating outside of the limits of either of these parameters required a ‘call-out’ from the pilot monitoring, which was the co-pilot.

The co-pilot had been employed by the operator for less than a year and was aware of the 150 ft standard for drop heights, although they also expressed a belief that the accident drop height was about 100 ft. The co-pilot also believed there was no minimum drop height and indicated that they had interpreted the 150 ft standard as a maximum drop height to mitigate the potential adverse effects of the wind on the fall of the retardant. On the day of the accident, all 4 drops were conducted significantly below 150 ft, which likely reinforced the co‑pilot’s interpretation of 150 ft as a maximum height and was consistent with the idea that the wind could reduce the accuracy and coverage level at higher drop heights. 

The operator’s FCOM pre-drop checklist included setting the radio altimeter decision height bug to the target drop height. The 150 ft drop height was assumed knowledge in the operator’s organisation as they had adopted the IAB recommended minimum drop heights, which complied with their USFS contract requirements. However, the setting of the decision height bug to the drop height was not included in their quick reference handbook normal checklist and was not practiced by them.

Despite the published FCOM requirement, the operator believed that setting the decision height bug at their target drop height would result in a distraction if the decision height light cycled on and off over undulating terrain. Therefore, as the captain had not briefed a target drop height and the co-pilot had interpreted the 150 ft as a maximum drop height, the co-pilot did not have a minimum reference drop height to monitor against the deviation limits in the FCOM. This probably led to the co‑pilot’s silence about the low height during the accident drop.

The co-pilot reported the airspeed indicator was one of their primary flight instruments they were scanning during the drop, and that they believed it was indicating their target drop speed of 118 kt throughout the run. Recorded data indicated the airspeed went below, and remained below, the lower deviation limit about 2 seconds before the start of the drop after fluctuating intermittently above and below the lower limit during the approach. However, the co-pilot did not detect and announce the low and decreasing airspeed. 

The second approach was flown through the smoke and the CVR indicated the captain was having trouble identifying the start of the drop. As the co-pilot was also looking for the target and observed BD123 deploy smoke over the target, they were potentially distracted from the airspeed indicator during the approach.

Within 2‍–‍3 seconds of the start of the drop, the captain started to advance the thrust levers at which stage the co-pilot was anticipating the flap retract call for the go-around. The ATSB was unable to determine why the co-pilot did not recognise the low and decaying airspeed. However, it is possible that the short duration of the drop might have resulted in the co-pilot shifting their attention to the vertical speed indicator, altimeter and flap lever in preparation for the go-around once the captain started to advance the thrust levers. 

The operator reported their crew resource management (CRM) philosophy was to brief first and then only announce deviations as this minimised distractions during the retardant drop. However, this practice did not employ all the resources at their disposal to mitigate the risk of a low energy state developing. While they considered the decision height bug to be a distraction if it was set at the target drop height, they had not considered setting the bug at the lower deviation limit to assist the pilot monitoring duties. They also did not brief a target drop height if the captain planned to vary the drop height from their 150 ft standard, which resulted in the co-pilot being left ‘out‑of‑the‑loop’ in terms of their responsibility to monitor the target drop height. 

The pilot monitoring call-outs in the operator’s retardant drop procedure were reactive to exceeding the target drop height or drop speed deviation limits. They did not include proactive ‘approaching target height/speed’ or ‘on height/speed’ call-outs. Proactive announcements confirm to other crewmembers that the pilot monitoring is conducting their duties and alerts the pilot flying to the flight path and energy state of the aircraft.

The operator’s expectations of pilot monitoring duties were presented in their CRM training but their controlled flight into terrain prevention training package did not present their CRM and SOPs as prevention strategies. Retardant drops are conducted to unsurveyed areas, which may result in lower safety margins than an approach to, or departure from, an airport. Hence the application of CRM and SOPs to managing the energy state of the aircraft should be as important for a retardant drop as for a departure or an arrival.

According to the US Federal Aviation Administration’s (FAA) advisory circular on standard operating procedures and pilot monitoring duties for flightdeck crewmembers, effective CRM is founded on SOPs. Therefore, while the operator’s pilot monitoring announcements were reactive to a low energy state, the ATSB could not conclude that this CRM-related aspect contributed to the accident due to the conflict between the operator’s procedures and actual practice. However, the ATSB concluded that the reactive nature of the pilot monitoring duties in their retardant drop procedure increased the risk of their aircraft entering an unrecoverable energy state and therefore have assessed it as a safety issue for Coulson Aviation.

Standard operating procedures

During the ATSB’s 2020 C-130 LAT accident investigation, it was apparent that the crew had developed an elevated level of risk tolerance for hazardous weather conditions. While the loss of the crews’ lives and lack of CVR data limited the ATSB’s ability to explore their pre-flight and tactical decision-making, the ATSB did note that there were no ‘circuit-breakers’ in the system to mitigate the crews’ acceptance of a task in the prevailing conditions. The operator had not imposed weather limits or a pre-flight risk assessment and the tasking agency did not have a policy and procedure to manage the task rejections that unfolded throughout the day.

Following the Bomber 139 accident, the co-pilot reported to the ATSB that they did not believe there was a minimum drop height. The ATSB subsequently found that neither the operator nor the tasking agency had published a drop height. While the operator had adopted their USFS contract standard of 150 ft as their standard target drop height, this was assumed knowledge within the organisation and was not published as a SOP. Consequently, captains could exercise their own judgement to improve accuracy if there was no contractual limit imposed. Their contract for services with the National Aerial Firefighting Centre (NAFC) did not impose a minimum drop height but required the operator to comply with the Member SOPs for their nominated operational base. The WA LAT SOPs did not impose a drop height limit and therefore the Bomber 139 crew could conduct their drops below 150 ft without breaching their contract or published SOPs.   

The history of the development of airtanker drop heights revealed the primary aims were to minimise the risk to ground personnel from falling debris and prevent shadowing of the vegetation. The drop height safety requirement for ground personnel was published by the USFS in 2000 (Lovellette) and was later reiterated with a safety alert in 2019 following the fatal injury of a ground firefighter from a retardant drop in 2018. The WA LAT SOPs managed the safety risk to ground personnel by prohibiting drops before the drop zone was confirmed clear. This confirmation was communicated between the Birddog aircraft and Bomber 139 on the accident flight, which meant that the ground personnel safety requirement was not applicable to the drop height on this occasion.

The history of drop tests also revealed that wind speed and drop height affected the drop pattern and consequently the effectiveness of the drop. Greater wind speeds distorted the drop pattern at the lower coverage levels (Suter, 2005). When released from a sufficient height to stop the forward momentum of the retardant, Suter (2005) noted that the cloud was susceptible to wind erosion. Therefore, lower drop heights improve the drop pattern at low coverage levels. This knowledge may incentivise pilots to conduct lower drops, and operators and tasking agencies to accept this practice.

In this accident, the ATSB found that a low drop height was a contributory factor to the low energy state and collision with terrain during the exit from the drop. According to the FAA’s AC 120-71B (2017) Standard Operating Procedures and Pilot Monitoring Duties for Flight Deck Crewmembers, crew performance is founded on SOPs and pilot monitoring performance can be improved with the development of effective SOPs. Without a published SOP, the Bomber 139 captain could exercise drop height discretion and the co-pilot did not have a deviation parameter to monitor and call-out. Therefore, the ATSB has assessed the lack of a minimum drop height SOP as a safety issue for Coulson Aviation. 

Furthermore, the fireground is an area of operations where multiple operators are required to interact. This introduces the problem of non-standard procedures between operators and the potential for unforeseen risks to emerge. In this accident, the Birddog operator had published a minimum drop height SOP. However, neither the LAT operator, nor the relevant WA government tasking agencies— the WA Department of Fire and Emergency Services and Department of Biodiversity, Conservation and Attractions—had published a minimum drop height.

The tasking agency SOPs provide a mechanism for the integration and standardisation of their aerial firefighting operation and for them to take the lead on setting the quality and safety criteria they require from the aerial firefighting effort. Therefore, the lack of a minimum drop height SOP is considered to be a safety issue for the WA Department of Fire and Emergency Services and Department of Biodiversity, Conservation and Attractions.

National large air tanker standard operating procedures

Aviation is a federally regulated activity but there is no federal authority with the operational responsibility for aerial firefighting. In 2003, the National Aerial Firefighting Centre (NAFC) was formed by the Australasian Fire and Emergency Services Authorities Council with the support of federal, state and territory governments. Their state and territory Members agreed that a single contract for each operator would facilitate the movement of assets between jurisdictions. However, the operating procedures that applied in each jurisdiction remained each Member’s responsibility.

While this may not increase risk for aerial firefighting assets that remain within a single jurisdiction throughout the season, LATs have been identified as having the capability to operate in more than one jurisdiction on the same day. Furthermore, the 2020 Royal Commission into National Natural Disaster Arrangements identified LATs as assets that should be tasked according to ‘greatest need’ on a national rather than jurisdictional basis.

The LATs and their aerial supervision do not operate in isolation and must be integrated into each jurisdiction where they operate. Therefore, any inconsistencies or omissions between tasking agencies’ SOPs will increase the complexity of risk management activities as it will be more difficult to assess and manage the associated risks. On the day of the Bomber 139 accident, the LAT and Birddog aircraft were conducting tactical flight profiles at the fireground, which were consistent with the Birddog operator and US National Wildfire Coordinating Group SOPs, but were not reflected in the WA LAT SOPs. Bomber 139 was also conducting drops without a prescribed minimum safe drop height or emergency overrun procedures under their contract for services.

The WA LAT target identification procedures (tactical flight profiles) were consistent with the 2018 NSW LAT procedures published at the time of the C-130 accident. However, NSW had published a LAT drop height of 150 ft but WA had not published a drop height, and where WA had prescribed the drop exits should be downhill, NSW had not made this explicit. Consequently, a different standard or no standard existed in each agencies’ document, which indicated an inconsistent approach to the development and implementation of LAT SOPs. 

In contrast, the US standards for LAT and aerial supervision operations reviewed by the ATSB consistently reported a 'minimum drop height of 150 ft’ for LATs. The USFS reported that the move in 1976 towards the development of national standards was accepted as a better way of operating. They captured the lessons learned from accidents and facilitated the movement of airtankers across state borders to operate at multiple firegrounds on the same day.

The ATSB noted that the US aerial firefighting standards were a collaborative effort with multiple signatories on each document. One of the challenges in compiling a standard is understanding what should, and should not, be included. The current process of each NAFC Member developing their own standards hampers an inter-agency collaborative process with stakeholder consultation to develop a set of best-practice SOPs. 

The differences in the NAFC Members’ LAT SOPs, noted by the ATSB, were not based upon a complete review, only the circumstances surrounding the C-130 and B737 accidents. A comprehensive review of all Member LAT SOPs, while beyond the scope of this ATSB investigation, would likely find further inconsistencies and omissions. This work is best conducted by the key stakeholders, who are the NAFC Members, with input from the LAT operators, aerial supervision operators and air attack supervisors.

At the time of the C-130 and B737 accidents, there was no agency responsible for producing national aerial firefighting SOPs. Following the Royal Commission into the 2019/2020 bushfire season, a national Aviation Safety Group (ASG) was formed from the NAFC Members and a priority list developed, which included standardised doctrine. However, the ASG’s list of priorities did not explicitly identify LAT SOPs, despite the 2 recent LAT accidents and the intent for them to be deployed on a national ‘greatest need’ basis.

The ATSB considers the absence of national LAT SOPs a safety issue and considers the Australasian Fire and Emergency Services Authorities Council collaboration model through the NAFC Strategic Committee the most appropriate pathway to address this deficiency. The safety issue has been raised with the understanding that the LAT SOPs will either incorporate the LAT aerial supervision procedures or that the LAT aerial supervision procedures will be developed and implemented concurrently, noting the ASG has already prioritised a national LAT Lead Plane pilot training standard.    

Findings

From the evidence available, the following findings are made with respect to the controlled flight into terrain involving Coulson Aviation Boeing 737-3H4, registered N619SW, at Fitzgerald River National Park, Western Australia on 6 February 2023. 

Contributing factors

• During the retardant drop downhill, the aircraft descended significantly below the operator’s standard target drop height and airspeed and entered a high rate of descent with the engines at idle. While the engines were starting to accelerate at completion of the drop, the airspeed and thrust were insufficient to climb above a ridgeline in the exit path, which resulted in the collision with terrain.
• Prior to the retardant drop, the aircraft captain (pilot flying) did not detect there was rising terrain in the exit from the drop, which likely contributed to the captain allowing the aircraft to enter a low energy state during the drop.
• After arrival at the fireground, the aircraft captain (pilot flying) declined a ‘Show-Me’ run and was briefed by the Birddog pilot that it would be a downhill drop. Bomber 139 then conducted a go-around from the high ground after the first drop and was led to the target through the smoke on the second drop. These factors likely contributed to the captain not expecting or detecting the rising terrain in the exit path.
• The co-pilot (pilot monitoring) did not identify and announce any deviations during the retardant drop, which could have alerted the aircraft captain (pilot flying) to the low-energy state of the aircraft when it descended below the target drop height with the engines at idle.
• The flight crew did not brief a target retardant drop height and, contrary to published standard operating procedures, did not set it as a decision height reference on the radio altimeter. Subsequently, the co-pilot (pilot monitoring), who did not believe there was a minimum drop height, did not alert the aircraft captain (pilot flying) to the low-energy state of the aircraft.
• Coulson Aviation and the relevant Western Australian Government Departments had not published a minimum retardant drop height in their respective operating procedures for large air tankers. Consequently, the co-pilot (pilot monitoring), who did not believe there was a minimum drop height, did not alert the aircraft captain (pilot flying) to a drop height deviation prior to the collision. (Safety issue)

Other factors that increased risk

• The Coulson Aviation practice of recalculating the target retardant drop speed after a partial drop reduced the post-drop stall speed and energy‑height safety margins. (Safety issue)
• The Coulson Aviation crew resource management practice of limiting the pilot monitoring (PM) announcements to deviations outside the target retardant drop parameter tolerances increased the risk of the aircraft entering an unrecoverable state before the PM would alert the pilot flying. (Safety issue)
• Australian states and territories that engage in Large Air Tanker (LAT) operations have developed their own separate standard operating procedures (SOPs) for LATs and aerial supervision assets. This can result in safety requirements being omitted or misunderstood by the different tasking agencies, such as a minimum drop height, resulting in inconsistencies in the development and application of LAT SOPs. (Safety issue)

Safety issues and actions

Drop speed recalculation

Safety issue number: AO-2023-008-SI-01

Safety issue description: The Coulson Aviation practice of recalculating the target retardant drop airspeed after a partial drop reduced the post-drop stall speed and energy‑height safety margins.

Inadequate procedural requirements

Safety issue number: AO-2023-008-SI-02

Safety issue description: Coulson Aviation and the relevant Western Australian Government Departments had not published a minimum retardant drop height in their respective operating procedures for large airtankers. Consequently, the co-pilot (pilot monitoring), who did not believe there was a minimum drop height, did not alert the aircraft captain (pilot flying) to a drop height deviation prior to the collision.

Inadequate procedural requirements

Safety issue number: AO-2023-008-SI-03

Safety issue description: Coulson Aviation and the relevant Western Australian Government Departments had not published a minimum retardant drop height in their respective operating procedures for large airtankers. Consequently, the co-pilot (pilot monitoring), who did not believe there was a minimum drop height, did not alert the aircraft captain (pilot flying) to a drop height deviation prior to the collision.

Crew resource management practice

Safety issue number: AO-2023-008-SI-04

Safety issue description: The Coulson Aviation crew resource management practice of limiting the pilot monitoring (PM) announcements to deviations outside the target retardant drop parameter tolerances increased the risk of the aircraft entering an unrecoverable state before the PM would alert the pilot flying.

Safety recommendation description: The Australian Transport Safety Bureau recommends that Coulson Aviation takes safety action to address their crew resource management procedures for retardant drops to reduce the risk of the aircraft entering an unrecoverable state before the pilot monitoring alerts the pilot flying.

National Large Air Tanker standard operating procedures

Safety issue number: AO-2023-008-SI-05

Safety issue description: Australian states and territories that engage in Large Air Tanker (LAT) operations have developed their own separate standard operating procedures (SOPs) for LATs and aerial supervision assets. This can result in safety requirements being omitted or misunderstood by the different tasking agencies, such as a minimum drop height, resulting in inconsistencies in the development and application of LAT SOPs.

Safety action not associated with an identified safety issue

Additional safety action by Coulson Aviation

3 May 2023

Coulson Aviation increased the safety margins for large air tanker drops in Australia from 150 ft and 1.25 stall speed to 200 ft and 1.35 stall speed. This was promulgated on Operations Bulletin 2023-01 on 8 February 2023 and their B737 Fireliner normal checklist was amended accordingly.

October 2023

In October 2023, Coulson Aviation provided the ATSB with a copy of their master corrective action plan (CAP) document, which they had developed in response to the Bomber 139 accident. This comprised flight activities, ground activities, Operations Bulletins and Flight Operations Manual amendments under the following topics:

CAP 1: raise minimum drop speed and height

CAP 2: recalibration of airspeed during partial load drops

CAP 3: pre-drop planning and familiarization

CAP 4: pilot situational awareness

CAP 5: flight training exercises

CAP 6: exit flight path

CAP 7: communication and coordination with Birddog

CAP 8: adherence to standard operating procedures

CAP 9: crew resource management

CAP 10: human factors

CAP 11: continuous training and proficiency assessment.

Additional safety action by the Birddog operator

The Birddog operator reported that as part of their continuous improvement they removed their Lead Plane procedures from an annex in their Company Operations Manual and introduced a Lead Plane Standard Operating Procedures manual with expanded procedures in October 2023.

Additional safety action by the Western Australian Department of Fire and Emergency Services and Department of Biodiversity, Conservation and Attractions

In November 2023, the Department of Fire and Emergency Services, and the Department of Biodiversity, Conservation and Attractions released the Western Australian Aerial Fire Suppression Procedures 2023‍–‍2024. The draft version of this document was reviewed by their contracted operators and included a chapter for Large Air Tanker (LAT) Operating Procedures. The document included standard drop heights for all the aircraft types employed, procedures for task rejection, standdown and reinstatement of aviation resources after a safety incident, and a section on tactical flight profiles with the addition of the Lead and Chase-Position profiles.

Additional safety action by the Western Australian Department of Biodiversity, Conservation and Attractions

The Western Australian Department of Biodiversity, Conservation and Attractions advised that published drop heights for 2024‍–‍25 will be reviewed with the intent of prescribing a ‘minimum’ safe drop height, noting the actual drop height to safely achieve the objective is the pilot‑in‑command’s decision.

Glossary

Sources and submissions

Sources of information

The sources of information during the investigation included: 

• Australasian Fire and Emergency Service Authorities Council
• Birddog 123 pilot
• Bomber 139 captain and co-pilot
• Coulson Aviation
• General Electric Aerospace
• National Aerial Firefighting Centre
• New South Wales Rural Fire Service
• Queensland Fire and Emergency Services
• The Boeing Company
• United States Forest Service
• United States National Transportation Safety Board
• Victorian Department of Energy, Environment and Climate Action
• Western Australian Department of Fire and Emergency Services.

References

Australian Transport Safety Bureau (2022), Collision with terrain involving Lockheed EC130Q, N134CG, 50 km north-east of Cooma-Snowy Mountains Airport (near Peak View), New South Wales, on 23 January 2020 (AO-2020-007). Retrieved from https://www.atsb.gov.au

Federal Aviation Administration (2004), Crew Resource Management Training (AC 120-51E). Retrieved from https://www.faa.gov

Federal Aviation Administration (2017), Standard Operating Procedures and Pilot Monitoring Duties for Flight Deck Crewmembers (AC 120-71B). Retrieved from https://www.faa.gov

Federal Aviation Administration (2018), Flight Test Guide for Certification of Transport Category Airplanes (AC 25-7D). Retrieved from https://www.faa.gov

Federal Aviation Administration (2021), Airplane Flying Handbook Chapter 4: Energy Management: Mastering Altitude and Airspeed Control (FAA-H-8083-3C). Retrieved from https://www.faa.gov

Interagency Airtanker Board (2013), Procedures and Criteria for the Interagency Airtanker Board (IAB). Retrieved from https://www.nwcg.gov

Lovellette G (2000), Safe Drop Height for Fixed-Wing Airtankers. Retrieved from https://www.fs.usda.gov

National Aerial Firefighting Centre (2021), National aerial firefighting strategy 2021-26. Retrieved from https://www.nafc.org.au

National Transportation Safety Board (1994), Safety study: A review of flightcrew-involved, major accidents of U.S. air carriers, 1978 through 1990 (NTSB/SS-94/01). Retrieved from https://www.ntsb.gov

National Transportation Safety Board (2014), Low altitude operation/event involving Lockheed P2V-7, N14447, Modena, Utah, 3 June 2012 (WPR12GA243). Retrieved from https://www.ntsb.gov

National Wildfire Coordinating Group (2017), Interagency Aerial Supervision Guide (PMS 505). Retrieved from https://gacc.nifc.gov

New South Wales Rural Fire Service (2018), Operating guidelines for air tankers operations. RFS, Sydney.

Suter A (2005), Aerial Delivery Systems User Information: Wind speed and drop height. Retrieved from https://www.fs.usda.gov

Transportation Safety Board of Canada (2004), Collision with Terrain, Air Spray (1967) Ltd., Lockheed L-188 Electra C-GFQA, Cranbrook, British Columbia 2.5 nm south, 16 July 2003 (A03P0194). Retrieved from https://www.tsb.gc.ca 

Transportation Safety Board of Canada (2004), Collision with Terrain, Conair Group Inc., Convair 580 C–FKFY, Lytton, British Columbia, 9 nm SE, 31 July 2010 (A10P0244). Retrieved from https://www.tsb.gc.ca

United States Department of Agriculture Forest Service (2019), Standards for Airtanker Operations. Retrieved from https://www.fs.usda.gov

United States Department of Agriculture Forest Service (2019), Interagency aviation safety alert: Retardant safe drop height (IASA 19-02). Retrieved from https://www.fs.usda.gov

Western Australian Department of Fire and Emergency Services and Department of Biodiversity, Conservation and Attractions (n.d.), Large air tankers operational procedures 2021-22, v3.3.

Submissions

Under section 26 of the Transport Safety Investigation Act 2003, the ATSB may provide a draft report, on a confidential basis, to any person whom the ATSB considers appropriate. That section allows a person receiving a draft report to make submissions to the ATSB about the draft report. 

A draft of this report was provided to the following directly involved parties:

• Australasian Fire and Emergency Service Authorities Council
• Birddog 123 pilot
• Bomber 139 captain and co-pilot
• Civil Aviation Safety Authority
• Coulson Aviation
• National Aerial Firefighting Centre
• New South Wales Rural Fire Service
• United States National Transportation Safety Board
• Western Australian Department of Fire and Emergency Services
• Western Australian Department of Biodiversity, Conservation and Attractions.

Submissions were received from:

• Australasian Fire and Emergency Services Authorities Council
• Birddog 123 pilot
• National Aerial Firefighting Centre
• Western Australian Department of Biodiversity, Conservation and Attractions.

The submissions were reviewed and, where considered appropriate, the text of the report was amended accordingly.

Appendices

Appendix A – SIGMET thunderstorm activity

Figure 18: Lightning strikes (red) at 1600

Appendix B – Flight data recorder plot

Figure 19: FDR plot of last 5 minutes of the accident flight with the last 2 drops

Purpose of safety investigations The objective of a safety investigation is to enhance transport safety. This is done through:  identifying safety issues and facilitating safety action to address those issues providing information about occurrences and their associated safety factors to facilitate learning within the transport industry. It is not a function of the ATSB to apportion blame or provide a means for determining liability. At the same time, an investigation report must include factual material of sufficient weight to support the analysis and findings. At all times the ATSB endeavours to balance the use of material that could imply adverse comment with the need to properly explain what happened, and why, in a fair and unbiased manner. The ATSB does not investigate for the purpose of taking administrative, regulatory or criminal action. Terminology An explanation of terminology used in ATSB investigation reports is available here. This includes terms such as occurrence, contributing factor, other factor that increased risk, and safety issue. Publishing information Released in accordance with section 25 of the Transport Safety Investigation Act 2003 Published by: Australian Transport Safety Bureau © Commonwealth of Australia 2024 Ownership of intellectual property rights in this publication Unless otherwise noted, copyright (and any other intellectual property rights, if any) in this report publication is owned by the Commonwealth of Australia. Creative Commons licence With the exception of the Commonwealth Coat of Arms, ATSB logo, and photos and graphics in which a third party holds copyright, this report is licensed under a Creative Commons Attribution 4.0 International licence. The CC BY 4.0 licence enables you to distribute, remix, adapt, and build upon our material in any medium or format, so long as attribution is given to the Australian Transport Safety Bureau.  Copyright in material obtained from other agencies, private individuals or organisations, belongs to those agencies, individuals or organisations. Where you wish to use their material, you will need to contact them directly.