Executive summary

What happened

On 17 September 2021, intermodal freight train 8796, operated by Aurizon, was en route from the Stuart terminal (near Townsville) to Brisbane, Queensland. One of the wagons was loaded with a flat rack carrying a heavy road vehicle tipping trailer that was out of gauge (over the permissible height for the route).

During the journey, the trailer’s hydraulic lifting post collided with the overhead structure of Alexandra Bridge in Rockhampton, causing damage to the bridge and trailer. There were no injuries and no damage to the train.

What the ATSB found

The Stuart terminal was an intermodal freight facility that was cooperatively operated between Linfox and Aurizon. Linfox was responsible for the movement of road freight to and from the terminal, and owned the wagons. Aurizon conducted the loading of rail wagons at the terminal and was responsible for rail operations and rail safety, including the safety and integrity of loads.

The tipping trailer was a Linfox asset rather than customer freight. As such, it was not managed through Linfox’s normal process, which required any non-standard loads to be referred to management for approval. The Linfox capacity officer who entered the load into the freight management system was not advised that the trailer was a tipping trailer and, as a result, it was entered into the system with generic, incorrect dimensions and with no indication of it being potentially out-of-gauge. In addition, unlike at other terminals where Linfox was directly involved in loading trains, Linfox did not have a process for internal, non-standard freight movements at the Stuart terminal to be reviewed and approved for conformance to the permissible loading profile for transport via rail.

The tipping trailer was secured onto a flat rack by Linfox staff, and then lifted onto the wagon by an Aurizon heavy lift operator. The heavy lift operator misjudged the height of the freight when loading the wagon, likely associated with the heavy lift operator not having any nearby reference objects and having a high level of expectancy that the load would not be out of gauge (given that non-standard loads were rarely encountered and there had been no prior indication that this load would be out of gauge).

Later, while conducting a walking inspection of the completed consist, the heavy lift operator did not detect the over-height load, probably due to expectancy and being focussed on checking that the loads were secure rather than their height. Similarly, the out-of-gauge load was not detected during 2 subsequent roll-by inspections (one at Stuart and one at Merinda).

At the Stuart terminal, Aurizon did not routinely apply a process to verify that the dimensions of non-standard loads were within the permissible loading profile. Aurizon had previously had a tool for measuring the height of loads at the terminal. However, at the time of the occurrence, it did not have any measuring equipment available to identify freight loads that were outside the permissible loading profile for transport via rail.

What has been done as a result

Aurizon reported that it carried out a range of safety actions as a result of this occurrence, including:

• installing a passive over-height warning device (jangle bar) as temporary corrective action, and initiating assessment in relation to automated controls
• updating its freight management system and booking system and implementing procedures for the identification of non-standard freight
• revising competency assessments for forklift and reach stacker operators
• reviewing its loading and securing training and standards.

Linfox advised that it would train forklift and operations staff in Aurizon loading and securing and produce a quick reference guide for loaders, and it arranged for non-standard items to be referred to and approved by the capacity and rail network manager.

Safety message

In the absence of measuring equipment, or nearby objects of a known and relevant height, it is difficult to accurately estimate the dimensions of loaded freight, especially when judging the height of tall freight from ground level. Tools to alleviate this limitation will be significantly more accurate and come with minimal cost and should be available to personnel involved in loading non-standard loads.

The occurrence

Freight arrangements

Intermodal freight train 8796, operated by Aurizon, was a regular service between Stuart (near Townsville) to Acacia Ridge (in Brisbane), Queensland on the North Coast Line (Figure 1). At the Stuart terminal, Aurizon was responsible for the loading of freight onto wagons and Linfox was responsible for the movement of road freight to and from the terminal. Queensland Rail (QR) was the rail infrastructure manager for most of the North Coast Line.

Figure 1: Shipping route

Source: Queensland Rail, annotated by ATSB

On 15 September 2021, the Linfox warehouse supervisor at the Stuart terminal made a request (via phone) to a Linfox capacity officer at Acacia Ridge for a (heavy road vehicle) trailer to be transported via rail from Stuart to Brisbane.

The trailer was a tipping skeletal trailer (Figure 2) that was designed to be towed by a prime mover for transporting shipping containers via road, with the capability of raising one end to unload its contents. It had an incorporated hydraulic lifting post to facilitate the tipping process. The height of the tipping trailer, including the hydraulic lifting post, was 3.070 m. The trailer was a Linfox asset based at the Stuart terminal and it was being transported to be used by Linfox at the Acacia Ridge terminal.

Figure 2: Tipping skeletal trailer dimensions

All measurements are in mm. Shown attached to a prime mover and loaded with a 20-foot container.

Source: Aurizon

The Linfox capacity officer entered the trailer into Aurizon’s A2B freight management system for rail transport. It was entered in the system as a flat rack unit,[1] with a length of 12.192 m and height of 2.743 m, which was the default height in the A2B system. It was classified as ‘general domestic’ freight.

The flat rack type used for the trailer was common to the freight operation, and it had a height 0.260 m. The total height above the wagon floor of the tipping trailer’s hydraulic lifting post on a flat rack was therefore 3.330 m. This was 0.590 m above the maximum permissible height of a load on the QR network.

In their conversation with the capacity officer, the Linfox supervisor did not describe the trailer as a tipping trailer and did not provide (and was not asked to provide) its dimensions. The capacity officer later stated that they were aware that the tipping version of the trailer was probably out of gauge,[2] but they had not been advised that this particular trailer was the tipping version. Accordingly, they thought they were arranging to transport a normal (non-tipping) trailer. They did not change the default dimensions of the load in the freight management system and the consignment note did not include an out-of-gauge warning or indication.

Loading and departure

At about 1150 local time on 16 September, Linfox personnel reversed the trailer with a prime mover onto a flat rack at the Stuart terminal and secured it for transport. A closed-circuit television (CCTV) recording showed that several Linfox personnel were involved in the loading process (Figure 3). The dimensions of the loaded flat rack were not measured during this process.

Figure 3: Loading and securing the tipping trailer on to a flat rack at Stuart Yard

Source: Aurizon

After the trailer was secured onto the flat rack, the Linfox supervisor notified an Aurizon heavy lift (forklift) operator that the trailer was ready to be loaded. At 1347, the heavy lift operator lifted the flat rack with the tipping trailer by forklift onto a flat rail wagon (number BCZY 46459). The load’s dimensions were not verified, and the heavy lift operator did not identify that the load was potentially out of gauge.

The wagons were then shunted onto the train. From 1740 to 1805, the heavy lift operator completed a test and inspection of the completed rail consist, including the wagon with the loaded trailer, checking for container and load security. The inspection was from an area beside the track with eye height below the top of the trailer. They did not identify that the load was potentially out of gauge during this process.

At 1912, train 8796 departed the Stuart terminal as a driver-only operation with a 2800 class locomotive. During the departure of the train, a roll-by inspection[3] was conducted by Aurizon personnel, which did not identify the over-height wagon.

At 2240, additional wagons were added at Merinda. A roll-by inspection at that location did not identify any problems.

A driver crew change occurred at Mackay and the train departed Mackay at 0225 on 17 September. On this section of the journey it had 26 container wagons and was 526 m long with a total weight of 1,164 t, and the consist was predominantly containerised freight and empty flat-bed wagons. The wagon carrying the trailer was the 23rd wagon.

Collision with Alexandra Bridge

At about 0723 on 17 September, the train entered the Alexandra Bridge structure, just outside of Rockhampton Yard, travelling at approximately 25 km/h. The Alexandra Bridge is a concrete pillar and steel truss bridge over a river located in the centre of Rockhampton, serviced by a single rail line and a pedestrian pathway.

Approximately 2 minutes later, the hydraulic lifting post on the trailer impacted the entry crossbeam of the bridge as its wagon passed onto the bridge (Figure 4). A CCTV recording showed the impact with the crossbeam at the northern end of the bridge causing damage to the girder and buckling of bridge braces. Additionally, the bracing rails on the trailer bent under the impact, forcing the hydraulic lifting post to bend backwards. This gave the trailer clearance under the bridge structure for the rest of its journey across the bridge.

Train 8796 continued to Rockhampton, with the driver unaware that the collision had occurred. At 0734 the train entered Rockhampton Yard.

Figure 4: Trailer collision with bridge

Source: Rockhampton Regional Council

Identification of the damaged trailer

The damaged tipping trailer was observed at Rockhampton Yard (Figure 5). The wagon was removed from the consist and QR was notified of the damage at about 1020.

At 1230, the rail traffic crew of Aurizon train 8798 reported damage to the overhead structure on Alexandra Bridge. QR track workers conducted an inspection of the rail corridor between Stuart and Rockhampton, and they determined that only Alexandra Bridge had been damaged (Figure 6). Rockhampton Regional Council confirmed the impact of the bridge through the review of a CCTV recording.

The bridge was deemed fit for the return of services by QR engineering personnel and reopened to rail traffic at 1530 on 17 September with a 10 km/h speed restriction, pending further assessment.

Figure 5: Damaged skeleton tipping trailer, Rockhampton Yard

Source: Aurizon (upper), Linfox (lower), annotated by the ATSB

Figure 6: Alexandra Bridge girder damage

Source: Queensland Rail, annotated by the ATSB

Context

Track and network information

The North Coast Line consists of 1,680 km of railway between Cairns and Brisbane. Queensland Rail (QR) was the rail infrastructure manager for about 1,567 km, including the section from Townsville to Rockhampton.

The maximum permissible load height on the on the North Coast Line from Townsville to Brisbane was 3.820 m above the rail, or 2.740 m above the wagon floor with a standard 1.080-m high wagon. Therefore, with a total height of 4.410 m, the wagon with the flat rack and tipping trailer was 0.590 m above the maximum permissible load height.  

Following the accident, QR measured the clearance between rail and the structure of Alexandra Bridge to be 4.312 m, which was 12 mm more than the bridge’s design clearance. Figure 7 compares the outline of a standard shipping container with the height of the Alexandra Bridge and the height of the tipping trailer’s hydraulic lifting post.

Figure 7: North Coast Line standard container loading outline[4] with overlays of tipping trailer and Alexandra Bridge beam heights

All measurements are in mm. The added Alexandra Bridge and tipping trailer height overlays are for illustrative purposes only.

Source: Aurizon, modified by ATSB

The track south of Rockhampton was electrified with sections of overhead contact wires as low as 4.349 m above the track level.

Stuart terminal

The Stuart terminal was an intermodal freight facility about 11 km south of Townsville that was cooperatively operated between Linfox and Aurizon. Linfox was responsible for the movement of road freight to and from the terminal and owned the wagons, with a ‘hook and pull’ (wagon hauling) agreement in place with Aurizon. Aurizon conducted the loading of rail wagons and was responsible for rail operations and rail safety, including the safety and integrity of loads. This arrangement meant that there was a combination of both Linfox and Aurizon personnel working at the terminal.

Most of the freight passing through Stuart Yard was in 20-foot or 40-foot containers, though occasionally non-containerised freight was also transported.

At the time of the occurrence, Aurizon was responsible for loading trains at Stuart and one other terminal on the North Coast Line. At other terminals (such as Acacia Ridge), Linfox was responsible for the loading.

Freight booking information

Aurizon and Linfox used a freight management system called A2B to coordinate the movement of items. 

Provisions were in place where freight known to be exceeding the standard outline could be transported. In those circumstances, when freight was known to be out of gauge, an authority to travel (ATT) authorisation was required to be obtained with specific restrictions (for example, only to travel through certain tracks or not to pass other rail traffic in certain areas). Even with an ATT in place, there was still maximum limitations due to various factors, such as tunnel sizes, overhead electrical equipment and bridges. Where an ATT was approved, the freight terminal would be advised directly, and a notation made on the consignment note applicable for the freight being transported, warning the loading operators of the oversized item.

Linfox reported that clients normally booked freight through its customer service team. Any non-standard freight bookings (or freight that did not fit into standard containers) would be referred to a manager for approval. In this case, however, the tipping trailer was an internal asset that was already in the yard at Stuart and the booking process bypassed this process. 

The Linfox warehouse supervisor at the Stuart terminal had only recently (within 3 months) commenced work at that terminal. They reported that they had arranged for the transport of tipping trailers by rail when working at another depot in another state, but had not previously arranged for transport of a tipping trailer from the Stuart terminal or on the North Coast Line.

Linfox advised that it had processes for checking the dimensions of freight at terminals where Linfox were involved in loading freight onto trains. However, since Linfox did not load freight onto rolling stock at the Stuart Terminal, there was no procedure in place for measurement of the dimensions of any non-standard freight received at the Stuart terminal prior to requesting Aurizon personnel to load the freight onto a train. The warehouse supervisor also advised that they were not aware of the specific dimensions of loads that were allowable on the North Coast Line. They were relying on the Aurizon personnel at the terminal to check the load and determine whether the load was suitable for rail travel. This was consistent with the process they were familiar with when working at other terminals.

Freight loading information

Aurizon’s Loading and Securing Standard provided instruction on the safe transport of different types of freight by rail or road throughout Australia. It noted that a rail infrastructure manager (RIM) was responsible for defining freight outlines and for the authorisation of out-of-gauge loads, and a rolling stock operator was responsible for ensuring freight complied with the RIM’s instructions. It also stated that, for rail transport, the rolling stock operator (in this case Aurizon) was responsible for ensuring:

• Loading that exceeds or infringes the permissible outlines applicable to the type of loading and the routes over which it will travel is not transported, unless it has a valid out of gauge certificate
• All loading that is suspected of being close to or outside the applicable outline is referred to the RIM for investigation

Aurizon’s Loading and Securing of Freight Manual emphasised that out-of-gauge loads were not allowed to travel on Aurizon trains without an ATT. It also stated that the workers responsible for loading and securing freight were responsible for (among other things) ensuring that freight was loaded so that its overall size was within the (specified) standard loading outlines. Aurizon provided such personnel with a checklist for loading and securing flat racks, which included the loading outline and maximum allowed dimensions.   

The Aurizon heavy lift (forklift) operator had worked in that role at the Stuart terminal for about 10 years, and they had undertaken the organisation’s online loading and securing training within the required time period. The heavy lift operator could not recall loading any out-of-gauge loads in the last 5 years. They stated they had knowledge of other (heavy road vehicle) trailers being transported on flat racks without any problems, and they also thought (but were not sure) that tipping trailers had been transported before. At the time they loaded this specific tipping trailer, they did not consider the height to be an issue. They had received no indication through the consignment note, their supervisor or the Linfox warehouse supervisor that the load was actually or potentially out of gauge. Other personnel who saw the loaded flat rack also indicated that they did not consider the height to be an issue.

Personnel reported that shipping containers were sometimes used as a general indication of the allowable freight envelope. On this occasion, although a 20-foot shipping container[5] was near the tipping trailer when it was loaded onto the flat rack by Linfox personnel, there was no container on the ground near the flat rack when it was lifted onto the wagon. The heavy lift operator also stated that there were no other containers on adjacent wagons at that time. There were containers loaded on the adjacent wagons when the completed rail consist was inspected at the Stuart terminal and at Merinda.

Aurizon advised the ATSB that pre-loaded flat racks that arrived at the terminal needed to pass an incoming inspection at the terminal’s entrance gate. However, it rarely received loaded flat racks via the entrance gate. Aurizon also advised that it was not a routine or consistent practice for the documented dimensions of non-standard freight to be physically verified, or for the train’s loading outline to be measured, during the loading process.

In addition, there were no tools or equipment provided for checking the dimensions of non-standard freight. Aurizon previously had a load height-measuring staff at Stuart to gauge a train’s height. This L-shaped staff or pole was held next to a wagon, and if the total load exceeded its height, that wagon would not be allowed to travel. This staff was a locally-implemented risk control; Aurizon’s safety management system did not specifically require it to be in place. After moving from previous facilities to the current Stuart terminal, the staff was no longer available to loading personnel.

The Stuart terminal did not have a viewing platform to assist with identifying out-of-gauge loads. There were also no other devices installed at the terminal to verify that loads were below the maximum permissible height or indicate over-height loads.

The ATSB has previously investigated a number of other occurrences involving loads out of gauge or not effectively secured. Some recent investigations include:

• On 15 June 2020, a wagon body became dislodged from the bogie during unloading at Coopers Plains, Queensland.[6] As part of the consist of train 2BW4, the wagon body later contacted and caused damage to platforms at Grafton, Coffs Harbour, Taree, Wingham and Dungong, New South Wales. The train had been inspected before departure, and again after damage to the Grafton platform was identified.
• On 18 August 2018, a collapsible rear end wall of a flat rack raised up en route and contacted an overpass on approach to Cooroy, Queensland, pulling down 1.3 km of high voltage overhead line equipment.[7] The ATSB found that personnel at the Acacia Ridge terminal did not check the collapsible end walls of the flat racks were secured on arrival at the terminal and after the flat racks were loaded. In addition, Aurizon, did not have an effective system in place for ensuring personnel required to check the securing of unusual loads (such as empty flat racks) prior to departure had sufficient knowledge of their responsibilities, and had ready access to relevant procedures, guidance and checklists.
• On 16 January 2018, an incorrectly-secured container on freight train 2BM9 collided with station infrastructure at Maitland, New South Wales, causing damage to gutter retaining brackets.[8] The investigation found that the departing train inspection did not detect the incorrectly-secured container, and that the collision with infrastructure was also partially due to raised track height at the station relative to the documented design. 

Findings

From the evidence available, the following findings are made with respect to the collision with infrastructure involving freight train 8796 at Rockhampton, Queensland, on 17 September 2021.

Contributing factors

• The dimensions of the tipping trailer were not entered correctly into the freight management system when the shipment was being arranged, and there was no indications that the trailer was potentially out of gauge.
• Linfox did not have a process for internal, non-standard freight movements at the Stuart terminal to be reviewed and approved for conformance to the permissible loading profile for transport via rail.
• Judging the load’s height without a reference, the heavy lift operator did not detect the over-height load when loading the wagon or later when conducting a walking inspection of the consist from ground level. This allowed the over-height wagon to enter the rail corridor.
• At the Stuart terminal, Aurizon did not routinely apply a process to verify that the dimensions of non-standard loads were within the permissible loading profile, and it did not have measuring equipment available to identify freight loads that were outside the permissible loading profile for transport via rail. (Safety issue)

Safety issues and actions

Absence of measuring equipment

Safety issue number: RO-2021-010-SI-01

Safety issue description: Aurizon did not have measuring equipment available at its Stuart Yard to identify freight loads that were outside the permissible loading profile for transport via rail.

Safety action not associated with an identified safety issue

Additional safety action by Linfox

Linfox advised that it would train forklift and operations staff in Aurizon loading and securing and produce a quick reference guide for loaders.

Linfox also advised:

An awareness campaign has been completed so that all non-standard items are to be referred to and approved by the QLD Capacity and Rail Network Manager. This is regardless of its nature as customer freight or internal shipments.

Linfox are also considering the feasibility of making alterations to the WAVE software system (which has replaced the previous A2B system) to provide a systematic prompt for dimensions when non-standard containers types (flat racks or open top container) are lodged.

Glossary

ATT                  Authority to travel

CCTV               Closed-circuit television

Out of gauge     Exceeding the limits of the approved kinematic envelope, or outline, of the rail corridor to remain clear of obstructions

QR                   Queensland Rail

RIM                  Rail infrastructure manager

Sources and submissions

Sources of information

The sources of information during the investigation included:

• Aurizon
• Linfox
• Queensland Rail
• Rockhampton Regional Council
• the heavy lift operator, capacity officer and freight supervisor.

References

Wickens CD, Hollands JG, Banbury S and Parasuraman R (2013) Engineering psychology and human performance, 4th edition, Pearson Boston, MA.

Wickens CD & McCarley JS (2008) Applied attention theory, CRC Press, Boca Raton, FL.

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:

• Aurizon
• Linfox
• the Office of the National Rail Safety Regulator (ONRSR)
• Queensland Rail
• the heavy lift operator, capacity officer and freight supervisor.

Submissions were received from Aurizon, Queensland Rail and ONRSR and additional information was received from Linfox. The submissions were reviewed and, where considered appropriate, the text of the report was amended accordingly.

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 2022 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 Coat of Arms, ATSB logo, and photos and graphics in which a third party holds copyright, this publication is licensed under a Creative Commons Attribution 3.0 Australia licence. Creative Commons Attribution 3.0 Australia Licence is a standard form licence agreement that allows you to copy, distribute, transmit and adapt this publication provided that you attribute the work. The ATSB’s preference is that you attribute this publication (and any material sourced from it) using the following wording: Source: 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]     Flat racks are a type of shipping platform designed for oversized loads that do not fit inside a standard shipping container.

[2]     Out of gauge: exceeding the limits of the approved kinematic envelope, or outline, of the rail corridor to remain clear of obstructions.

[3]     Roll-by inspection: an inspection conducted by a qualified worker to identify issues with the rail consist, normally for wheel lockups or dragging equipment. It was conducted while standing at ground level adjacent to the train as it slowly ‘rolls’ past the worker under locomotive power.

[4]     The loading outline is a 2-dimensional cross-section of the maximum permissible envelope or shape of a rail vehicle at rest.

[5]     The height of a standard 20-foot shipping container is about 2.6 m.

[6]     ATSB investigation RO-2020-009, Wagon out of gauge on freight train 2BW4, Main North rail line, New South Wales, on 16 June 2020.

[7]     ATSB investigation RO-2018-011, Dewirement involving freight train YC77, Cooroy, Queensland, on 18 August 2018.

[8]     ATSB investigation RO-2018-003, Loading irregularity on train 2BM9, Maitland, NSW on 16 January 2018.

[9]     For example, ATSB Occurrence Investigation AO-2009-072 (reopened), Fuel planning event, weather-related event and ditching involving Israel Aircraft Industries Westwind 1124A, VH-NGA, 6.4 km WSW of Norfolk Island Airport, 18 November 2009.

Safety summary

What happened

On the afternoon of 18 August 2021, an Airbus Helicopters Deutschland BO105 CBS‑5 helicopter registered VH-NVH and operated by Surf Life Saving Queensland, departed Archerfield Aerodrome, Queensland to conduct aerial work operations. There were 3 crew on board.

During initial climb, the pilot noticed the onset of abnormal airframe vibration, which became more severe as airspeed increased. The pilot returned the helicopter to Archerfield, where it was landed without incident. The pilot subsequently identified a crack in one of the main rotor blades.

What the ATSB found

The main rotor blade crack was consistent with the in-flight vibration experienced. The crack originated at the location of a previous blade repair, and minor damage was also present in the same location on the other 3 rotor blades. There were no manufacturing or operational factors identified for the blade damage.

Safety message

This occurrence is a reminder for pilots to be vigilant for changes in aircraft noise and vibration. While the threshold for acknowledgement of changes can be subjective, based on an individual’s familiarity and experience, mild deviations from normal flight conditions could still be an indication of a developing technical issue.

Executive summary

Why the ATSB conducted this study

Between 1 July 2014 and 31 August 2022, 24 rail or road users were fatally or seriously injured in collisions between trains and heavy vehicles at Australian level crossings. The ATSB conducted this safety study to improve understanding of the risks associated with level crossing collisions involving heavy vehicles. The goals of this study were to:

• Compare the severity of level crossing collisions involving heavy vehicles to collisions involving light road vehicles.
• Provide a statistical description of level crossing collisions involving heavy vehicles.
• Identify common contributing factors associated with level crossing collisions involving heavy vehicles.
• Identify systemic safety issues associated with the safety system for level crossings, including the design of level crossings used by heavy vehicles.

To accomplish these goals, the ATSB used quantitative and qualitative methods to analyse data collected by rail regulators in Australia and the United States. The ATSB also reviewed all reported level crossing collisions involving heavy vehicles in Australia from July 2014 to August 2022 (49 collisions), and collected records from rollingstock operators, rail infrastructure managers, police, and other organisations for each of these accidents.

What the ATSB found

The annual number of level crossing collisions between road vehicles and trains remained relatively constant between July 2014 and June 2022.

Data from Australian and United States rail regulators showed that level crossing collisions involving heavy vehicles were more likely to lead to injuries to the occupants of rail vehicles, to damage to rail vehicles and track, and to derailment of rail vehicles. Derailment of rail vehicles was also more likely to occur in level crossing collisions in which the road vehicle struck the train, compared to those where the train struck the vehicle. Comparison of records of the number of total road vehicles registered in Australia, and records to the number of kilometres travelled by different road vehicle types, showed that heavy vehicles had a higher rate of level crossing collisions. As a result, heavy vehicles present a greater risk of level crossing collision as a function of both likelihood (per road vehicle and road kilometres travelled) and consequence compared to light road vehicles.

The majority of level crossing collisions arose from heavy vehicle drivers entering the level crossings following some form of unintentional error or omission. In at least 14 accidents, it was likely the heavy vehicle driver intentionally entered the level crossing in a manner contrary to road rules, however even in these instances the intention was to proceed through the crossing prior to the arrival of a train.

The review of level crossing collisions identified common factors which may have contributed to the actions of heavy vehicle drivers. These included:

• In at least 12 collisions the heavy vehicle driver had regularly used the level crossing prior to the collision with the train. The drivers' previous experience at the level crossings may have led to a low expectancy for trains and contributed to them not detecting a requirement to stop and give way.
• In at least 14 collisions, the heavy vehicle driver’s view of the track or level crossing protection equipment was obstructed by vegetation, the design of the heavy vehicle cab, poor crossing lighting, or sun glare.
• Consistent with prior research showing that train horns have limited effectiveness for alerting road vehicle drivers approaching level crossings, in at least 25 accidents the horn was not effective at alerting the heavy vehicle driver to the presence of the train.

These observations were based on a partial set of the 49 collisions involving heavy vehicles, since in many instances information about the actions of heavy vehicle drivers and the factors which influenced these actions was not available from operator or police reports. The review identified that such information is typically only available when an independent no-blame transport safety investigation is conducted.

The collisions reviewed in this study mostly occurred at level crossings designed according to the required standards. However, this study identified one systemic safety issue with level crossing design standards. Of 6 collisions where the heavy vehicle driver did not identify the presence of activated flashing lights, 5 occurred on crossings with curved road approaches, all of which were right curves. Following a review of the applicable standards for designing and assessing level crossings (AS1742.7:2016 Manual of uniform traffic control devices, Part 7: Railway crossings), and research literature concerning visual attention in curved driving, the ATSB determined that methods used to calculate safe stopping distances (and determine the need and location of active advanced warning signs) for road approaches to level crossings did not account for the likelihood of detecting the level crossing ahead based on the normal visual focal points of road drivers negotiating a curved road.

The collisions primarily resulted from level crossing warnings or the presence of trains not being detected, being detected late, or being perceived incorrectly. Many of the errors and omissions which lead to level crossing accidents reflect the inherent variability of human performance.

The review identified that passively controlled level crossings, in particular, rely on the road vehicle driver visually detecting the presence of a train and identifying a requirement to stop and give way, they are susceptible to situations where drivers do not look for, or detect the presence of, a train, and there are limited effective recovery controls to prevent a collision when this occurs. The use of redundant controls, and particularly fail-safe engineering controls, improve the safety at passively controlled level crossings. The study considered one such form of control (in-vehicle alerting systems for level crossings), however there are significant barriers to the implementation of such systems.

What has been done as a result

Standards Australia reported they will review the standard AS1742.7 and determine if additional information or guidance may be provided to manage risks associated with curved road approaches to level crossings.

Safety message

This study highlights the increased risk posed by level crossing collisions with heavy road vehicles. Interventions which target these vehicles are likely to provide significant leverage in improving the overall safety at level crossings.

The accidents reviewed in this study demonstrate the potential for driver errors at level crossing to cause significant harm. As also demonstrated in this study, the performance of road vehicle drivers is inherently susceptible to unintentional errors. While the level crossing safety systems rely on road vehicle drivers always detecting a level crossing (and at passive crossings, the presence of trains) it is certain that this will fail from time to time and result in accidents in the future. The possibility of the future use of engineering controls which alert road users to a requirement to stop will almost certainly provide an enhanced level of safety at level crossings, by reducing the reliance on road vehicle drivers to attend to and detect the presence of trains.

Background

The Australian rail network comprised about 46,000 km of track and had an annual throughput of approximately 250 million km of combined passenger and freight journeys in 2021.[1] Across this network, records from the Office of the National Rail Safety Regulator (ONRSR) indicated there were almost 11,000 operational[2] level crossings where there is an interface between rail traffic and road vehicles.[3] At these interfaces there is a risk of collision, which is managed using different forms of traffic control equipment and associated procedures.

Relative to the volume of traffic conveyed across the rail network, collisions between trains and road vehicles at level crossings are rare. Rail safety data produced by ONRSR showed that for every million km travelled by freight trains between 2016 and 2021, there was an average of about 0.18 collisions with road vehicles at level crossings. The rate for passenger trains was about 0.10 per million km.

Though rare, level crossing collisions can have severe consequences. Collisions between passenger cars, motorcycles or bicycles often result in tragic consequences for road users. Where a heavy vehicle is involved in a level crossing collision, the rail system risks associated with the collision are much greater, including the increased potential for multiple fatalities onboard the train. In recent decades, several collisions between heavy vehicles and trains have resulted in multiple fatalities of train occupants:

• On 27 November 2008, 2 train drivers were fatally injured when a B-double combination truck collided with a passenger train in Rungoo, Queensland.[4]
• On 5 June 2007, 11 train passengers were fatally injured when a semi-trailer hit a train at a level crossing in Kerang, Victoria.[5]
• On 13 October 2002, 3 people in the cabin of a steam train were fatally injured when it collided with a B-double combination truck at a level crossing in Benalla, Victoria.[6]

To understand the trends and characteristics of level crossing collisions involving heavy vehicles requires sufficient information to enable detailed analysis. The ATSB, as the national independent investigator of rail safety occurrences, investigated about 20% of these occurrences between July 2014 and July 2022.

Safety study goals

This safety study aimed to improve understanding of the risks of level crossing collisions involving heavy vehicles. The goals of this study were to:

• Compare the severity of level crossing collisions involving heavy vehicles to level crossing collisions involving other road vehicles.
• Provide a statistical description of level crossing collisions involving heavy vehicles.
• Describe common themes associated with level crossing collisions involving heavy vehicles.
• Identify systemic issues associated with the safety system for level crossings, including the design of level crossings used by heavy vehicles.

Level crossing safety systems: an overview

Level crossings are the physical interface between road and rail traffic. Both modes of transport are operated as entirely separate entities and have different rules, procedures, characteristics and operational limitations. Trains are typically significantly larger than road vehicles, and when a road vehicle enters a level crossing in the path of a train, the only action that a train driver can take is to try to alert the road vehicle driver with the train horn and apply train braking. The train may not slow significantly, if at all, before the collision occurs.

Given the limitations on braking and accelerating trains, by necessity, road vehicle drivers must give way to trains at level crossings. Road traffic law codifies the right of way of rail traffic, with the Australian road rules (model law) stating:

Level crossing traffic control systems use various signs, lights and other devices (level crossing protection equipment) to firstly alert an approaching road vehicle driver to the presence of the level crossing, and secondly instruct the road vehicle driver to take appropriate action to give way to trains.

Passive control level crossings control the movement of road traffic using signs and devices (including Give Way or Stop signs). Passive level crossing equipment is functionally static, meaning that there is no change to the equipment when a train is approaching compared to when no train is approaching. As such, passive control level crossings rely on road vehicle drivers (and pedestrians) detecting both the sign itself and then the approach or presence of a train by direct observation.

Active control level crossings control the movement of road traffic using devices such as flashing light signals, gates or barriers, or a combination of these. Active level crossing devices are activated prior to, and during, the passage of a train through the crossing. If the crossing devices are activated (level crossing lights are flashing and/or boom gates are down), then road vehicle drivers' attention should be drawn to their active status, and road rules require the road vehicle driver to stop and wait until the crossing devices are no longer active, removing the need for the road vehicle driver to detect the presence of a train.

Methods and sources

Overview

To analyse trends and characteristics (including accident severity) of level crossing accidents in Australia, this study used records of level crossing collisions and near misses in the national rail safety data collected by the Office of the National Rail Safety Regulator (ONRSR).

The study also used data from the United States Federal Railroad Administration (FRA) to support analyses of level crossing accident severity. The road and rail systems in the US are broadly equivalent to those in Australia.[7] Both jurisdictions (Australia and the United States) use similar forms of level crossing protection mechanisms, and the fundamentals of level crossing collisions are unlikely to be significantly different; level crossing collisions involve relatively small road vehicles colliding with comparatively larger trains. The ATSB considered that level crossing accident data from the United States provided valid data for the purpose of the analyses.

Australian level crossing collision data

Office of National Rail Safety Regulator national rail safety database

The Rail Safety National Law Act 2012 required railway operators report all notifiable occurrences to ONRSR or another authority specified by ONRSR. The Rail Safety National Law Regulations 2012 specified that collisions between trains and vehicles at level crossings were Category A occurrences, whereas near misses at level crossings were Category B occurrences.[8] The Rail Safety National Law does not require road vehicle operators to report level crossing accident information to ONRSR.    

The national rail safety database recorded information provided to ONRSR under the reporting requirements specified in the Rail Safety National Law. This database commenced with the establishment of ONRSR on 20 January 2013, however coverage of different Australian jurisdictions (States) occurred gradually. ONRSR advised the dataset for level crossing occurrences was reliable from 1 July 2014 onwards. The ATSB analysis used occurrences from the national rail safety database between 1 July 2014 and 30 June 2022.

The national occurrence database recorded:

• a narrative description of the collision or near miss, which typically recorded the initial wording of the notification provided by the operator and sometimes included follow-up information
• the date, time and location of the level crossing collision or near miss
• the operator and rail infrastructure manager involved
• the primary level crossing protection equipment (flashing light, boom gate, stop sign or give way sign) at the level crossing
• the type of rollingstock and road vehicle involved in the collision or near miss
• fatalities and injuries sustained by rail passengers, rail operating crew and road vehicle occupants.

For the time period analysed, the national rail safety database did not record:

• the level of damage sustained by road vehicles, and rail and road infrastructure
• the speed, direction, mass and length of rail or road vehicles
• the actions of the road user.

The national rail safety database recorded 283 level crossing collisions with vehicles between 1 July 2014 and 30 June 2022. Of these, 220 were collisions with light passenger vehicles, with 44 collisions involving heavy freight vehicles.[9] The remaining 20 collisions involved buses (4), special purpose machinery (5), bicycles (5), motorcycles (1), a dangerous good vehicle (1) and other vehicles (4). To analyse the severity and characteristics of occurrences involving heavy vehicles, the ATSB identified the level crossing collisions involving heavy vehicles recorded in the ONRSR national rail safety database for additional evidence collection.

A set of 47 collisions was identified in the national rail safety database, which comprised:

• 42 collisions with vehicles recorded as heavy freight vehicles[10]
• 3 collisions with buses[11]
• one collision each with vehicles coded special purpose machinery,[12] and dangerous goods vehicles.

Of the remaining collisions in the national rail safety database, the ATSB identified a subset of 216 level crossing collisions involving light road vehicles for comparative analyses. Light vehicles were defined as passenger cars and motorcycles, but not bicycles or farm or roadwork equipment. The comparative analysis also excluded collisions with hi-rail/road-rail vehicles.[13]

In July 2022, ONRSR introduced new requirements for operators submitting notifiable occurrence reports. Under the new requirements, rail operators reporting level crossing collisions and near misses were required to report the damage sustained by rail vehicles and infrastructure, whether a derailment occurred, and the cause of the collision.

Additional information collection

Between July and August 2022, the ATSB commenced 2 investigations into level crossing collisions involving heavy vehicles, these collisions were added to the 47 heavy vehicle level crossing collisions identified from the national rail safety database.

For each of the 49 in-scope collisions, the ATSB sought information from rail operators, rail infrastructure managers, police, state road authorities and other parties. These records included investigation reports and level crossing assessments. Follow-up enquires were made to enhance the dataset. However, in many instances important information was not recorded in investigation reports and was not available for follow-up enquiry due to the passage of time from the events. As such, some of the information presented in subsequent sections is drawn from a partial sample of the identified level crossing collisions.

Level crossing accident data collected in United States

To support analysis of level crossing collision severity, the ATSB obtained records from the United States Federal Railroad Administration (FRA). Due to greater rail and road traffic volumes, there were about 60 times more recorded level crossing collisions in the United States than Australia, and the FRA datasets provided additional statistical power for the ATSB analyses. Regulations in the United States required that collisions between railway equipment and vehicles at level crossings (referred to as ‘grade crossings’ in the United States) were reported to the FRA. The regulations required that reports of level crossing collisions included the:

• type of road user involved
• speed, direction and position of the road user
• environmental temperature, light and weather conditions
• train type, consist length and speed
• level crossing equipment
• actions of the road user.

In addition, for accidents exceeding a threshold damage value (about A$16,000 in 2022), operators were required to report the:

• mass of the train consist
• cost of rail equipment and track damage
• number of derailed train cars.

These data were contained in two databases. The Highway-Rail Grade Crossing Accident (HRA) database contained data for all level crossing accidents, and the Rail Equipment Accident/Incident (REA) database contained data on accidents exceeding the damage cost threshold.

ATSB analysed data from both datasets from 2012 to 2021. After data cleaning and removal of irrelevant collisions, there were 19,495 collisions in the HRA dataset, and 2,283 records of highway rail collisions (level crossing collision) in the REA dataset. There were about 700 duplicate records in the REA dataset. The ATSB identified a final sample of 1,901 unique HRA records with matching REA records.

Descriptive statistics

Australian level crossing statistics

ONRSR provided the number of road crossings, based on information reported by rail operators. ONRSR stated that while the number of road crossings had been validated by reasonability checks, it was not possible to determine the accuracy of the data reported by operators.

The ONRSR database showed that in December 2023, there were almost 11,000 operational level crossings which managed traffic between trains and road traffic. This included almost 3,000 active control crossings, and almost 7,000 passive control crossings (there were about 1,500 operational crossings which were recorded as unprotected). ONRSR advised it was not possible to disaggregate their data on the basis of whether level crossings were on public or private roads. Table 1 shows the distribution of operational level crossings by protection type in each Australian state and territory.

Table 1: Counts of road level crossings, by primary protection type and jurisdiction

Source: ONRSR

During the development of the safety study, the ATSB received submissions from stakeholders in some jurisdictions which identified that the number of level crossings presented in Table 1 were not consistent with other records, including those stored in the Australian Level Crossing Assessment Method dataset. ONRSR also identified that there were known differences between the number of crossings identified in the ONRSR database and the ALCAM dataset, and  the data had not been cross-validated between the datasets.

Australian level crossing collisions

Characteristics of all level crossing collisions and near misses

There were 283 collisions and 4,319 near misses between trains and road vehicles recorded in the Australian national rail safety database. These totals included collisions involving hi-rail vehicles, bicycles, and farm and roadwork equipment.

Key statistics involving the type of train included:

• There were 120 collisions and 1,801 near misses involving freight trains.
• 122 collisions and 1,895 near misses involved regular passenger services.

Key statistics involving the type of crossing included:

• 160 collisions and 3,133 near misses occurred at active control crossings.
• 115 collisions and 1,086 near misses occurred at passive control crossings.

Other descriptive statistics for collisions and near misses are described in Appendix A.

Characteristics of level crossing collisions involving heavy vehicles

Environmental conditions

Of the 49 collisions reviewed in the safety study, 3 occurred in dark conditions (after astronomical twilight).[14] There were a further 7 collisions when the sun’s elevation was between -18° and 6°, meaning the sun was low in the sky and twilight conditions were apparent.

Information about the road conditions was recorded for 40 of the collisions. In 37 collisions the road was reported to be dry, and in 3 collisions the road was reported to be wet. Thirty-three of the 49 collisions occurred in rural locations, and 16 collisions occurred in built-up or urban areas.

Train characteristics

More than two-thirds of level crossing collisions with heavy vehicles involved freight rail operations (Table 2).

Table 2: Operation category of trains involved in level crossing collisions with heavy vehicles

The maximum reported train mass was 10,109 tonnes, with 18 collisions involving trains with a reported mass of over 1,000 tonnes. The maximum train length was 2,964 m, with 6 collisions involving trains 1,000 m or longer. Thirteen collisions involved driver only operations. Where the number of passengers was recorded, the maximum was 250 (Table 3).

Table 3: Maximum and average train mass, length and passenger numbers

Information about the time and distance of emergency braking prior to collision was available for 36 collisions (Table 4). In 13 instances, the train crew either did not brake, or braked at about the point of impact (2 seconds or less prior to impact). These included instances where the train crew saw the road vehicle slowing or stopped for the crossing, and therefore did not anticipate that a collision would occur. These instances also included situations where the train crew’s vision of the level crossing was blocked by curved track and vegetation, preventing the train crew from identifying the potential collision earlier.

In 4 instances, train crew made emergency brake applications for 10 or more seconds prior to the collision, including up to 213 m braking distance prior to the collision. Three of these collisions involved instances were at level crossings controlled by boom gates, and involved heavy vehicles stopped on the crossing. The other instance involved a very large freight train (5,330 tonnes) which was unable to stop after a truck driver stopped then proceeded from a Stop sign. Due to the train braking over a long distance, the collision speed in these occurrences was significantly reduced, probably leading to reduced consequences (injuries and damage) to the train and road vehicle occupants than if braking had been applied at a shorter distance to the collision point.

The average reported speed of the trains prior to braking for the collision was 62 km/h, with a maximum of 160 km/h reported. The average reported collision speed was about 51 km/h, with a maximum reported speed of 110 km/h.

Table 4: Train speed and braking prior to level crossing collisions

Heavy vehicle information

Over half (28) of the heavy vehicles involved in level crossing collision were articulated vehicles (semi-trailers or trucks with trailers). Seventeen vehicles were unarticulated vehicles (garbage trucks, flat-bed and box trucks) and 3 were buses. In terms of the classes of heavy vehicles described by the National Heavy Vehicle Regulator,[15] there were 27 General Access Vehicles, 14 Restricted Access Vehicles, and there was insufficient information to classify 8 vehicles.

The maximum recorded gross mass was 148 tonnes, and 6 vehicles had a recorded mass of 70 tonnes or greater. The greatest number of occupants recorded was 20 (a passenger bus), with the next highest being 3. Thirty-seven collisions involved single-occupant vehicles, and 3 collisions involved circumstances where the occupants had vacated the vehicle prior to the collision.

Table 5: Mass, length and occupant numbers for heavy vehicles involved in level crossing collisions

Reported mechanical problems were associated with 4 of the collisions identified in this study. In 2 instances, post-accident assessments identified problems with truck braking systems which may have contributed to the trucks failing to stop ahead of the crossing. In 2 other collisions, the heavy vehicles stalled on the level crossing and were unable to clear the crossing prior to the train arriving.

Police records indicated that one heavy vehicle driver was affected by alcohol or other drugs, and was reported to be fatigued/drowsy prior to the collision.

Level crossing characteristics

There was a relatively even distribution of collisions involving heavy vehicles at passive control crossings and collisions at active control crossings. Considering that heavy vehicles accounted for about 18% of level crossing collisions recorded in the national rail safety occurrence database, heavy vehicles were significantly under-represented for collisions occurring at Boom gates, and over-represented for collisions at Give way signs (Table 6).

Table 6: Level crossing collisions by protection equipment type

Most occurrences (39) occurred on sealed roads, with 10 collisions occurring on unsealed roads. The average road speed limit, where recorded, was 77.1 km/h, and 15 collisions occurred on roads where the speed limit was 100 km/h or greater. Where recorded, the average track speed at the level crossing locations of the occurrences was 82.6 km/h, with a maximum track speed of 160 km/h.

Problems with level crossing design and maintenance

There are several standards applicable to the design and maintenance of level crossings in Australia. These include:

• Australian Standard AS 7658:2020 Level Crossings: Rail Industry Requirements. This standard was developed and maintained by the Rail Industry Safety and Standards Board (RISSB). It describes minimum operational and engineering requirements of the life cycle of a level crossing.
• Australian Standard AS 1742.7(2016), Manual of uniform traffic control devices. Part 7: Railway crossings. This standard was developed by the Australian Standard Committee MS‑012. It described design guidance for the installation of level crossing protection equipment, including methods for assessing the sighting distances required at different level crossings. Additional discussion of AS 1742.7 is provided in Standards and other guidance relevant to placement of crossings on curved road approaches, and the formula used to calculate sighting distances for level crossings are described in Appendix B.

Key aspects of these standards include that:

• AS 7658 required that rail infrastructure managers establish and maintain procedures for the rail portion of the level crossing, and that an interface agreement exists with the road manager to ensure that ‘all facets of the monitoring and maintenance of the level crossings is undertaken’. This includes procedures for inspection and testing and vegetation management. The standard further required that ‘removal of plant growth shall be required to ensure retention of sighting distance’.
• AS 7658 stated that whistle boards may be installed on the approach to level crossings.
• AS 7658 stated that that ‘Adequacy of street lighting at a level crossing shall be assessed against the requirements in AS/NZS 1158 [Road lighting] during a level crossing risk assessment. Street lighting provided for level crossings shall comply with AS/NZS 1158 and impacts on train drivers, road users and pedestrians shall be evaluated, and outcomes documented.’
• AS1742.7 stated that level crossings should be located so that there is sufficient distance between the crossing and a downstream intersection to accommodate the road design vehicle[16] and a safety factor of 5 m. The intent of these provisions is to prevent so-called ‘short stacking’, where a long vehicle stops across a crossing while queueing for another intersection.

The safety study considered evidence of problems with the design or maintenance of the level crossing locations of the 49 collisions reviewed in this study. Analysis of the design of AS1742.7 is described in Analysis of level crossing safety systems. Two ATSB investigations, 4 rail infrastructure manager reports and one State government investigation identified problems with the design or maintenance the level crossing:

• ATSB investigation RO-2014-024 identified that a stand of trees had self-sown within the rail reserve. When a truck approached the level crossing, the driver’s view was partially obscured by this vegetation. The investigation noted that previous inspections had recorded the presence of the trees, but this had not led to the removal of the trees.
• ATSB investigation RO-2016-009 identified that the Stop sign level crossing required road vehicle drivers sight approaching rail traffic at an obtuse angle (greater than 90° but less than 180°). When a semi-trailer truck stopped at the crossing, the driver’s view of an approaching train was restricted and they did not identify the risk of collision.
• A rail infrastructure manager report noted that a Stop sign level crossing was located in a dark industrial area. There was poor lighting on the crossing itself and contrasting high levels of lighting in the background and foreground. In these dark conditions, a truck driver approaching the Stop sign level crossing did not detect that a rail safety worker was controlling the crossing or that a freight train was indexing in reverse over the crossing.
• A rail infrastructure manager and operator report noted that a temporary Stop sign level crossing had been installed without other approach signage, and with no whistle board installed for approaching trains. A train driver approaching the crossing did not sound the train horn until they identified a truck, which had stopped at the level crossing. The truck driver did not hear the horn and proceeded into the path of the train.
• A state government investigation report identified that a level crossing protected by flashing light signals was designed with a stop line placed too close to the lights assembly. When a bus stopped at the crossing while waiting for traffic to clear, the driver’s view of the near-side lights was obstructed. When traffic cleared, the bus driver, who had an unobscured view of the far-side flashing lights, proceeded into the crossing against the flashing lights.
• A rail infrastructure manager report identified that the road configuration near a Stop sign level crossing was not appropriate for some heavy vehicle classes as there was insufficient room to clear the crossing when queued at an intersection beyond the crossing. There was no signage or pavement markings to discourage road users from queuing across the crossing. A heavy vehicle driver approaching the crossing did not stop at the Stop sign (causing the train drivers to apply emergency brakes), then was unable to clear the crossing before the train arrived.
• A rail infrastructure manager and operator report identified that a level crossing flashing light assembly did not face towards a side road approaching the crossing. A heavy vehicle driver approaching from the side road did not detect the requirement to stop at the level crossing.  

In total, there were at least 7 collisions between heavy vehicles and trains where problems with the design or maintenance of the level crossings were probably contributory to the collisions. Reports from a further 2 collisions identified that the sighting provided to road users was insufficient considering the requirements of Australian Standards AS1742.7. In these instances, the sighting deficiencies were for rail or road approaches other than those involved in the collisions, so were not contributory to the collision. Reports for 8 other occurrences identified various other minor problems, including faded signs and minor discrepancies between the signage provided and the requirements of AS1742.7.

In summary, the evidence available to this review did not indicate significant problems with the design or maintenance of most level crossings where collisions between heavy vehicles and trains occurred, in terms of compliance with applicable design standards. The absence of such evidence, however, is not necessarily evidence that no problems with the design and maintenance of crossings existed. As described further in Study limitations, this review did not involve site inspections of level crossings (other than those previously investigated by the ATSB), and the ATSB did not seek to otherwise audit the compliance of the level crossings with relevant standards.

Level crossing collisions in the United States

Collisions and fatalities by vehicle types

The number of level crossing collisions and fatalities recorded in the US Federal Railroad Administration (FRA) Highway-Rail Grade Crossing Accident (HRA) dataset is summarised by vehicle type (Table 7). The vehicle type variable was then categorised as Heavy (bus, truck or truck-trailer) and Light (all other vehicles), with the number of collisions and collision fatalities for each category shown in Table 8.

Table 7: Level crossing collision statistics by road vehicle type, FRA (HRA) data, June 2012 to June 2021

[1] This category describes passenger vehicles such as sedans, but does not include motorcycles

Table 8: Level crossing collision statistics for 'Heavy' and 'Light’ vehicles (excluding pedestrian collisions)

Types of trains involved

Concerning all road vehicles, 74% of level crossing collisions involved freight trains. Collisions involving heavy vehicles accounted for about 25% of all collisions with freight trains, and about 16% of collisions with passenger trains.

Table 9: Level crossing collisions by train type and vehicle type, FRA HRA data

Level crossing protection equipment

More than half (56%) of level crossing collisions were recorded at active control level crossings. Heavy vehicles were involved in about 25% of all level crossing collisions (active and passive controlled crossing combined). They were over-represented in collisions at passive control crossings, and under-represented for collisions at active control crossings (Table 10).

Table 10: Level crossing collisions by crossing level crossing protection equipment type and vehicle type

Note: ‘Proportion heavy’ describes the proportion of accidents at each level crossing type which involved heavy vehicles. For example, 37% of accidents at Stop sign crossings involved heavy vehicles.

Road vehicle driver actions associated with level crossing collisions

The most common road vehicle driver action associated with level crossing collisions was a failure to stop prior to a level crossing, accounting for 40% of collisions (Table 11). Heavy vehicles were proportionately more likely to be involved in collisions where:

• The road user did not stop at the level crossing (1.3 times more likely).
• The road user stopped at the crossing then proceeded into the path of an oncoming train (1.6 times more likely).

Table 11: Level crossing collisions by road vehicle driver actions and vehicle type

The road user action attributed to level crossing collisions varied significantly as a function of the type of level crossing protection equipment in use. Table 12 describes the proportion of collisions attributable to different motorist actions, by crossing and vehicle type.

Table 12: Motor vehicle actions by crossing and road vehicle type

Note: Percentages shown in this table describe the proportion of accidents for each crossing type associated with each action type. For example, 22% of Heavy vehicle collisions at Stop sign controlled crossings involved the motor vehicle driver stopping on the crossing.

Vehicle speed

Most collisions recorded in the FRA HRA dataset involved trains colliding at speed with stopped or slow-moving road vehicles. The mean recorded train speed was 47 km/h, with a maximum reported speed of 177 km/h. The mean estimated collision speed for road vehicles was about 11 km/h for heavy vehicles and about 14 km/h for other road vehicles.

Comparison of Australian and United States Level Crossing collision data

There were consistent characteristics in the level crossing collision data for Australia and the United States. There was a similar proportion of level collisions involving heavy vehicles, with 18% of collisions in Australia involving heavy vehicles, compared to 25% in the United States. Similarly, data for both countries showed that heavy vehicles were under-represented in accidents occurring at crossings protected by boom gates, and over-represented in accidents occurring at passively controlled crossings.

The FRA data showed that the most common road user action associated with level crossing collisions was a failure to stop. As described in Errors and violations associated with level crossing collisions, this action was also the most common road user action identified in the 49 collisions involving heavy vehicles in Australia.

Analysis of heavy vehicle collision severity and frequency

Severity of level crossing collisions involving heavy vehicles

Previous research and reviews

In 2008, the ATSB published a

Analyses of US level crossing collision datasets have previously shown that when heavy vehicles were involved in a collision, there was a much greater likelihood of a derailment occurring (Chadwick, Saat and Barkan, 2012).[17] Other research noted that the damage costs of level crossing collisions involving commercial vehicles (which includes heavy vehicles) are 3-4 times greater than collisions involving other road vehicles (Hellman and Poirier, 2019). 

Injuries and fatalities to rail passengers and crew

Australia

The ATSB’s analysis of the national rail safety database identified 47 injuries to rail passengers and crew in 285 level crossing collisions with road vehicles,[18] 7 of these injuries were classified by the ATSB as serious as they resulted in hospital admission. The majority of injuries to rail occupants (7 serious and 34 minor) resulted from 49 reported collisions involving heavy vehicles, this included 18 passengers and a conductor who sustained injuries after a collision between a passenger train and a truck on 15 July 2016 at Phalps Road, Victoria.

There were no fatalities of railway employees or passengers resulting from a level crossing collision in the period July 2014 to August 2022.

United States

In the collisions recorded in the FRA HRA dataset, there was an average of 25 rail injuries (injuries to rail passengers or staff) per 100 collisions involving a heavy vehicle, compared to 3.5 per 100 collisions involving light vehicles. A Mann-Witney U test[19] found that the difference in average rail injuries by vehicle type was statistically significant (p<.005).

Fatalities to rail occupants were extremely rare, with 7 collisions (out of 19,495) resulting in 11 rail fatalities. There were 5 separate collisions involving heavy vehicles which each caused one rail fatality, whereas one collision involving a light vehicle resulted in 5 fatalities[20] (another collision with a light vehicle resulted in one fatality). Table 13 shows the total number of fatalities and injuries of rail and road users recorded in the FRA datasets.

Table 13: Fatalities and injuries by road user type in US level crossing collisions, 2011 to 2021

Fatalities and injuries to road users

Australia

There were 23 road fatalities from 283 level crossing collisions recorded in the national rail safety database, with 5 fatalities resulting from collisions involving heavy vehicles and 17 resulting from collisions with non-heavy vehicles (one fatality involved a collision with a cyclist). The average number of road fatalities per collisions involving heavy vehicles (0.11 fatalities per collision) was greater than collisions involving non-heavy vehicles (0.08 fatalities per collision), however the difference was small.

The ATSB review of 49 heavy vehicle collisions identified that road vehicle drivers and passengers sustained 33 injuries from level crossing collisions with heavy vehicles (0.67 injuries per collision), including 12 serious injuries. This included 17 injuries (6 serious) sustained by the occupants of a passenger bus which was struck by a train at Draper Street in Brisbane in 2015.[21] The national rail safety database showed that there were 60 injuries to road users in the 216 collisions involving light vehicles (0.28 injuries per collision).

United States

Analysis of the FRA HRA dataset (19,495 collisions) showed there were 1,770 total road fatalities. A Mann-Witney test showed that the average number of road fatalities per collision was statistically significantly lower for collisions involving heavy vehicles (0.05 fatalities per collision), compared to collisions involving light vehicles (0.11 fatalities per collision;  p<.001).

Similarly, there were fewer injuries to road users for collisions involving heavy vehicles (0.23 per collision), compared to collisions involving light vehicles (0.36 per collision; p<.001).

Rail damage

Australia

The national rail safety database did not record estimated damage for level crossing collisions. As such it was not possible to compare the damage sustained in Australian level crossing collisions involving heavy vehicles, compared to collisions involving light vehicles. For the 49 collisions reviewed in this study, there was typically insufficient information to quantify the damage sustained by rail vehicles, road vehicles or infrastructure in most collisions.

The ATSB reviewed the damage reported in various sources, and identified that:

• 11 collisions resulted in the derailment of at least 1 rail car.
• In 16 instances the rail vehicle sustained substantial damage, which required major repairs or significantly affected the performance of the vehicle. In 31 instances rail vehicles sustained only minor damage such as scratches or small dents.
• In 33 instances the road vehicle sustained substantial damage, including several collisions where the vehicle was destroyed. In 13 instances the road vehicle sustained minor damage.
• In 23 collisions there was no reported damage to infrastructure. In 10 instances there was minor infrastructure damage reported, and in 7 instances there was substantial infrastructure, such as large sections of destroyed track.

The following examples illustrate the significant rail damage resulting from collisions with heavy vehicles:

• A collision between a truck, trailer and dog combination with a grain train resulted in extensive damage to the locomotive, with total repair costs estimated at $480,000.
• A collision between a prime mover and trailer with a freight train resulted in damage to 20 m of track, 1,250 sleepers and 90 tonnes of ballast, with an estimated repair cost for infrastructure assets of over $300,000.
• A collision between a prime mover and a grain train led to the derailment of the train, resulting in significant damage to two locomotives and multiple wagons.

United States

The FRA Rail Equipment Accident/Incident (REA) dataset recorded the estimated rail equipment (rollingstock) and track damage for accidents which exceeded the damage threshold (about A$16,000). Of 1,901 level crossing collisions recorded in this dataset, the average equipment damage for collisions involving heavy vehicles was around USD$78,000, compared to USD$31,000 for collisions involving light vehicles. The average track damage for level crossing collisions involving heavy vehicles was USD$31,000, compared to USD$19,000 for collisions involving light vehicles. Mann-Witney tests showed both of these differences were statistically significant (p<.001).

Derailment

Australia

A mixed-methods approach was used to identify the number of derailments resulting from heavy vehicle and light vehicle level crossing collisions:

• All level crossing collisions (heavy and light road vehicles) in the national rail safety database were cross referenced with those also classified as derailments and/or identified as a derailment in the accident description field.
• For collisions involving heavy vehicles, operator and rail infrastructure manager reports were used to supplement the data from the nation rail safety database. Combined these methods identified that 11 of the 49 collisions resulted in derailment (9 of 47 collisions prior to July 2022).
• For collisions involving light vehicles, there was only one of 216 collisions that resulted in derailment. This was a collision between a passenger train and an abandoned light vehicle.

Fisher’s exact test was conducted to examine the relationship between road vehicle type and derailment for the 283 heavy and light vehicle collisions in the national rail safety database. The relationship was significant (one-tailed p<0.001), showing that collisions involving heavy vehicles were more likely to lead to derailment.

United States

The FRA data showed that there were 141 derailments in 4,886 collisions involving heavy vehicles, compared to 43 derailments in 14,609 collisions involving light vehicles.[22] Fisher’s exact test was showed that the difference in distribution between derailments and vehicle type was also significant in the FRA data (one-tailed p p<.001).

Summary of analysis of accident severity

Consistent with previous research, level crossing accident data from Australia showed that collisions involving heavy vehicles were more likely to result in injury to rail and road vehicle occupants, and more likely to cause derailment of rollingstock. Due to limitations in the available data, it was not possible to compare the accident damage sustained to rollingstock, infrastructure and road vehicles in level crossing collisions involving heavy vehicles and light vehicles, for collisions occurring in Australia.

The data from the US was broadly consistent with the analysis of the Australian data, showing that collisions involving heavy vehicles resulted in more frequent injury to rail occupants and more frequent derailments. The data also showed that rail damage was significantly greater for collisions involving heavy vehicles.

In contrast to the Australian data, the US data showed there were significantly more injuries to the occupants of light road vehicles involved in level crossing collisions, compared to occupants of heavy road vehicles. The difference between these data sources is probably explained by the small number of collisions in the Australian sample, and thus the influence of a single multi-injury accident in the Australia data, which substantially increased the average number of road vehicle injuries in heavy vehicle collisions.

Over-representation of heavy vehicles in level crossing collisions

Previous research and reviews

The Independent Transport Safety Regulator for New South Wales (ITSR) described level crossing statistics for Australia from 2000 to 2009 (ITSR, 2011). The ITSR report showed that heavy vehicles were over-represented in level crossing collisions, accounting for 20% of collisions and 23% of fatal collisions while making up only 2.5% of vehicle registrations and 6% of kilometres travelled. The report also showed that level crossing collisions involving heavy vehicles had double the fatality rate per collision as collisions involving light vehicles.

The industry body Austroads (2010) reviewed level crossing collisions occurring in Australia and New Zealand from 2003 to 2007. This report found that of 405 total collisions, 20% involved heavy vehicles, with 14% being articulated heavy vehicles.

Analysis of level crossing collisions in Canada showed that occupants of heavy vehicles accounted for 40% of all level crossing collision fatalities from 1983 to 2001 (Caird, 2002). This review noted that light trucks, accounting for 27% of fatalities, typically operated in rural and industrial urban areas where there are more level crossings. A later analysis of Canadian accident records found heavy vehicles were overrepresented in level crossing collisions at rural, passive level crossings where the driver of the vehicle did not stop (Rudin-Brown et al., 2014).

Data used to normalise collision numbers by vehicle type

The Bureau of Infrastructure and Transport Research Economics (BITRE) Australian Infrastructure and Transport Statistics - Yearbook 2022 provided estimates of the number of kilometres travelled by vehicle type, by financial year. This data showed that between the financial years 2014-15 and 2021-22, heavy vehicles travelled on average 21.65 billion km each year and light vehicles travelled on average 1.83 trillion km.

The Australian Bureau of Statistics Motor Vehicle Census provided estimates of the total number of registered motor vehicles by vehicle type, by calendar year. This data showed that from 2015 to 2021, there were on average 707,000 heavy vehicles registered in Australia,[23] and about 18.47 million light vehicles.

Comparison of heavy vehicle and light vehicle level crossing collision rates

As described in Australian level crossing collision data, from 1 July 2014 to 30 June 2022, there were 47 level crossing collisions involving heavy vehicles, and 216 collisions involving light vehicles. As shown in Table 14, heavy vehicles were over-represented in level crossing collisions as a function of both the number of collisions per vehicle (around 2 times greater), and the number collisions per kilometre travelled (around 4 times greater).

Table 14: Ratios of collisions by vehicle type to distance travelled and number of registered vehicles

Caution should be exercised in making inferences regarding whether the over-representation of heavy vehicles in level crossing collisions necessarily indicates a greater level of risk. It may also reflect a greater likelihood of encountering level crossings, or a greater likelihood of encountering crossings of higher risk. It is plausible that heavy vehicles are more likely to be operated on rural roads, which are less likely to have level crossings removed through grade separation or be protected by higher level active controls (boom gates). There are no known data describing the rate at which different types of vehicles encounter and traverse level crossings of different types and characteristics, and thus this explanation could not be tested.

Number of level crossing collisions over time

Prior to the establishment of the Office of the National Rail Safety Regulator (ONRSR) in 2013, records of the number of level crossing collisions in Australian jurisdictions were collected by state and territory regulators. The ATSB collected rail collision information from these regulators, and published various statistics including the number of level crossing collisions. The ATSB publication Australian Rail Safety Occurrence Data 1 July 2002 to 30 June 2012 showed that in the decade to July 2012, there was a notable drop in the number of level crossing collisions with vehicles. The number of collisions declined from 82 in the year to July 2002 to 49 in the year to July 2012.

As described in Australian level crossing collision data the national rail safety database was reliable from 1 July 2014 onwards. As such, there was a transitional period from the discontinuation of reporting of state and territory regulator collision statistics in July 2012, to the availability of annual totals commencing 1 July 2015.

There was a substantial drop in the number of level crossing collisions from 49 in the year to July 2012, to 30 in the year to July 2015. From July 2014, the number of level crossing collisions has remained relatively constant, at around 30 to 40 per year (Figure 1). A Mann-Kendall test was conducted on the number of collision per financial year between July 2014 and June 2022 (Figure 2), the test did not identify any statistically significant trend (p = 0.61).

Figure 1: Level crossing collision with vehicle, Financial year totals 2002 to 2022

Figure 2: Monthly level crossing collisions, July 2014 to June 2022, with regression line

Level crossing collision sequence and derailment

Level crossing collisions can either involve the train striking a vehicle (TSV) or a vehicle striking the train (VST). In TSV collisions the impact involves the front of the train striking either a moving or stationary road vehicle, whereas in VST collisions the impact involves a road vehicle striking the side of a moving or stationary train. These different scenarios generally produce significantly different physical forces on the train, affecting the likelihood of derailment (Cherchas et al., 1982). Chadwick (2017) found that VST collisions were disproportionately more likely to result in derailment.

In addition to different direction of the physical forces exerted upon the trains in these scenarios, there is also an expected difference in the speed of the road vehicle at the time of the collision. For instance, in 13 collisions which involved a heavy vehicle stopping foul of the level crossing, the road vehicle was stationary. In many of these collisions, the train drivers identified the stationary vehicle and commenced braking, and the train speed was thus also reduced.

The national rail safety database did not include information about whether road vehicle struck, or was struck by, the train. Of the 49 collisions between trains and heavy vehicles reviewed in this study, there were at least 39 TSV collisions of which 7 (17.9%) resulted in derailments. Of 7 VST collisions, 4 (57.1%) resulted in derailment. Fisher’s exact test showed a significant association between collision sequence and derailment (one-tailed p= 0.046), such that VST collisions were more likely to lead to derailment.

Of 4,886 collisions involving heavy vehicles recorded in the US FRA dataset there were 4,354 TSV collisions, of which 110 (2.3%) resulted in derailment. Of 532 VST collisions, 31 (5.8%) resulted in derailment. Fisher’s exact test showed that VST collisions were significantly more likely to result in derailment (one-tailed p<.001).

Analysis of heavy vehicle level crossing collisions in Australia

This section provides an overview of the characteristics of the 49 heavy vehicle accidents in Australia between July 2014 and August 2022. This section first describes the actions of heavy vehicle drivers associated with the level crossing collisions. The analysis then focuses on some of the factors identified in the study which may have contributed to the actions of the heavy vehicle drivers.

Actions of heavy vehicle drivers

The ATSB reviewed various sources of evidence to determine the circumstances which led to the road user entering the level crossing and not giving way to the train. There were 3 main types of road vehicle actions which precipitated level crossing collisions:

Unsurprisingly, the different actions associated with level crossing collisions varied depending on the type of level crossing protection equipment in use. There was one collision where there was insufficient information to determine the action of the road vehicle driver. The number of collisions associated with each action and the distribution by crossing type is shown in Table 15Table 15: Heavy vehicle driver actions preceding level crossing collisions, by level crossing primary protection type

Collisions at boom gate crossings

All 11 collisions at crossings protected by boom gates involved the road user stopping foul of the crossing. In at least one of these collisions, the driver was unintentionally foul of the crossing due to misjudging the length of their vehicle relative to the available distance on the far side of the crossing.

In 3 collisions, records indicated that the heavy vehicle driver had been unable to exit the level crossing due to unanticipated events. In 2 of these collisions, the heavy vehicle was involved in a collision or near collision with another road vehicle and became stuck. In the other collision, the heavy vehicle reportedly entered the crossing while queueing for an adjoining road, but encountered a mechanical problem and was unable to move from the crossing.

Collisions at flashing light crossings

Most collisions at crossings protected by flashing light signals (8 of 12) involved circumstances where the heavy vehicle driver did not stop before entering the crossing. In 6 of these cases, the heavy vehicle driver observed the flashing lights late on the approach and braked and skidded into the crossing. This meant that at the time of collision, most of the road vehicles were either travelling at a very slow speed or had in fact come to a stop foul of the crossing. In one of the cases, the driver detected the crossing and braked, however records indicated that the vehicle had defective brakes.

In 3 of the 12 collisions the heavy vehicle driver stopped at the crossing and inadvertently proceeded into the path of a train. Various unique factors contributed to each of these collisions:

• In one accident the stop lines at the level crossing were placed too close to the primary level crossing flashing light assembly. This meant that the driver of a passenger bus, who was seated very close to the front of the vehicle, was unable to see the flashing level crossing lights. In that instance, the level crossing also included a secondary flashing light assembly at far side of the crossing, however the bus driver did not look at those lights and inadvertently proceeded into the crossing just before the arrival of a train.
• In another accident, the truck driver heard the train driver sound the train horn and mistakenly thought this was an invitation to enter the crossing ahead of the train.
• In another accident, the truck driver was distracted by something related to the performance of their vehicle and did not pay attention to the level crossing.

One collision involved a heavy vehicle stopping on the crossing (prior to the activation of the flashing lights) and being unable to clear the crossing prior to the arrival of a train due to another heavy vehicle being stuck on the road ahead.

Collisions at Stop sign crossings

In 11 instances, the heavy vehicle driver did not stop at the Stop sign crossing. In 3 cases, the driver braked too late and inadvertently entered the crossing, indicating the crossing was not detected in time to stop. In 4 collisions the drivers attempted to conduct a ‘rolling stop’[24] and thus intentionally did not stop prior to entering the crossing, and probably failed to detect the presence of a train. In one collision, the truck driver reportedly stated they had not stopped or looked for trains before proceeding through the crossing.

In 7 of the 19 collisions at Stop sign level crossings, the heavy vehicle driver stopped their vehicle at the crossing and inadvertently proceeded into the path of an oncoming train.

There was one reported collision where the heavy vehicle driver detected the train and mistakenly determined it was safe to enter the crossing. As such, the other accidents probably indicated problems with either the sighting at the level crossing, the drivers’ behaviour when looking for trains, or the conspicuity of the trains, all of which may have led to the heavy vehicle drivers entering the crossing without identifying a train was approaching.

Collisions at Give way sign crossings

In 5 of the 6 collisions at Give way crossings, the road vehicle driver either did not see the train until they were on the crossing or did not see the train at all. This included one collision where a truck driver did not stop at a level crossing ahead of an approaching train, causing the train drivers to apply emergency brakes. The driver then stopped at an intersection ahead of the crossing, while still foul of the train line.

Errors and violations associated with level crossing collisions

Inadvertent failure to detect level crossing equipment

Research and background

In order to give way to trains at level crossings, road vehicle drivers must first identify that they are approaching a level crossing. The level crossing safety system provides information to drivers about the presence of an upcoming level crossing using roadside signs and roadway pavement markings. In the case of active level crossings, the presence of the crossing is also indicated by boom-gates across the roadway and/or flashing light beacons.

Yeh and Multer (2008) note that some road vehicle drivers inadvertently miss cues relating to the presence of a crossing, reducing their ability to stop in time. The section Factors affecting heavy vehicle driver  provides discussion about some of the reasons for these errors.

Analysis of heavy vehicle level crossing collisions

As identified in Actions of heavy vehicle drivers there were 24 collisions in which the heavy vehicle driver did not stop prior to entering the level crossing. One plausible reason for these collisions was that the drivers did not detect they were approaching a level crossing until it was too late.

To identify the instances where level crossing equipment was not detected, records including police reports, train driver statements, rail infrastructure manager reports and ATSB investigation interviews were reviewed. The ATSB identified:

• There were no instances of heavy vehicle drivers not detecting the presence of a boom-gate controlled active level crossing.
• There were at least 6 collisions where the heavy vehicle driver did not detect that they were approaching an active flashing lights-controlled level crossing until it was too late to stop. In several of these instances, braking occurred before the crossing, however due to the speed and mass of the heavy vehicles there was insufficient distance for them to stop. In one example, an ATSB investigation identified that the road-train truck driver commenced braking at about 180 m from the crossing while travelling at 90 km/h but was unable to stop.
• There were at least 3 collisions at Stop sign level crossings where the heavy vehicle driver did not identify the level crossing until it was too late to stop. In one of these collisions, the heavy vehicle driver reported to police that they were completely unaware of the crossing, and a rail infrastructure manager report identified that sun glare obscured the driver’s vision of the crossing. In the other 2 collisions, the driver identified the level crossing too late to stop.
• There were 5 collisions where the heavy vehicle driver did not stop prior to entering a level crossing controlled by Give way signs. The road rules for Give way sign-controlled crossings do not require a vehicle driver to come to a complete stop, except to give way to trains if one is approaching the crossing. Due to this, the failure of the drivers to stop may have resulted from a failure to detect the presence of the level crossing, or a failure to detect the presence of trains. There was insufficient evidence to determine which of these ‘failures’ contributed to the 5 accidents at Give way sign crossings.

This analysis indicates that in a subset of collisions, heavy vehicle drivers did not detect the presence of the upcoming level crossing. While only a minority of instances, these collisions reflect a dangerous state in the level crossing system. If a driver does not identify the presence of a Passive control crossing, they will not be looking for trains, and there are limited redundant controls to alert them of the requirement to stop. AS1742.7 states that Stop sign crossings are required to be installed when visibility on the road approach to a crossing does not provide adequate visibility for a road vehicle driver to see an approaching train, and this poor visibility will further reduce the likelihood of the train being detected visually by opportunistic (rather than deliberative) scanning. The train horn may provide an aural indication of a requirement to stop, however there are known limitations with the audibility of train horns (see Train horn audibility).

Concerning collisions at active level crossings, the errors were mainly a failure to detect that the crossing was activated, rather than not detecting the crossing. This also presents a highly dangerous condition, since a driver approaching an active control level crossing and perceiving the crossing as not active (lights not flashing), will parse that information as indicating the crossing is safe and no train is approaching. This highlights the importance of ensuring that the signal produced by active level crossings are conspicuous.

Deliberate violation of level crossing rules

Research and background

Part of the safety system for level crossings are the procedural risk controls for level crossing users, including road rules applicable to road vehicle drivers. When road vehicle drivers intentionally disobey the requirements of road rules at level crossings, they greatly reduce the effectiveness of the safety system.

Research has identified that some road vehicle drivers deliberately violate road rules requiring them to stop at level crossings. For example, one study showed that of 22 drivers, 3 did not stop at rural Stop sign level crossings (Beanland and colleagues, 2017). These drivers noted trains were infrequent and suggested there was ample sight distance to look for trains (Stop sign crossings are often installed where there is not adequate distance to look for trains while approaching the crossing).

Gou and Bellavigna-Ladoux (2003) noted that a common form of deliberate non-compliance at level crossings is conducting a ‘rolling stop’. This involves slowing the vehicle until a decision is made to proceed into the crossing, without coming to a complete stop. When conducting a ‘rolling stop’ a road vehicle driver will necessarily spend less time at the stop point for a passive level crossing and therefore will probably conduct less time scanning for oncoming trains, increasing the likelihood of an incorrect decision to proceed into the crossing when it is not safe.

Heavy vehicle drivers may be more likely to engage in ‘rolling stop’ behaviour when approaching a Stop sign level crossing. As described in the ATSB investigation RO-2007-001, Level crossing collision at Back Creek, NSW:

The design standards for level crossings (Appendix B) include a variable for the acceleration capabilities of the design vehicle, and in doing so provide allowance for the increased time required by heavy vehicle drivers to depress and release clutch and engage the required gears to proceed from a stop and through the crossing. This may not, however, translate to heavy vehicle drivers always ensuring their vehicles come to a stop at Stop sign controlled crossings, since they may seek to avoid placing stresses on drivetrain components or simply prefer to maintain forward movement.

Another common violation occurs when road vehicle drivers enter a level crossing and stop foul of the rail tracks. A common scenario is the driver entering the crossing while the road ahead was blocked by other traffic, mistakenly anticipating traffic ahead would clear and they will be able to proceed through. This is contrary to road traffic laws. For example, the Australian road rules (model law) states:

These violations have been attributed to various factors; road vehicle drivers may perceive there to be a low level of risk associated with entering a crossing and waiting for traffic to clear. These actions may also be influenced by social pressure and perceived social norms (Yeh and Multer, 2008). In one study, 30% of drivers reported that they would violate boom gate crossing rules if they did not detect a train and they saw another driver do so (Witte and Donohue, 2000).

Analysis of heavy vehicle level crossing collisions

In at least 4 collisions, the heavy vehicle drivers engaged in deliberate non-compliance with level crossing rules for Stop signs by conducting a ‘rolling stop’ on approach to the crossing. In another collision, the heavy vehicle driver did not stop at a Stop sign and did not look for trains. These actions increased the risk of collision by reducing the likelihood that the drivers would identify the presence of trains.

There were at least 6 collisions where the heavy vehicle driver intentionally entered the level crossing when traffic or another blockage prevented them from proceeding clear of the train lines. This did not include instances where the driver mistakenly thought their vehicle was clear of the crossing, or instances where the driver had an accident or breakdown on the crossing.

In an additional 3 collisions, the heavy vehicle driver intentionally entered and stopped foul of the crossing, but did not exit the crossing in time to prevent a collision. In these 3 accidents there was no evidence of a blockage preventing the heavy vehicle from clearing the crossing and were considered intentional non-compliance with level crossing rules for the purpose of this analysis.

In addition, there were 2 collisions where the driver of the heavy vehicle was engaged in high-risk behaviour prior to unintentionally entering the level crossing or unintentionally stopping foul of the crossing. In one instance, the driver was distracted by their mobile phone, and in another the driver was found to be drug affected. These accidents were not considered intentional violations of level crossing rules in this analysis; these drivers did not intentionally enter the level crossing.

Failure to detect trains at passive control crossings

Research and background

At passive control crossings, road vehicle drivers must visually search for and detect the presence of a train to determine when they can and cannot proceed through a crossing. There are no recovery controls to prevent a collision if a road vehicle driver does not detect that a train is approaching.

Road vehicle drivers sometimes do not look for oncoming trains when approaching a passive control crossing. Research using head movement measurements of heavy vehicle drivers found that less than 60% looked in at least one direction of the train line when approaching a passive control level crossing (Ngamdung and da Silva, 2012). Failure to look for trains may be associated with a number of factors, including not identifying the presence of a crossing, distraction or preoccupation, or a low expectancy of the presence of a train at the crossing.

Even when road vehicle drivers look in the direction of approaching train, they may not detect the train nor identify a requirement to stop. Rudin-Brown and colleagues (2014) identify that many level crossing collisions result from so-called ‘Looked But Failed to See’ errors.

Analysis of heavy vehicle level crossing collisions

As identified in Actions of heavy vehicle drivers there were 26 collisions at passive control level crossings. Of these collisions, at least 2 involved the heavy vehicle driver detecting the train before entering the crossing:

• In one instance, the driver of the heavy vehicle stopped for a Stop sign level crossing and identified that a train was nearby. The train’s lead locomotive cab was unoccupied, and was conducting an indexing movement across the crossing at low speed. The truck driver incorrectly determined that they could transit the crossing safely, and the train scraped the side of the truck prior to it clearing the crossing.
• In one instance, the driver approached a Stop sign level crossing and, according to the rail operator report for this collision, the truck was unable to stop for the crossing due to a mechanical problem.

Concerning the other collisions at passive control level crossings, there was limited information to determine whether the heavy vehicle driver looked for trains and detected the presence of a train before proceeding into the crossing. Heavy vehicle drivers were almost never interviewed, except in collisions which were investigated by the ATSB. Driver statements were sometimes included in police reports; however drivers may be disinclined to inform the police if they had not looked for trains.

While there was very limited direct evidence concerning whether a heavy vehicle driver looked for trains, the ATSB considered that an indirect indicator of heavy vehicle drivers attempting to comply with the requirement to look for and give way to trains was whether the heavy vehicle slowed or stopped during the approach to the crossing. However, it is acknowledged that this indirect evidence cannot rule out the possibility that some drivers may have slowed when approaching the level crossing due to other factors, such as the road condition near the crossing, or to accommodate the type and weight of their load.

The ATSB reviewed rail operator, rail infrastructure manager and police reports to classify the braking behaviours of heavy vehicles approaching passive control crossings.

• In at least 12 collisions, the driver either slowed or stopped while approaching passive control crossings but either did not detect the train, or deliberately entered the crossing ahead of the train.
• In at least 4 collisions, the driver did not slow or stop for the crossing until too late. In one of these collisions, the ATSB report found the driver did not approach the crossing with sufficient caution to stop after noticing the train. In another collision, the rail infrastructure manager report stated the vehicle was travelling too fast to stop at the crossing. In another collision, the driver stated they did not detect the presence of the crossing, and only commenced braking after seeing the train shortly before the collision.
• In one collision, the rail operator report stated that the heavy vehicle driver reported that they had not slowed or looked for trains at the Stop sign crossing.
• In 7 collisions, there was insufficient evidence available to determine whether the heavy vehicle slowed on approach to the passive control level crossing.

The analysis therefore identifies that in at least 46% (12 of 26) of collisions the observed braking behaviour indicated that the heavy vehicle drivers made some attempt to comply with the requirement to look for and give way to trains, however probably did not detect the presence of the train.

An alternative explanation in these cases was that some the heavy vehicle drivers who slowed or stopped when approaching passive control level crossings did identify the presence of a train, but elected to enter the crossing, in an attempt to ‘beat the train’ through the crossing. There was no direct evidence of any heavy vehicle drivers deliberately proceeding in front of an approaching train, although drivers may be unlikely to make admissions of such behaviours to police or to rail operator investigations. In 2 cases, the heavy vehicle driver was fatally injured in the accident and there was limited information concerning whether they had observed the train prior to entering the crossing. In 4 cases, the heavy vehicle drivers stated that they had looked for trains prior to entering the crossing. In other cases, other circumstances related to the accident indicated it was unlikely the heavy vehicle driver had deliberately entered the crossing. Considering all the available evidence of the 12 cases where heavy vehicle drivers slowed or stopped while approaching passive control crossings, the ATSB determined the drivers had probably not detected the presence of a train.

Factors affecting heavy vehicle driver behaviour at level crossings

Overview and prior research

The analysis of level crossing collision data showed that collisions typically occurred in fine weather, in daylight, on dry, sealed roads. Previous analysis found that most level crossing accidents did not involve alcohol or excessive speed (ATSB, 2002), which are common causes for other road accidents (OECD, 2021).

The majority of collisions probably involved some form of unintentional lapse, where the heavy vehicle driver did not identify a train was approaching the crossing. Even in the subset of collisions where a deliberate violation of road rules was identified, the heavy vehicle driver probably intended and expected to clear the crossing safely.

The safety systems approach to accident analysis seeks to understand the contextual factors associated with individual errors and violations. This involves identifying factors that exist within the system which increase the likelihood of errors and other failures (latent conditions). For the level crossing safety system, this includes considering how the experiences and tasks of road users shape their performance at level crossings. It also includes considering the design of vehicles, trains and level crossing equipment, and whether these support the tasks of the human operators at level crossings (road vehicle and train drivers).

Familiarity and expectancy

Research

When drivers become familiar with a particular crossing or a particular type of crossing, and when the driver has previously not observed trains at that crossing (or type of crossing), an unconscious expectancy for no trains at that crossing or that type of crossing may form (Rudin-Brown and colleagues, 2014). Researchers have argued that drivers generally expect there to be no trains present at level crossings (Eck, 2002), and that these expectations are ‘the greatest challenge’ in overcoming safety problems at passive level crossings (Salmon and others, 2013) and the ‘root of unintentional noncompliance at…level crossing’ (Eck, 2002).

Research has shown that drivers familiar with a level crossing are more frequently involved in level crossing collisions than drivers unfamiliar with an area (Abraham and colleagues, 1998). A coronial investigation into 12 fatal level crossing collisions in Victoria found that all except 3 drivers were very familiar with the crossing and all but one never or rarely saw a train at the level crossing (Coroner’s Court of Victoria, 2013).

There are several ways that this low expectancy of trains can contribute to level crossing collisions. Low expectancy for trains may influence deliberative, planned behaviour where a driver enters a level crossing contrary to the requirements of road rules. Reason (1990) identifies that routine violations of rules are often the product of environments where violations are infrequently punished, which may be taken to mean formal punishment in the case of law enforcement, or other negative consequences such as collisions. As Yeh and Multer (2008) summarise, a low expectancy and low frequency of trains at level crossings may cause drivers to simply disregard crossing protection equipment.

The other way low expectancy for trains can influence level crossing collisions is at the unconscious level, affecting how road vehicle drivers scan for and detect important information about the crossing and the presence of trains. Reason (1990) identifies that people engaged in well-practiced, routine tasks are vulnerable to unintentional slips of attention. In these contexts, attention and behaviour is highly influenced by the ‘motor schema’ or ‘script’ for the task, and there is a high likelihood that an operator will miss cues which indicate a need to divert from the typical routine.

The task of driving a road vehicle is generally highly-practiced, and occurs in roadway environments where other demands and distractions are often prevalent. In this context, road vehicle drivers are required to depart from the active and dominant motor schema of continuing to drive along the road, and adopt a motor schema of looking and preparing to stop for trains. For motorists with a low expectancy for encountering a train at a crossing, there is a high likelihood that they may miss cues which identify they need to stop for trains.  

Other psychological theories identify that experiences form the basis for ‘schemas’ or ‘mental models’ of the task environment. These mental models influence where a person will search for information, what they are looking for, and which information they will detect (Wickens and McCarley 2008; Wickens and others 2013). Drivers with a low expectancy of trains may therefore be less likely to look for trains, or look for activated flashing lights. Even in situations where a driver looks directly at a train or flashing light assembly, they may not detect the requirement to stop and proceed into the crossing, reflecting a so-called ‘looked-but-failed-to-see’ error.

Analysis of heavy vehicle level crossing collisions

As described in Errors and violations associated with level crossing, 9 accidents were due to the heavy vehicle driver intentionally stopping foul of a level crossing. There was insufficient evidence to establish whether any of these drivers were familiar with the level crossing, or what expectancy they had when entering the crossing. It is possible, however, that some or all these drivers previously queued over the same crossing or other intersections, with no negative consequences. It is also possible that these drivers had observed other drivers queueing over level crossings, normalising the behaviour and reducing the expectancy of negative consequences.

There were 12 collisions where the ATSB identified that the heavy vehicle driver was familiar with the level crossing prior to the collision with the train, and all of these involved the driver either not stopping at the crossing or stopping and proceeding into the path of the train. There were 6 collisions where the heavy vehicle driver was familiar with a passive control level crossing and did not detect the presence of the train. In 5 collisions, the heavy vehicle driver was familiar with a crossing protected by flashing lights and did not detect that the lights were activated.

The heavy vehicle drivers were typically engaged in driving for business or employment, and the collisions often occurred on routes they traversed several times a day or week. Some examples included:

• One driver reported that they travelled across the same level crossing about 20 times per day for work, and was not expecting a train.
• One driver reported they had been driving the same route for 2 months and the train they collided with was the first train they had seen at the level crossing.

It was not possible to precisely describe the effects of familiarity for each of these 12 collisions. However, it is highly likely that in some collisions the drivers’ familiarity with the level crossings and low expectancy for encountering trains affected their attention to and perception of the crossing environment. With a low expectancy of trains, the drivers may have either not allocated sufficient attention to the crossing or have looked but not detected a requirement to stop.

In one collision the heavy vehicle driver entered a level crossing after identifying that the level crossing lights were flashing and detecting the train. The driver heard the train driver sound the train horn and mistakenly thought this was an invitation to enter the crossing ahead of the train. They reported to police that they had previously been ‘let through’ the crossing in similar circumstances, and therefore had an expectancy that they could cross ahead of the train.

Distraction and divided attention

Research

Distraction occurs when a driver’s attention (meaning where the driver is looking, what they are manipulating/touching, and what they are thinking about) is diverted from activities critical for safe driving towards a competing activity (Parnell and others, 2016). There is a significant body of evidence demonstrating that distraction impairs driving performance and contributes to accidents (Young and others, 2007).

The task of driving a road vehicle often requires attending to complex road environments. Due to a limited ability to attend to multiple sources of information at once, road vehicle drivers operating in complex or cluttered areas (such as where there are multiple intersections, signs, or road users) may be less likely to attend to and detect trains (Rudin-Brown and others, 2014).

Reviews of level crossing collisions have identified that distraction can be a contributory factor to driver non-compliance (Rudin-Brown and others, 2014). In-cab studies of heavy vehicle drivers have shown that drivers sometimes attend to other tasks (including texting or eating) while negotiating level crossings (Ngamdung and da Silva, 2012), or may be distracted by factors at the crossing such as poor surface conditions (Eck, 2002).

Analysis of heavy vehicle level crossing collisions

Records indicated that at least 6 of the heavy vehicle drivers were probably distracted prior to the collision, or were dividing their attention between driving the vehicle and another task or thought. In some of these instances, the driver was reportedly focussed on other traffic or complicated road signage, and did not detect the level crossing flashing lights or the approaching train. In one example, an ATSB investigation obtained in-vehicle video footage which showed the truck driver was distracted by re-affixing a mobile phone mount which had fallen from the truck windscreen. The truck driver did not notice the flashing active crossing warning lights and proceeded into the crossing, colliding with a freight train.

Obstructed vision due to vegetation, sun glare and poor crossing lighting

Research

For a road vehicle driver to detect an oncoming train at a passive level crossing, the train needs to be visible within their field of view. In some instances, however, the design and maintenance of level crossings is such that the driver’s view is obstructed.

Another visual problem encountered by road vehicle drivers approaching a level crossing is disabling sun glare. Disability glare occurs when light enters the eye and ‘washes out’ the image being perceived (Sanders and McCormick, 1993). The disabling effect of glare is greatest when the sun is in the direct line of sight of the driver. Thus, sun glare is greatest at the times just before sunset and just after sunrise, when the sun is low in the sky and thus objects which are being viewed at ground level fall in a narrower angle relative to the sun.

After sunset, night-time darkness affects the ability of drivers to safely negotiate level crossings by reducing the availability of visual cues, particularly for passive control crossings. Where level crossings are unlit, level crossing signage may be difficult to detect. Yeh and Multer (2008) note that illumination of level crossings improves the detection of the crossings at night.

Analysis of heavy vehicle level crossing collisions

In at least 9 collisions reviewed by this study, the view of the heavy vehicle driver was probably obstructed by nearby vegetation. In at least 8 instances, vegetation probably obstructed visibility of the rail track, which may have reduced the ability of the driver to detect oncoming trains. In one instance, vegetation obstructed the visibility of the rail track and the level crossing protection equipment.

In one collision, the rail infrastructure manager and police reports identified that the driver of the heavy vehicle was probably unable to detect the presence of the level crossing due to sun glare.

In another collision, the rail infrastructure manager identified that a Stop sign was located in a dark industrial area, with poor lighting on the crossing itself and contrasting bright lights from surrounding industrial installations and other road traffic. The infrastructure manager report found that when a heavy vehicle approached the crossing at night, the poor lighting conditions reduced the driver’s ability to detect an approaching train.

Obstructed visibility due to vehicle design

Context

In addition to environmental obstructions, the visibility of trains and level crossing equipment may be affected by vehicle design characteristics. Heavy vehicles often have large ‘A’ and ‘B pillars, planar mirrors, and other structures like exhaust snorkels. Research has shown that restricted fields of view in heavy vehicles are a common cause of other road accidents (Blower, 2007; Niewoehner and Berg, 2005).

The Australian Design Rule 93/00 – Forward Field of View specified the allowable restrictions within the drivers’ forward 180° field of view for passenger and heavy vehicles supplied to the Australian market.[25] The regulations did not prescribe any requirements for field of view behind the driver.

The design standards for level crossings (Australian Standard AS1724.7:2016) required that the sighting opportunity available to road users is provided at visual angles which do not require excessive head movement or sight obstruction by the vehicle itself. The standard states that for the sighting distance when stopped at a level crossing, the maximum allowed viewing angles are 110° to the left and 140° right.

The combination of vehicle design which does not prohibit restricted visibility behind the driver seated position, and crossing design which allows for rail traffic to approach from obtuse angles (greater than 90°), leads to a potential zone of restricted visibility where the driver may need to lean forward to view along the train track.

Analysis of heavy vehicle level crossing collisions

In at least 5 collisions reviewed in this safety study, the design of the heavy vehicle cab probably restricted the driver’s view of the crossing equipment or train. These included 3 collisions which had previously been investigated by the ATSB:

• RO-2015-016: The ATSB found that the truck ‘A’ pillar and air snorkel probably restricted the driver’s view of the flashing lights as the truck approached the crossing.
• RO-2016-009: The ATSB found that the truck driver’s view of the track when looking left was restricted by the structure of the cab and the absence of windows behind the driver. When seated in an upright position, the viewable area was limited to 90°, meaning only 29 m of track was visible. This could be extended by up to 104° (60 m) by hunching forward.
• RO-2017-011: The ATSB found that the truck driver would not have been able to see the approaching train from their stopped position, due to their line of sight being obscured by the truck’s B-pillar.

In another collision, the shallow-nosed design of the bus cab meant that when the driver was stopped at the stop line, they were very close to the stop line and had a restricted view of the flashing crossing lights on the near-side of the crossing, contributing to them entering the crossing into the path of a train. In the other collision, the driver identified that large air cleaners on their vehicle may have obstructed vision of the train at the passive control level crossing.

This analysis demonstrated that the visibility of some heavy vehicle drivers was restricted by objects around the level crossings, the design of the heavy vehicle and the crossing, and the environmental conditions at the time of the collisions.

Train horn audibility

Australian Standards and other requirements

Locomotives operating in Australia must be equipped with train horns, which are used to ensure individuals in and around the rail corridor are aware that rail traffic is approaching. The Rail Industry Safety Standards Board (RISSB) Code of Practice for Rail Traffic Horn Use stated that:

The Australian Standard for Railway Rolling Stock Audible Warning Devices (AS 7532:2016) described requirements for the audibility of train horns:

• The country warning device must produce at least 88 dB at 200 m ahead of the rolling stock, and at least 106 dB at 30 m ahead of the rolling stock.
• The town warning device must produce at least 90 dB at 100 m from the rolling stock and 101 dB at 30 m from the rolling stock.

AS 7532:2016 did not specify the types of train horns accepted for use, or the frequencies at which the train horn must produce sound.

The RISSB Code of Practice for Rail Traffic Horn Use stated that activating the train horn was a requirement at passive control level crossings, and optional at active control crossings. The code of practice also required that the train horn must be sounded at the whistle board, where placed, and may be sounded at other locations including during the approach to and on the crossing.

Research on effectiveness of train horns

Research examining level crossing collisions in the US between December 1995 and August 1996 found that 55 of 60 train drivers sounded the locomotive horn prior to the collision, but only 4 of 14 road vehicle drivers reported hearing the horn (described in Yeh and Multer, 2008). In a 2013 coronial review of 12 fatal level crossing accidents in Victoria, the Coroner noted that none of the drivers had heard the locomotive horn (Coroner’s Court of Victoria, 2013). These reviews highlight that in many collisions it is evident that train horns were unreliable for alerting road users to the presence of a train.

Supporting the utility of train horns, Yeh and Multer (2008) identified that initiatives prohibiting the use of train horns (whistle bans) in US jurisdictions have led to increased rates of level crossing collisions. For example, they report analysis conducted by the FRA in 1995 was reported:

Follow-up analysis conducted by the FRA (2000) found that:

Several laboratory studies have investigated the acoustic qualities of train horns, and the audibility of train horn sound in simulated conditions. Findings from this research include:

• Different types of train horns produce different patterns of sound. 5-chime horns were found to produce a much greater volume of sound at higher frequencies (above 800 Hz) than 3-chime horns (Rapoza and Raslear, 2001).
• Train horn sound is attenuated as a function of the distance of the train from the vehicle, consistent with the inverse square law for sound. Because a vehicle travelling at a greater speed will need to detect the train from further away in order to stop, the train horn must be proportionally louder (proportional to the square of the distance between the train and road vehicle).
• The level of sound available to road vehicle drivers is affected by the sound-attenuating properties of the vehicle body, referred to as insertion loss.
• Train horn sound is also attenuated by whether the road vehicle windows are open or closed, whether the engine is running, and whether the radio and/or fan is on (Rapoza and Raslear, 2001; Casali and others 2002; Dolan and Rainey, 2005).

There were mixed findings regarding the detectability of train horns.

• Raslear and Rapoza (2001) used a probability model to predict the likelihood that road vehicle drivers would detect an approaching train in time to prevent collision. The model predicted that a road vehicle driver approaching a passive control level crossing would almost always detect a 5-chime warning horn. However, a 3-chime warning horn would only be detectable 75% of the time when the train was travelling at 32 km/h, dropping to 25% of the time when the train was travelling at 64 km/h.
• Dolan and Rainey (2005) conducted a laboratory study which involved producing simulated train horn sounds at different volumes, along with simulated background noises to replicate different vehicle conditions. They found that 50% of participants could detect the simulated train horn sound at a volume of 10 dB below the simulated background noise. One limitation of the applicability of this research is that the participants were not driving a vehicle and were primed to listen for the train horn signal.
• Casali and others (2002) measured the sound produced by a 5-chime train horn, and the sound attenuating properties of a sample of 1990s model trucks under different operational conditions (engine on/off, windows up/down, radio on/off). This research concluded that the noise produced by the train horn only exceeded the ‘masked threshold’ (being 13 dB louder than background noise) when the engine was at idle and the windows were down.

In summary, there is no known research which has examined how frequently real drivers respond to train horns in real driving conditions. Consolidation of the research indicates that the requirement to use a train horn improves the safety of level crossings overall, however collision case-studies indicate that there are numerous instances where the train horn is not effective for alerting drivers to the presence of a train. There is a research gap, therefore, in identifying the circumstances in which train horns are not effective, and how often train horns are not effective for preventing collisions.

Analysis of heavy vehicle level crossing collisions

Records indicated that in at least 44 of the 49 level crossing collisions reviewed in the study the train crew sounded the horn as required prior to the collision. There were at 2 instances where the horn was not sounded as required at the whistle board, and 3 instances where there was insufficient information to determine whether the horn had been sounded.

Considering the effectiveness of the train horn for alerting heavy vehicle drivers to the presence of trains, the ATSB noted that:

• There were 13 collisions which involved the heavy vehicle driver stopping foul of the crossing. In most of these, the heavy vehicle driver was not able to exit the crossing due to being blocked by traffic or another form of impediment, and thus detecting the train via the train horn was unlikely to have prevented these collisions. In 3 of the collisions the ATSB did not identify evidence of a blockage preventing the heavy vehicle from clearing the crossing.
• In 3 collisions, the heavy vehicle driver had identified the presence of a train but did not stop. These included 2 collisions where the heavy vehicle driver encountered a mechanical problem and was unable to stop, and one collision where the truck driver heard the train horn and mistakenly interpreted this to be permission to entering the crossing. The detectability of the train horn was not considered to have affected these collisions.
• There were 3 additional collisions where there was insufficient information to determine whether the vehicle driver detected the presence of the train.

Excluding these instances, the ATSB identified 25 collisions where the train horn was sounded and not detected by the heavy vehicle driver. The heavy vehicle drivers, not detecting the presence of a train, proceeded into the path of the train resulting in collision. It is possible that had the train horn, or another warning, been more conspicuous to the heavy vehicle drivers, then these 25 collisions may not have occurred.

The safety study reviewed collisions only, and did not examine the number of near misses and other instances where heavy vehicle drivers were alerted to the presence of the train by the train horn.

Other audible warning devices

Australian Standard AS1742.7 (Railway level crossings) states that flashing light signals may include the provision of audible warning devices. Further, the standard requires that where active control of pedestrian traffic is provided (red flashing man signal), an audible signal shall be provided. The safety study did not analyse the effectiveness of other audible level crossing warning devices, such as audible ‘bells’ provided at some level crossings. The ATSB observes, however, that such audible warnings are produced at a much lower volume than locomotive horns, and thus are likely to be less audible to approaching road vehicles. Audible level crossing signals are primarily engineered to be effective for alerting pedestrians.

Visual conspicuity of trains

Research

Due to the safety system for passive control level crossings relying on road vehicle drivers visually detecting the presence of a train in order to identify a requirement to stop and give way, there has been substantial interest in the extent to which the visual characteristics of trains make them difficult to detect. The concept of conspicuity refers to the material characteristics of an object which affect the likelihood it will be detected. The greatest determinant of visual conspicuity is the level of contrast between the object and its surrounding environment.

Trains are often constructed with dark colours, and may be poorly lit, reducing their conspicuity for road vehicle drivers, particularly at night (Rudin-Brown and others., 2014; Yeh and Multer, 2008). A review conducted by the Australian Centre for Rail Innovation (ACRI) identified Australian case studies where factors such as dull colour schemes, dirty train exterior and low contrast between the train colour and the surrounding environment may have reduced the conspicuity of freight trains (ACRI, 2022). The ACRI freight train visibility review identified 30 opportunities to increase the conspicuity of trains, of which 2 were carried forward to field testing (flashing train beacons and conversion of headlights from halogen to LED). The results of this testing showed that LED conversion of locomotive headlights resulted in improved visibility in misty conditions only, with insignificant improvement in clear weather conditions. The beacon light was only effective at improving visibility in situations where the train and observer were close to the crossing, and the level crossing angle was obtuse (between 90° and 180°) (Kassa, Wan and White, 2023). Other research has examined the effects of some common treatments for improving train conspicuity:

• Studies have shown that trains equipped with auxiliary alerting lights (additional to the train headlight) are detected at greater distances (Carroll et al 1995).
• Trains which have reflectorised markings installed on freight cars (wagons) are detected at greater distances than unmarked cars (see Edquist and others (2009), for a review).
• Several studies have examined the benefits of strobe lighting on trains, however only small effects have been observed (Cairney, 2003).

In addition to poor illumination, the visual image of a freight train available to a road vehicle driver approaching a level crossing may provide a low likelihood of detection. When a driver is approaching a level crossing and attending to the road, an approaching train will be in the driver’s peripheral field of view. There will often be no visual cues indicating the presence of a train in the driver’s central (foveal) field of view, where detection of objects is most likely (Edquist and others, 2009; Yeh and Multer, 2008).

The human perceptual system is particularly adapted to detecting movement, and vision in the peripheral field of view is particularly sensitive to movement. Unfortunately, due to the geometry of a collision, a train will typically present as an unchanging retinal image to the driver. That is, the train will not appear to move across the driver’s retina when it is on a collision course with the road vehicle driver (Rudin-Brown and others, 2014). These perceptual challenges are similar to those described in the ATSB report

Relevant industry standards

Australian Standard AS 7531:2015 Lighting and Visibility described the requirements for rolling stock lighting and visibility. This standard required that all locomotives and self-propelled passenger rolling stock have at least one white headlight on each leading end. For new and modified rolling stock, the standard required locomotives and self-propelled rolling stock have active visibility lights, which were required to flash for at least 15 seconds after the horn is sounded. The standard stated that the primary purpose of these lights was to enhance the visibility of the train from the perspective of a driver of a road vehicle approaching a crossing.

AS7531:2015 also required all new and modified locomotives and lead vehicles of passenger trains to have high-visibility colour livery applied. Under this requirement, at least one square metre of the forward-facing area of locomotives and passenger rolling stock was required to be yellow, orange, red or white.

In December 2023 (after the study period of 1 July 2014 to 31 August 2022), the Rail Industry Safety and Standards Board approved AS7531:2023. This version of the standard included additional rolling stock lighting and visibility design principles, and provided additional requirements for high-visibility livery. The lighting and visibility design principles added to the standards in the 2023 revision included the following:

In January 2024, ONRSR announced that it would develop a Code of Practice for train visibility in Australia, with a final draft expected by mid-2024. The intent of this code was to ‘assist rail transport operators to strengthen the overall safety management systems that underpin their operations where trains interact with people, drivers and vehicles - with an emphasis on risk controls for train visibility’. ONRSR stated that it would consider the Code of Practice when conducting compliance activities, and that the code would be admissible to proceedings related to compliance with the Rail Safety National law.

Analysis of heavy vehicle level crossing collisions

The ATSB reviewed operator reports and other records to identify how conspicuity equipment such as headlights, visibility (ditch) lights and livery had been utilised for the trains involved in the 49 collisions with heavy vehicles reviewed in this study. Overall, this information was reported infrequently, with key observations being:

• In 24 collisions, the records indicated that the train headlight was on prior to the collision with the vehicle, while in one collision, records indicated the headlight was off. In 24 collisions there was insufficient evidence to establish status of the headlight.
• For 8 collisions, records indicated that the train had visibility (ditch) lights installed, whereas in 4 collisions the train did not have visibility lights. There were 37 collisions where information about the train visibility lights was not reported.
• In 10 collisions, operators reported on the installation of reflective delineators. Four trains had reflective delineators installed on rollingstock, and 6 did not.

Of the 26 collisions at passive control level crossings, 2 occurred at night and in both of these cases the train had its headlights activated. In one of these collisions, the train was being operated in the reverse direction, meaning the train headlights were not effective at illuminating the train in the direction of travel towards the crossing.

In summary, due to limitations in available data from operator reports and other documents, this study was not able to evaluate whether trains involved in level crossing collisions were equipped with the lighting and other visibility equipment required by Australian standards. Furthermore, since this study only examined trains which were involved in collisions, it was not possible to identify factors associated with greater of poorer conspicuity, or for that matter the effectiveness of different types of equipment used to enhance train conspicuity. This is discussed further in Study limitations.

Other observations which may be useful for further analysis of train conspicuity include:

• Very few of the collisions reviewed in this study occurred at night, when problems associated with conspicuity are probably greatest.
• The majority of collisions (at least 39 of 49 collisions) involved the train striking the heavy vehicle, indicating that problems associated with train conspicuity are more likely to involve the conspicuity of the locomotive, rather than the other rollingstock.

Analysis of level crossing safety systems

Curved road approaches

Standards and other guidance relevant to placement of crossings on curved road approaches

Australian Standard for level crossing design

The criteria for the placement of level crossing protection equipment were described in Australian Standard AS 1742.7:2016 Manual of uniform traffic control devices. AS 1742.7 stated that level crossings:

The standard required that if a level crossing was located on a curved road, then appropriate curve signs were to be used, or alternatively additional level crossing warning signage may be placed along the curve. The standard further stated that:

AS 1742.7 sought to ensure that the placement of the primary active level crossing control device (such as the flashing light assembly) provided sufficient time and distance for the road user to identify a crossing ahead and stop after identifying the control device is activated. This distance was referred to as S₁ or the safe stopping distance (SSD).

Sighting distances described in AS 1742.7, and shown in Figure 3, were defined as:

• S₁, the minimum road distance which must be available to the road design vehicle, at the point the driver is able to detect a requirement to stop prior to the nearest rail.
• S₂, the minimum distance of a train from the crossing at which a road vehicle driver at distance S₁ from the crossing can proceed at speed and safely clear the crossing ahead of the train.
• S₃, the minimum distance of an approaching train from the centre of the crossing, when the driver of the road vehicle, stopped at the crossing, must first see an approaching train in order to safety cross the tracks.
• X_L, sighting angle (left).

Figure 3: Sighting distances for level crossings described in Australian Standard AS 1742.7:2016

Source: Image created by ATSB, illustrating concepts described in AS 1742.7:2016

AS 1742.7 provided guidance ‘in order to ensure that a motor vehicle driver can see along the prescribed sight triangles without excessive head movement’. This was achieved through the provision of maximum sighting angles for the S₂ and S₃ distances.

AS 1742.7 provided instructions for calculating S₁ at each crossing, considering various factors related to the expected road user and the road construction (the formula for calculating S₁ is provided in Appendix B). The standard did not require additional S₁ distance for crossings placed on or after curved road approaches.

Procedural guidance for level crossing sighting surveys

Level crossing sighting surveys are conducted to determine if level crossings comply with requirements including those described in AS 1742.7. The Australian Level Crossing Assessment Model (ALCAM) Level Crossing Assessment Handbook was produced by the National ALCAM Committee as a guide for trained surveyors, providing methods for carrying out ALCAM surveys of level crossings in Australia and New Zealand. These procedures stated that to measure the available sighting distance for primary level crossing protection equipment (S₁), surveyors should:

• Drive all approach roads in the vicinity of the crossing to obtain the maximum sighting distance when the crossing control can be clearly seen, and record the available sighting distance available for each approach.
• Calculate the required safe stopping distance position for the roads passing through the crossing. If there are side roads turning towards the crossing or if the crossing is approached from a nearby intersection, safe stopping distances for these roads were also to be determined.
• Compare the available and required (calculated) safe stopping distances.

The handbook also identified that additional warning signs may be installed where the road approach to level crossings is curved, observing that ‘duplicated advance warning signs are generally placed on curved road approaches where visibility to the left-hand sign assembly is restricted by roadside vegetation, terrain or structures’.

Level crossing collisions

Of 6 collisions where the road driver braked too late to stop at a flashing light level crossing, 5 occurred at crossings on or following a curved road approach. All 5 were right curves. None of these collisions occurred at crossings with active advanced flashing lights (RX-11) installed. Three of the collisions had previously been investigated by the ATSB:

• ATSB investigation RO-2015-016, Level crossing collision between freight train 8834N and road-train truck, Tullamore Rd, Narromine, NSW, on 23 September 2015. This investigation found that ‘The viewing angle approaching the railway crossing (from the southwest) through the right-hand curve meant that at higher road speed there was reduced opportunity for the truck driver to identify that the flashing lights were operating’.
• ATSB investigation RO-2017-005, Level crossing collision between freight train 8426N and road-train truck, Cobb Highway, Ivanhoe, NSW, on 11 July 2017. This investigation noted that it was possible the driver did not identify the flashing crossing lights due to a focus on negotiating the curved approach to the crossing.
• ATSB investigation RO-2021-003, Level crossing collision between freight train 2C74 and road-train truck Yarri Road, Parkeston, Western Australia, on 22 February 2021. This investigation noted that the truck driver’s sighting of the flashing crossing lights was through a curve. With the truck driver distracted by another activity and possibly only looking directly ahead of the vehicle, they were probably not effectively looking through the curve to the crossing lights. Figure 4 illustrates the driver’s view of the level crossing while navigating this curved road.

Figure 4: Image showing overhead and driver view of level crossing relative to vehicle line of travel, from ATSB investigation RO-2021-003

Source: ATSB investigation RO-2021-003, Level crossing collision between freight train 2C74 and road-train truck Yarri Road, Parkeston, Western Australia, on 22 February 2021.

Records from ALCAM surveys showed the recorded S₁ distances for the road approaches travelled by the heavy vehicles, for the 5 collisions in this study which occurred on curved road approaches to flashing light level crossing (no boom gates installed). The ATSB reviewed satellite imagery to identify the position of the road vehicle at the point at which the S₁ distance was identified, and the visual angle from that position to the level crossing protection equipment (flashing lights). These distances and angles are shown in Table 16.

Table 16: Visual angle identified for sighting of flashing lights at selected level crossing collisions at crossings with curved road approaches

Level crossing collisions investigated in other jurisdictions have also been attributed to the effects of curved road approaches, with examples including:

• Canadian Transportation Safety Board Railway Investigation Report R13T0192, Crossing collision involving passenger train and double decker bus on 18 September 2013. This collision occurred at a level crossing protected by flashing lights and boom gates. The investigation identified that ‘while negotiating the … [left] curve on the approach to the crossing, the [bus] driver would have generally gazed toward the tangent point at the centreline of the road and made anticipatory glances toward the occlusion point where the view of the road ahead was obstructed by trees, shrubs, foliage, and roadway signage. In addition to distractions that likely influenced the driver, the additional driver workload associated with negotiating the left-hand curve[26] on approach to the crossing likely decreased the driver’s ability to detect the activated (flashing lights)’.
• Canadian Transportation Safety Board Railway Investigation Report R16D0092, Crossing collision involving passenger train and tractor-trailer 20 September 2016. The collision occurred at a crossing protected by flashing lights and boom gates. The investigation noted that the left curve approaching the crossing obstructed visibility of the level crossing equipment, and the crossing only became completely visible at 730 ft (223 m) from the crossing. The report found that ‘While the tractor-trailer driver was negotiating the curve, his visual attention was probably focused more on the outside of the curve rather than far ahead. Therefore, the driver did not immediately notice that the warning system was activated.’ The investigation noted that once the curve straightened, the driver had only 500 ft (152 m) to sight the crossing, which was insufficient distance to stop the heavy vehicle.
• US National Transportation Safety Board Highway Accident Investigation Report HAR1402, Highway-Railroad Grade Collision involving freight train and truck on May 28 2013. The collision occurred at a Stop sign level crossing, and the investigation found that vegetation and a sharp horizontal curve near the crossing limited the truck driver’s ability to see oncoming trains.

Visual perception during curved road approaches

During vision, light energy passes through the eye and is focussed by specialised anatomy onto photoreceptor cells at the back of the eye. The centre of the focussed image falls upon the centre of the retina (the fovea) where there is the greatest concentration of photoreceptors, and particularly cone photoreceptors which are used for detail and colour vision. In combination with head and eye movements which bring objects into the centre of the field of view, this is referred to as visual fixation, and can be colloquially understood of as the process of looking at something.

Visual acuity (the ability to perceive detail) is greatest for objects which are directly fixated upon, and acuity declines as objects become more eccentric to the centre of the fixation point. When objects are more peripheral to the field of view, they will be less likely to be detected, and perceived less accurately with longer reaction times (see Carrasco and others, 1995 for example). Wolfe and colleagues (2019) found that drivers were more likely to detect brake lights of vehicles on the road straight ahead, and were slower and less accurate when detecting brake lights of vehicles which were in more peripheral locations.  

There are no known standards for the visual angle provided for roadside signs and signals.

Research on visual fixations during curved road driving

When navigating straight roads, drivers primarily focus on the straight-ahead position the vehicle will be in the future, with experienced drivers focusing on the visual focus of expansion (which corresponds to the horizon straight ahead) (Underwood, 2007). The visual information about the vehicle’s current and future positions are located within the same field of view, so the driver is not required to look outside their straight-ahead orientation to steer the vehicle.

Scanning behaviour on curved roads is significantly more complex, since a driver must intermittently look towards the vehicle’s current and future positions to steer the vehicle (Transportation Safety Board of Canada, 2015).

Research shows that during driving on curved roads, visual attention is primarily focussed on the 1–2 second headway position of the vehicle, and these ‘guiding fixations’ are thought to provide the driver with just-in-time information to control steering in the curve (Lehtonen and others, 2014).  Due to the geometry of a curve, drivers must look at more obtuse visual angles to observe locations further along the roadway, with such glances described as ‘look ahead’ fixations.

Shinar and others (1977) found that the mean fixation for drivers was straight ahead for right curves, and 3.6° to the left for left curves,[27] with only 5% of the time directed at the opposite side of the road for left curves and 24% for right curves. This was similar to the result found by Cohen and Studach (1977), they showed that when approaching and travelling through a left curve, drivers fixated on the left side of the road and to a lesser extent the middle of the road. Further, when navigating a right curve, drivers fixate on both sides and the middle of the road.

Olsen and others (1989) also examined driver fixation through left and right curved roads during day-time driving (Figure 5). They found that for left curves, the drivers’ fixation was primarily on the left and centre of the road between 100 and 300 ft (30–91 m) in front of the vehicle. With the opposite result (right and centre) for right curves.

Figure 5: Percentage of drivers’ fixation time per road region for left and right curved rural roads during daytime

Researched was conducted in a region where driving is on the right-side of the road. Directions have been reversed for Australian conditions.

Source: ATSB, based on Olson and others (1989). Driver eye fixations under different operating conditions. The University of Michigan Transportation Research Institute.

Lehtonen and others (2014) used eye-tracking methods to measure the distribution of drivers’ visual attention while negotiating a curved road. This research showed that the majority of drivers’ fixations were within 3–4° of the 2-second headway position of the vehicle. Between 65–85% of fixations during curved driving were ‘guiding fixations’ (5° either side of the 2-second headway position), while 2–28% of fixations were ‘look ahead’ fixations (more than 5° either side of the 2-second headway position). The distribution of fixations by angular eccentricity from the 2-second headway position resembled a normal distribution, such that look-ahead fixations with smaller angular eccentricity were much more common than fixations at greater visual angles.

Another study (Lehtonen and others, 2013) showed that while driving along a curved rural road, drivers typically fixated within 6° either side of the future position of the vehicle, although look-ahead fixations (more than 6° of the headway position of the vehicle) were frequent. Between 8–33% of fixations were categorised as look ahead fixations. Under higher cognitive load look-ahead fixations were shorter and directed closer to the position of the vehicle. The average look-ahead fixation was made 8.0 seconds and 112 m prior to the fixation point under lower workload conditions, and 6.9 seconds and 95 m prior under higher workload conditions.

Lehtonen, Lappi and Summala (2012) found that drivers navigating curved roads regularly glanced towards the furthest visible point on a roadway prior to an obstruction (occlusion point). These fixations on the ‘occlusion point’ were interpreted as visual anticipation of potential hazards and the upcoming road alignment. Fixations on this occlusion point were significantly less frequent for right curves²⁷ and in conditions where the driver was under increased cognitive load. Drivers were also less likely to fixate on the occlusion point over consecutive runs, although this effect was not statistically significant.

This research was conducted in Finland (where road vehicles drive on the right side of the road and the driver sits on the left side of the vehicle). Explaining the much lower rates of occlusion point fixation for (Finnish) left curves, Lehtonen and others (2012), suggested:

Notably, each of the 5 level crossing collisions on curved roads identified in the safety study occurred on right curved road (again noting that in Finland vehicle driver on the opposite side of the road). There is no known dataset which describes the number of left and right curved roads approaching level crossings in Australia.

In summary, research has found that while most fixations are focussed on the immediate future position of the vehicle on curves, drivers do engage in scanning of locations further along a curve including at locations which are eccentric in visual angle to the driver. Some of this behaviour is thought to be anticipatory searching along the most distant viewable part of the curve, to identify upcoming hazards.

The degree of the curve, however, may affect the rate at which these anticipatory glances occur. The results reported by Lehtonen and others (2014) indicate that where the occlusion point of the curve required the driver to turn their head to fixate upon it, anticipatory glances were less frequent. This may also have an adverse implication for fixation upon objects as drivers proceed along a curve (and towards an upcoming level crossing), as an object will become more obtuse to a vehicle travelling a curve until the vehicle reaches the maximum turning point.

The observation that anticipatory scanning is reduced under increased workload may have implications for heavy vehicle drivers. Case study examples suggest that driving a heavy vehicle along a curved road requires focussed attention towards vehicle handling, particularly in degraded road conditions such as uneven road surface and worn shoulders[28] or wet roads.

The implication for road vehicle drivers approaching a level crossing along a curved road approach is that they will rely on these anticipatory look-ahead fixations to identify level crossing protection equipment, as this equipment will often not fall within the normal focal point of their guiding fixations. This provides a lower opportunity for drivers to detect level crossings and advanced warning signs compared to when navigating a straight road, when the future position of the vehicle will fall within the guiding fixations. To the extent that look-ahead fixations are constrained by factors such as workload, or do not capture particularly eccentric locations on a curved road, there is an increased likelihood that level crossing protection equipment will not be detected.

It was not possible to determine to what extent roadside and pavement signage warning drivers of the presence of an upcoming level crossing affects the frequency and eccentricity of their look ahead fixations for a level crossing. Drivers may slow their vehicle or engage in more conscious scanning when they encounter such warning signs. This behaviour would reduce the risk of level crossing protection equipment not being detected.

Prior investigations and changes to Australian Standards

In July 2016, the ATSB published an investigation report into a level crossing collision between a road-train truck and a freight train, which occurred on a flashing lights-controlled level crossing with a curved road approach in Narromine, NSW. The report identified that the then-current version of AS 1742.7 (2007 version):

Then, in March 2018, the ATSB published a report into a similar accident at Ivanhoe, NSW. This report identified that AS 1742.7 had been updated to the 2016 revision, including providing consideration to the treatment of crossings with curved road approaches. The ATSB report stated that:

In January 2019, AS 1742.7(2016) was updated. Prior to the revised text amendment, the standard included the following paragraph:

The revised text amendment replaced this paragraph with:

Additional evidence from Standards Australia

In July 2023, the ATSB met with representatives from the Standards Australia technical committee responsible for AS 1742.7: 2016. The committee provided advice about how the standards should be applied in the case of curved road approaches, with observations including:

• The standards are expected to be applied by trained and competent surveyors.
• The inclusion of text stating that level crossing owners may use active advanced warnings is significant, since crossing owners will need to justify why they have not utilised this form of protection equipment.
• The signage and pavement markings required and recommended in the standard would affect the behaviour of road drivers. For example, a road driver who is alerted to the presence of an upcoming crossing may expect and search for a crossing. This means crossing equipment located outside a primary field of view may be detected more readily.

Standards Australia provided additional analysis in September 2023, which stated:

Summary

Research has shown that when driving on a curved road, road vehicle drivers mainly fixate on the immediate headway position of the vehicle and rely on anticipatory look-ahead fixations to observe objects further ahead. This provides less opportunity to detect level crossing protection equipment, compared to driving on a straight road, especially on right curved roads where an advanced warning sign is located not duplicated on the right side.

Standards and guidance for designing and assessing level crossings do not identify the difference in driver sighting behaviour while navigating curved roads. The treatment of curved roads in the standards indicates that road designers and assessors must primarily consider if there is a clear line of sight through a curve. The standards do not convey that while there may be a line of sight through a curve, if this occurs through a large visual angle then there is an increased chance of drivers not sighting the level crossing.

The standards include the provision of active advanced warning on locations along a curved approach road, which would provide a mechanism of locating the crossing warnings within drivers’ central fields of view on curved roads. However, there was no guidance which indicates under what conditions the placement of advanced warning signs is necessary, and the use of these warnings is not mandatory.

This review identified that of 6 collisions which involved a heavy vehicle driver not stopping at a level crossing protected by flashing lights, 5 occurred following a right curved approach road and none of these locations had active advanced flashing light warnings. In 4 of these cases, the assessed available sighting distance was measured from a location where the sighting of the level crossing was from a large visual angle.

The ATSB considers that the absence of procedures or guidance which account for the normal sighting behaviour of drivers on curved roads for determining safety stopping distances and location of advanced warning signs reflects a condition of significant risk.

Safety systems at passive controlled level crossings

Safety system theory and level crossings

This report has described the development of level crossing collisions involving heavy vehicles, including describing the actions of the heavy vehicle drivers which led to the collisions with trains.

The systems safety approach identifies that rather than being the product of individual actions and events, safety is an emergent property of a system. While individual errors and violations affect the development of adverse events, these actions and omissions typically occur in a context where they ‘made sense at the time’. The systemic approach seeks to understand the latent conditions which shaped the behaviour of individual operators and contributed to the conditions at the time of an accident.

The systemic approach also seeks to understand the redundancies available in the system, with the concept of ‘defences-in-depth’ referring to the use of multiple, overlapping risk controls to prevent catastrophic accidents. Administrative controls, such as procedural rules, provide the weakest form of protection against the likelihood and consequence of individual errors by frontline personnel. In contrast, engineering controls provide the greatest level of protection through providing a physical fail-safe barrier which prevents or reduces harm.

The safety system at passive controlled level crossings does not include any engineering controls to alert road vehicle drivers (or pedestrians) about the presence of a train. At passive control crossings, road vehicle drivers must visually search for and detect the presence of a train in order to identify when they can or cannot proceed through a crossing. As such, the design of passive controlled level crossings is primarily dependent on the actions of road vehicle driver to detect both the crossing and detect the presence of trains. If a road vehicle driver does not look for trains, or looks for trains but does not detect that one is approaching, there are limited effective recovery controls to prevent a collision from occurring. The driver may be alerted to the presence of a train by the train horn, however in many instances this is not effective (see Train horn audibility).

With no limited effective recovery controls for instances where the road user does not attend to the crossing or does not detect the presence of a train, the safety system for passive level crossings inherently relies on the road user always attending to upcoming crossings, and always detecting trains.

A human-centred approach to level crossing errors

A large subset of heavy vehicle drivers involved in level crossing collisions were aware that they were approaching or proceeding through a level crossing. The breakdown in safety predominantly occurred due to the drivers not detecting that a train was approaching a crossing. This study identified that of 26 accidents at passively controlled crossings:

• In at least 20 collisions the heavy vehicle driver probably did not detect the presence of the train, or detected the train when it was too late to stop.
• In at least 1 collision the heavy vehicle driver detected the train but incorrectly determined that they could transit the crossing safely.
• In 1 collision the heavy vehicle driver detected the train but records indicated the truck had defective brakes and could not stop.
• In 4 collisions there was not sufficient evidence to determine whether the heavy vehicle driver had detected the presence of a train.

Overall, in a large majority of the 26 collisions at passive controlled level crossings, the heavy vehicle driver unintentionally proceeded into the crossing without detecting the presence of the train. These collisions reflected unintentional slips and lapses, where the drivers unintentionally deviated from the intended course of action (to give way to trains).

Reason (1990) notes that such errors typically occur in the context of a routine, well-practiced task where behaviour is largely automatic. Attention is captured by a distraction or other demand, and the operator misses the cues to depart from the ‘normal’ routine.

Applying this conceptual framework to level crossing collisions, driver errors take place in the context of the highly-practiced task of road vehicle driving, in roadway environments where other demands and distractions are often prevalent. Road vehicle driving can become highly automatic, and many road vehicle drivers will recall the experience of arriving at a common destination without a recollection of the journey. The phenomenon of ‘highway hypnosis’ or ‘driving without attention mode’ describes the degradation of vigilance towards the roadway associated with factors including monotonous driving conditions (Karrer and colleagues, 2005). In the context of a potential level crossing collision, drivers are required to depart from the current (and often automatic) schema of continuing to drive the motor vehicle along the road, and adopt a schema of looking and preparing to stop for trains.

The task of approaching a passive control level crossing is thus consistent with the conditions described by Reason (1990) as being conducive to the development of errors involving inattention. Road vehicle drivers are vulnerable to becoming ‘captured’ by focus on other aspects of road vehicle driving, and not attend to the roadside or the rail track to detect level crossings and trains.

As described in Factors affecting heavy vehicle driver behaviour at level crossings above, there are various factors which may contribute to road vehicle drivers not looking for, or detecting the presence of, a train at a passive control level crossing. As highlighted by Edquist and others (2009), when considering the perceptual and cognitive factors inherent in the task of approaching a passive control level crossing:

This is consistent with analysis by the Transportation Safety Board of Canada[29] (TSB), which found that:

The US National Transportation Safety Board (NTSB) safety study of collisions at passive control level crossings[30] made a similar conclusion, arguing that the long-term objective to reducing collisions at level crossings should be to eliminate passive control crossings.

In summary, many of the collisions and near misses identified in this study speak to the inherent variability of human performance, in the context of a system which is conducive to unintentional errors. Level crossing safety may be improved by the introduction of other risk controls to reduce the frequency of level crossing collisions or reduce their consequences.

Potential safety improvements for passive control crossings: the case for in-vehicle warning systems

Background

One proposed method for improving the safety of passive level crossings is using in-vehicle alerting technology to alert road vehicle drivers of a requirement to stop and give way to an approaching train. Such applications would reduce the safety system’s reliance on the driver to search for and detect the presence of a train, by attracting the driver’s attention and thereby providing a salient cue to the driver to stop.

In-vehicle alerting systems for level crossings have been proposed by previous investigations and reviews of level crossing collisions. Following a review of collisions at passive control level crossings, the NTSB (1998) noted:

Reporting on an accident at a passive control level crossing, the Canadian TSB found:

In the Australian context, a 2009 Parliamentary committee Level crossing safety recommended that:

Then, in 2013, the Coronial inquest into multiple fatal level crossing collisions (including the collision at Kerang, Victoria) recommended rail operators, infrastructure managers and regulators:

In summary, there is broad agreement among informed stakeholders both within Australia and internationally on the inherent challenges of relying on road users to attend to and detect the presence of trains at passive control crossings. Multiple stakeholders have endorsed the development of in-vehicle alerting technology for level crossings to improve the safety at passive control level crossings.

Current maturity of in-vehicle level crossing warning systems

The ATSB sought information about the development and trial of level crossing alerting technology following the recommendations made in the Coroners Court of Victoria’s Coronial Investigation of Twenty-six Rail Crossing Deaths in Victoria, Australia (2013). The Victorian Government Department of Transport and Planning advised that trials had been conducted, however the technologies were found to be unreliable. The Department further advised that the development of technology was being monitored, with a 5- to 10-year timeframe expected for a viable technology. Trials of level crossing technology cited by the Department included:

• In 2013, the La Trobe University Centre for Technology Infusion reported on the development and evaluation of 5.9 GHz Dedicated Short Range Communication (DSRC) system which provided audio and visual alerts to road users approaching a crossing. Field trials were conducted in rural and urban locations, involving 124 participants over 4.5 months. The trials included variants where trains and vehicles communicated directly, and variants where the vehicles communicated through a roadside unit intermediary.
  The results of the field trials showed that connectivity between vehicles was dependent on there being a clear line of sight between receiving units, and that the system provided much less connectivity range when no roadside unit was deployed.
• The Collaborative Research Centre for Rail Innovation (2014) reported on a separate trial of the La Trobe University DRSC system. This report noted that ‘drivers experienced a high number of hardware issues with the Latrobe system: display falling off, warning messages when no train was approaching. Further, participants had to start the system manually every time they started driving, which was found annoying. All of this resulted in a low usage of the system during the trial’.

A review of international research into level crossing warning systems identified the US Federal Railroad Administration Rail-Crossing Violation Warning (RCVW) project (Withers and Utterback, 2021). This RCVW detects the status of active control level crossings. This system would produce in-vehicle messages for all vehicles approaching an active crossing of the requirement to stop (inform message), and an alert if the vehicle is not predicted to stop based on its speed (warning message). Field testing has shown that system provided a reliable warning system for active control level crossings, using currently available technology. Notably, this technology does not detect whether a train is approaching the crossing (relying on existing active crossing equipment), and thus does not provide a solution for passive controlled crossings.

In summary, there is no known research which has demonstrated a proof-of-concept for a system for providing alerts to road vehicle users when approaching level crossings. Research conducted at Australian level crossings indicated that the available technology was not reliable. Given this research is a decade old, additional research is now required to determine whether technological advancement will now provide a useable platform for providing warnings to drivers approaching level crossings.

Human factors considerations for in-vehicle level crossing warning systems

In addition to the absence of mature technology for providing in-vehicle level crossing warnings, there are potentially significant implications for how such warnings would affect the behaviour of road vehicle drivers at level crossings. The phenomenon of behavioural adaptation has been observed in response to other changes to the road traffic system, where there is an observed aggregate change in the behaviour of road users following system changes, often resulting in the estimated benefits of the changes to be under-achieved. For example, research has identified that drivers presented with adaptive cruise control attended more frequently to a secondary task and had lower reaction times (Rudin-Brown and Parker, 2004). Similar results have also been observed in studies of adaptation to forward collision warning systems (Reinmueller and others, 2010).  

In the case of in-vehicle level crossing warnings, it is possible that road vehicle drivers may adapt to the presence of the warnings, and that this may affect their attention to traditional level crossing control equipment (Wullems and others, 2014).

Simulator research showed that participants who drove a vehicle equipped with an in-vehicle alerting system were less likely to comply with level crossing rules when there were no trains present. Drivers also approached the crossings at faster speeds and glanced less frequently at the rail track (Larue and others 2015). These changes were only observed when the driver was presented with a visual warning, and driving behaviour improved when an aural warning was presented.

As identified by Wullems and others (2014), the potential for over-reliance on in-vehicle warnings and changes to driving behaviour at level crossings presents potential hazards, unless the systems are perfectly reliable. Any failure of the system would produce a potentially very hazardous system state, particularly if the road user does not identify the failure.

It may also be hazardous to incrementally introduce the equipment necessary to provide in-vehicle warnings. If road vehicle drivers were to come to expect in-vehicle warnings for potential level crossing collisions, then level crossings and/or trains which were not equipped with the equipment required to produce these warnings would potentially become less safe than they are currently.

Other barriers to the implementation of in-vehicle level crossing warning systems

Any foreseeable system for providing in-vehicle level crossing warnings requires road vehicles be equipped with devices for receiving and displaying warnings, and the potential benefit of any warning system would depend on the installation uptake of such equipment. Considering the size of the road vehicle fleet (over 20 million road vehicles were registered in Australia in 2022), this could be a complex and costly exercise, requiring the co-ordination of stakeholders outside the rail industry. However, with advancements in navigation systems in many on-road vehicles today that allow for software updates, this is likely to be less complex in the future.

A final barrier to the implementation of in-vehicle level crossing warning systems relates to the ownership of risk at level crossings, and how responsibility for in-vehicle warnings could be managed. The existing regime for level crossing safety includes shared responsibility between rail and road infrastructure managers (as reflected in level crossing interface agreements), and with railway operators and road vehicle drivers for complying with relevant rules and procedures. Wullems and others (2014) identified that the introduction of in-vehicle level crossing warning systems would pose unresolved questions relating to where responsibility lies for ensuring that warnings are produced by rail vehicles and/or infrastructure, and being received by road vehicles. The ongoing monitoring, management and maintenance of such a system may require significant cost for rail industry organisations.

Wullems and others (2014) further identified that failures in the system which led to collision may produce a liability for rail operators or infrastructure managers, whereas in the existing regime these stakeholders can discharge their responsibility by providing compliant level crossing infrastructure and operating rollingstock correctly.

Whereas the use of in-vehicle warnings has been mooted as a low-cost method for improving safety at passive control crossings, there are significant changes required to implement such technology which may involve significant cost. The size of these costs is unknown, and will need to be more precisely understood to evaluate the safety case of potential alerting technologies, should fit-for-purpose technologies emerge.

Discussion

Study limitations

This study reviewed level crossing collisions involving heavy road vehicles, to understand their characteristics and trends, and in turn to identify risk factors and opportunities for safety action. This study did not review the characteristics and trends of heavy vehicle movements across level crossings which did not result in a collision. This limited the inferential capability of the study, such that it was not possible to quantify how specific characteristics of the vehicles, crossings, or other variables identified in the study contributed to the development of collisions.

Similarly, the study did not conduct an in-depth review of the characteristics and trends of level crossing collisions involving other road vehicles. As such, it was not possible to quantify how the different characteristics of heavy vehicles, or the different road and crossing conditions they may be exposed to, contributed to the observed differences in collision frequency and consequence. A richer dataset for all level crossing collisions would support such analysis.

Another limiting feature of this study was that the information used to understand the characteristics of level crossing collisions involving heavy vehicles was primarily drawn from investigation reports produced by rollingstock and infrastructure managers, and from police reports. These reports were produced in the context of the operator and infrastructure managers’ specific interests and/or legal requirements, and not with a view to support broader analysis such as this.

The data used in this study was obtained from sources that included either live databases where the data can be updated through additional data gathering or data cleaning exercises. For one level crossing accident reviewed in this study, the source material included an Office of the Chief Investigator investigation that was ongoing at the time the study was published.

The ATSB did not conduct any site inspections or audits against relevant standards of the level crossings involved in the occurrences in this safety study. Previous ATSB investigations of level crossing collisions have included site inspections and other analyses of the crossings, and the safety study utilised records from these observations. Otherwise, the safety study relied on observations of level faults or deficiencies recorded in rail infrastructure manager reports and other documents.

Information about the factors which affected the performance of heavy vehicle drivers was infrequently reported in operator and police reports. In most instances, these reports only identified that the road vehicle driver failed to give way, and these records rarely assessed human factors causes which could explain the drivers' errors.  

In contrast with operator and police reports, investigations conducted by the ATSB seek to understand the operation of the overall safety system, and to identify opportunities to improve that system. The objective is not to attribute blame or liability. The ATSB also has strict limitations on how information collected from involved parties can be used outside of the no-blame investigation process, which probably increases the likelihood of obtaining evidence such as statements from heavy vehicle drivers.

The independent no-blame investigations conducted by the ATSB were much more likely to provide sufficient information to describe the actions of the heavy vehicle drivers, and to provide information which explained these actions. As shown in Table 17, the proportion of collisions where the ATSB was unable to code key details about the heavy vehicle driver’s actions or the factors associated with those actions was much greater for collisions not investigated by the ATSB.

Table 17: Counts of collisions where road vehicle driver actions and related variables could not be coded due to missing information

There are some important caveats in drawing conclusions based on these differences. Most importantly, the collisions not investigated by the ATSB were different events, and may not have involved the factors identified in the collisions investigated by the ATSB. Also, the ATSB typically allocates more time and resources to conducting an investigation than a rail organisation would, and it is difficult to distinguish the effects of this resourcing with the benefits of conducting no-blame investigations.

Nonetheless, the ATSB considered that the information obtained in this study demonstrated that meaningful understanding of why heavy vehicle drivers enter level crossings contrary to the road rules was typically only available following an independent no-blame investigation. As these actions represent the fundamental breakdown of safety at level crossings, this supports the benefit of regularly investigating significant collisions at level crossings.

Overall observations as to heavy vehicle level crossing risk

The information reviewed in this safety study indicated that heavy vehicles pose a greater risk at level crossings than light vehicles, as a function of greater accident consequence for rail users. The collisions involving heavy vehicles resulting in more significant consequences was consistent with the observations of previous research, and can be easily understood as a simple reflection of the increase in physical forces imparted on a rail vehicle as the mass of the road vehicle involved in a collision increases (for a fixed speed).

Also consistent with previous analyses, this study identified that heavy vehicles are more likely to be involved in level crossing collisions, compared to light vehicles, per road distance travelled. The increase in level crossing collision frequency for heavy vehicles may be partially explained by a greater exposure to level crossings, and in particular greater exposure to level crossings with lower levels of protection (especially in rural areas). The study was unable to test the extent to which different exposure to level crossing hazards explained the difference in collision rates.

Nonetheless, the over-representation of heavy vehicles in level crossing collisions is likely to be due, in some part, to the fundamental physical property of heavy vehicles: that they are larger and take longer to stop prior to a crossing, and clear a crossing, as a function of both time and distance. Although the design of level crossings seeks to accommodate the stopping limitations of heavy vehicles (see Appendix B), it is inevitable that a heavier vehicle will be less likely to stop within a specific distance and time, all other factors being equal. Stated alternatively, all safety improvements at level crossings seek to increase the margins available for a road vehicle to stop and give way to trains, and any improvements which increase the ability of heavy vehicles to stop at level crossings will have a greater benefit to smaller vehicles whose braking capabilities are generally greater. As such, it is likely that heavy vehicles will continue to be over-represented in level crossing collisions.

The characteristics of heavy vehicle driving may also represent systemic risk factors which partially account for the over-representation of heavy vehicles in level crossing collisions. Plausible mechanisms for such systemic effects include:

• The handling characteristics of heavy vehicles may contribute to a greater tendency of heavy vehicle drivers to engage in ‘rolling-stop’ violations at level crossings, compared to the drivers of other road vehicles.
• The louder operating noises produced by heavy vehicle engines and heavier cab construction may result in heavy vehicle drivers being less likely to hear train horns, compared to the occupants of other road vehicles.
• The construction of heavy vehicle cabs may also provide more visual obstructions, making it more difficult for heavy vehicle drivers to detect trains and level crossing equipment.
• The nature of heavy vehicle driving is often one of repeated journeys along the same route. This familiarity can contribute to the expectancy that level crossings along the route will not be occupied by trains.

As identified in Study limitations, this study did not include data from heavy vehicle movements across level crossings that did not result in a level crossing collision, and did not conduct a thorough review of collisions involving non-heavy road vehicles. Due to this, it was not possible to accurately determine the characteristics of heavy vehicles that contributed to the greater frequency of level crossing collisions (per road distance travelled).

Overall observations as to level crossing collision characteristics

Of the 49 level crossing collisions involving heavy vehicles, at least 7 were partially attributable to problems with the design or maintenance of the level crossings. Generally, however, the collisions reviewed in this study occurred at level crossings designed according to the required standards. Also, there were no collisions attributed to the crossing providing insufficient sighting for a stopped heavy vehicle to start up and clear the crossing prior to a train arriving.

However, the study identified one systemic problem with the design criteria for level crossings. The methods used to calculate safe stopping distances for road approaches to level crossings did not account for the effect on sighting a level crossing from the normal focal points of drivers negotiating a curved road. While this problem did not only relate to heavy vehicles for the design of level crossings, heavy vehicle drivers are likely to be particularly affected in situations where insufficient sighting distance is provided.

The majority of level crossing collisions arose from heavy vehicle drivers entering the level crossings following some form of unintentional error or omission. Even in the case where heavy vehicle drivers deliberately engaged in actions contrary to the road rules, the intention was to proceed through the crossing prior to the arrival of a train. The collisions were primarily caused by level crossing warnings or the presence of trains not being detected, being detected late, or being perceived incorrectly. This form of omission or misperception can also affect other road users, and the forms and causes of errors observed in heavy vehicle drivers were consistent with the errors made by the drivers of other road vehicles, both at level crossings and in other areas of the road network.

As described in A human-centred approach to level crossing errors, many of the errors and omissions which lead to level crossing accidents reflect the inherent variability of human performance. While the level crossing safety system continues to heavily rely on road vehicle drivers adhering to the administrative controls of road rules and other procedures to prevent collisions between vehicles and trains, it is a certainty that from time to time, this will not occur and so collisions will continue to occur. However, the use of engineering controls which alert road users to a requirement to stop will almost certainly provide an enhanced level of safety at level crossings, by reducing the reliance on road users to attend to and detect the presence of trains.

Findings

From the evidence available, the following findings are made with respect to the review of level crossing collisions between trains and heavy road vehicles.

Factors that increased risk

• Of the 49 level crossing collisions involving heavy vehicles, at least 7 partially involved problems with the design or maintenance of the level crossings. Concerning compliance with applicable design standards, no significant issues were identified with the design or maintenance of most crossings.
• Level crossing collisions between trains and heavy vehicles were associated with greater levels of rail injuries and rail damage than collisions involving light vehicles.
• Heavy vehicles are involved in level crossing accidents at a greater rate per road kilometre travelled than light vehicles.
• Although level crossing collisions between heavy vehicles and trains were more likely to involve the train striking the heavy vehicle, accidents where the heavy vehicle struck the train were more likely to cause a derailment.
• All level crossing collisions involving heavy vehicles resulted from the heavy vehicle driver not giving way to trains. There were three actions associated with the collisions:
  • There were at least 24 accidents where the heavy vehicle did not stop prior to entering the crossing.
  • There were at least 11 accidents where the heavy vehicle stopped at the crossing then proceeded into the path of a train.
  • There were at least 13 accidents where the heavy vehicle entered a level crossing and stopped foul of the train line.
• In at least 14 collisions it is likely that the heavy vehicle driver intentionally entered the level crossing in a manner which was contrary to road rules. These included 6 collisions where the driver intentionally entered the crossing without being able to drive clear of the crossing, 4 collisions where the driver engaged in a 'rolling stop' while approaching a Stop sign level crossing, 3 collisions where the driver remained stopped foul of the crossing while there was no obstacle preventing them from exiting, and one collision where the driver did not stop or look for trains at a Stop sign crossing.
• Of 26 collisions at passive control crossings, there were at least 12 accidents where the heavy vehicle driver slowed or stopped but probably did not detect the train, and entered the crossing into the path of the approaching train.
• There were at least 12 level crossing collisions where the driver of the heavy vehicle had regularly used the level crossing prior to the collision with a train. This included 6 accidents where the heavy vehicle driver proceeded into a passive control crossing without identifying the presence of a train, and 5 accidents where the heavy vehicle driver did not identify activated flashing level crossing lights. The drivers' previous experience at the level crossings likely led to a low expectancy for trains and, in at least some collisions, contributed to them not detecting a requirement to stop and give way.
• Of the 49 level crossing collisions involving heavy vehicles, there were at least 14 collisions where the heavy vehicle driver’s view was obstructed by vegetation, the design of the heavy vehicle cab, poor crossing lighting, or sun glare.
• Previous research and reviews have identified that train horns have limited effectiveness for alerting road vehicle drivers to the presence of trains. Consistent with this, in at least 25 collisions, the horn was not effective at alerting the heavy vehicle driver to the presence of the train.
• The methods used in the Australian Standard AS 1742.7:2016 to calculate safe stopping distances, and determine the need and location of advanced warning signs for road approaches to level crossings, did not account for the likelihood of detecting the level crossing ahead based on the normal visual focal points of road drivers negotiating a curved road. While the standards included guidance for the use of active warning signs for curved road approaches to flashing light controlled crossings, this was not mandatory. There were 6 collisions in which driver of a heavy vehicle did not detect that level crossing flashing light signals were activated until it was too late to stop. In 5 of these collisions, the drivers approached the crossing along a right curved approach road and none of these crossings had active advance flashing light signals (Safety issue).
• Of 26 collisions at passive control level crossings, a large majority involved the heavy vehicle driver not detecting the presence of the train. These crossings rely on the road vehicle driver detecting the presence of a train and identifying a requirement to stop and give way, and are susceptible to situations where motorists do not look for, or detect, the presence of a train. There are limited effective recovery controls to prevent a collision when this occurs.

Other findings

• Information which supports an understanding of the factors associated with heavy vehicle drivers entering a level crossing without giving way to trains is typically only available when an independent no-blame transport safety investigation is conducted.
• The annual number of level crossing collisions between road vehicles and trains remained relatively constant between July 2014 and June 2022.
• Prior to July 2022, the national rail safety database did not include sufficient information to enable detailed analysis of level crossing collision characteristics.

Safety issues and actions

Standards and guidance for placement of crossing equipment on curved road approaches

Safety issue number: RS-2021-001-SI-01

Safety issue description: The methods used in the Australian Standard AS 1742.7:2016 to calculate safe stopping distances, and determine the need and location of advanced warning signs for road approaches to level crossings, did not account for the likelihood of detecting the level crossing ahead based on the normal visual focal points of road drivers negotiating a curved road. While the standards included guidance for the use of active warning signs for curved road approaches to flashing light controlled crossings, this was not mandatory. There were 6 collisions in which driver of a heavy vehicle did not detect that level crossing flashing light signals were activated until it was too late to stop. In 5 of these collisions, the drivers approached the crossing along a right curved approach road and none of these crossings had active advance flashing light signals.

Sources and submissions

Sources of information

The sources of information during the investigation included the:

• Office of the National Rail Safety Regulator
• Rail operators
• Rail infrastructure managers
• Australian Level Crossing Assessment Model Committee
• Federal Railroad Administration
• Rail Industry Safety and Standards Board
• New South Wales Police
• South Australia Police
• Victoria Police
• Queensland Police Service
• National Heavy Vehicle Regulator

References

Abraham, J., Datta, T. K., & Datta, S. (1998). Driver behavior at rail-highway crossings. Transportation Research Record, 1648(1), 28-34.

Australian Centre for Rail Innovation (2022). Freight train visibility review.

Australian Transport Safety Bureau (2002). Monograph 10: Level Crossing Accidents. Canberra.

Austroads (2010). Measures for Managing Safety of Heavy Vehicles at Passive and Active Railway Level Crossings

Austroads (2017). Improved Railway Road Design for Heavy Vehicles

Beanland, V., Salmon, P. M., Filtness, A. J., Lenné, M. G., & Stanton, N. A. (2017). To stop or not to stop: Contrasting compliant and non-compliant driver behaviour at rural rail level crossings. Accident Analysis and Prevention, 108C, 209-219. doi: 10.1016/j.aap.2017.09.004

Blower, D. F. (2007). Truck mirrors, fields of view, and serious truck crashes. University of Michigan, Ann Arbor, Transportation Research Institute.

Caird, J. K. (2002). Human factors analysis of highway-railway grade crossing accidents in Canada.

Carrasco, M., Evert, D. L., Chang, I., & Katz, S. M. (1995). The eccentricity effect: Target eccentricity affects performance on conjunction searches. Perception & psychophysics, 57, 1241-1261.

Carroll, A. A., Multer, J., & Markos, S. H. (1995). Safety of highway-railroad grade crossings: use of auxiliary external alerting devices to improve locomotive conspicuity (No. DOT-VNTSC-FRA-95-10). United States. Federal Railroad Administration. Office of Research and Development.

Casali, J. G., Robinson, G. S., & Lee, S. E. (2002). Masked thresholds and predicting the audibility of auditory displays: An example for long-haul trucks. In Proceedings of the Human Factors and Ergonomics Society Annual Meeting (Vol. 46, No. 19, pp. 1697-1701).

Chadwick, S. (2017). Quantitative analyses of train derailment probability at highway-rail grade crossings.

Chadwick, S. G., Saat, M. R., & Barkan, C. P. (2012). Analysis of factors affecting train derailments at highway-rail grade crossings. In Proceedings of the Transportation Research Board 91st Annual Meeting.

Cherchas, D. B., English, G. W., Ritchie, N., McIlveen, E. R., & Schwier, C. (1982). Prediction of the probability of rail vehicle derailment during grade crossing collisions.

Cohen, A. S., Studach, H. (1977). Eye movements while driving cars around curves. Perceptual and Motor Skills, 44, 683-389.

Coroner’s Court of Victoria. (2013). Coronial Investigation of Twenty-six Rail Crossing Deaths in Victoria, Australia

Dolan, T. G., & Rainey, J. E. (2005). Audibility of train horns in passenger vehicles. Human factors, 47(3), 613-629.

Eck, R. W. (2002). A context-sensitive approach to improving safety at passive crossings. In international symposium on railroad-highway grade crossing research and safety, 7^th, 2002, Melbourne, Victoria, Australia.

Edquist, J., Stephan, K., Wigglesworth, E., & Lenne, M. (2009). A literature review of human factors safety issues at Australian level crossings.

Federal Railroad Administration (1995). Nationwide Study of Train Whistle Bans. U.S. Department of Transportation, Federal Railroad Administration, Office of Safety: Washington, D.C.

Federal Railroad Administration. (2000). Updated Analysis of Train Whistle Bans. U.S. Department of Transportation, Federal Railroad Administration, Office of Safety: Washington, D.C.

Gou, M., & Bellavigna-Ladoux, O.(2003). Impact of heavy vehicles on crossing safety – Development of an adapted design tool. Transport Canada.

Hellman, A. D., & Poirier, P. J. (2019). Highway-Rail Intersection Crash Taxonomy for Connected Vehicle Safety Research: 2008–2017–Ten Year Trend (No. DOT/FRA/ORD-19/25). United States. Department of Transportation. Federal Railroad Administration. Office of Research, Development, and Technology.

House of Representatives Standing Committee on Infrastructure, Transport, Regional Development and Local Government. (2009). Level crossing safety. Australian Government, Canberra.

Independent Transport Safety Regulator. (2011). Transport Safety Bulletin: Level crossing accidents in Australia. New South Wales Government.

Karrer, Briest, Vohringer-Khunt, Baumgarten and Schleicher (2005). ‘Driving without awareness’ in Underwood (ed) Traffic and Transport Psychology: Theory and Application.

Kassa, E., Wan, I., & White, J. (2023). Assessment of Trials to Improve Train Conspicuousness Approaching Passive Level Crossings, Report No: Monash/RT/2023/1742, Monash Institute of Railway Technology.

Larue, G.S., Filtness, A.J., Wood, J.M., Demmel, S., Watling, C.N., Naweed, A. and Rakotonirainy, A., (2018). Is it safe to cross? Identification of trains and their approach speed at level crossings. Safety science, 103, pp.33-42.

Larue, G., Kim, I., Rakotonirainy, A., Haworth, N., Fereira, L. Driver’s behavioural changes with new intelligent transport system interventions at railway level crossings - A driving simulator study. Accident Analysis and Prevention, 81, pp. 74-85.

Lehtonen, E., Lappi, O., Kotkanen, H., & Summala, H. (2013). Look-ahead fixations in curve driving. Ergonomics, 56(1), 34-44.

Lehtonen, E., Lappi, O., Koirikivi, I., & Summala, H. (2014). Effect of driving experience on anticipatory look-ahead fixations in real curve driving. Accident Analysis & Prevention, 70, 195-208.

Lehtonen, E., Lappi, O., & Summala, H. (2012). Anticipatory eye movements when approaching a curve on a rural road depend on working memory load. Transportation research part F: traffic psychology and behaviour, 15(3), 369-377.

Mok, S. C., & Savage, I. (2005). Why Has Safety Improved at Rail‐Highway Grade Crossings?. Risk Analysis: An International Journal, 25(4), 867-881.

Ngamdung and da Silva. (2012). Driver Behaviour Analysis at Highway-Rail Grade Crossings using Field Operational Test Data – Heavy Trucks. US Department of Transportation, Federal Railroad Administration.

Niewoehner, W., & Berg, F. A. (2005). Endangerment of pedestrians and bicyclists at intersections by right turning trucks. Statistics, 1-15.

Olsen, P. L., Battle, D. S., Aoki, T. (1989). Driver eye fixations under different operating conditions. The University of Michigan Transportation Research Institute.

Organisation for Economic Cooperation and Development. (2021). Road Safety Report 2021 – Australia. Presented at the International Transport Forum.

Parnell, K. J., Stanton, N. A., & Plant, K. L. (2016). Exploring the mechanisms of distraction from in-vehicle technology: The development of the PARRC model. Safety science, 87, 25-37.

Proctor, R.W. and Proctor, J.D. (2021). Sensation and perception. In Handbook of Human Factors and Ergonomics (eds G. Salvendy and W. Karwowski).

Rapoza, A. S., & Raslear, T. G. (2001). Analysis of railroad horn detectability. Transportation research record, 1756(1), 57-62.

Read, Beanland, Lenne, Stanton, Salmon, Mulvihill, Walker and Cornellison (2017). A systems analysis of rail level crossings. In Integrating Human Factors Methods and Systems Thinking for Transport Analysis and Design. CRC press.

Reason, J. (1990). Human error. Cambridge university press.

Reason, J. (1995). Understanding adverse events: human factors. BMJ Quality & Safety, 4(2), 80-89.

Reinmueller, K., Kiesel, A., & Steinhauser, M. (2020). Adverse behavioral adaptation to adaptive forward collision warning systems: An investigation of primary and secondary task performance. Accident analysis & prevention, 146, 105718.

Rudin-Brown, C. M., & Parker, H. A. (2004). Behavioural adaptation to adaptive cruise control (ACC): implications for preventive strategies. Transportation Research Part F: Traffic Psychology and Behaviour, 7(2), 59-76.

Rudin-Brown, C. M., George, M. F. S., & Stuart, J. J. (2014). Human factors issues of accidents at passively controlled rural level crossings. Transportation research record, 2458(1), 96-103.

Sanders, M. S., McCormick, E. J. (1993). Human factors in engineering and design. United States: McGraw-Hill.

Salmon, P. M., Read, G. J., Stanton, N. A., & Lenné, M. G. (2013). The crash at Kerang: Investigating systemic and psychological factors leading to unintentional non-compliance at rail level crossings. Accident Analysis & Prevention, 50, 1278-1288.

Shinar, D., McDowell, E. D., Rockwell, T. H. (1977). Eye movements in cure negotiation. Human Factors, 19(1), 63-71.

Sun, D., El-Basyouny, K., & Kwon, T. J. (2018). Sun glare: network characterization and safety effects. Transportation research record, 2672(16), 79-92.

Transportation Safety Board of Canada (2015). Railway Investigation Report R13T0192, Crossing collision involving passenger train and double decker bus on 18 September 2013.

Underwood, G. (2007). Visual attention and the transition from novice to advanced driver. Ergonomics, 50(8), 1235-1249.

Wickens CD, Hollands JG, Banbury S & Parasuraman R. (2013), Engineering psychology and human performance, 4th edition, Pearson Boston, MA.

Wickens, CD & McCarley, JS. (2008), Applied attention theory, CRC Press, Boca Raton, FL.

Withers, J., & Utterback, J. (2021). Rail Crossing Violation Warning Application-Phase II [Research Results] (No. RR 21-12). United States. Department of Transportation. Federal Railroad Administration.

Witte, K., & Donohue, W. A. (2000). Preventing vehicle crashes with trains at grade crossings: the risk seeker challenge. Accident Analysis & Prevention, 32(1), 127-139.

Wolfe, B., Sawyer, B. D., Kosovicheva, A., Reimer, B., & Rosenholtz, R. (2019). Detection of brake lights while distracted: Separating peripheral vision from cognitive load. Attention, Perception, & Psychophysics, 81, 2798-2813.

Wolfe, B., Dobres, J., Rosenholtz, R., & Reimer, B. (2017). More than the Useful Field: Considering peripheral vision in driving. Applied ergonomics, 65, 316-325.

Wullems, C., Wayth, R., Galea, V., & Nelson-Furnell, P. (2014). In-vehicle railway level crossing warning systems: can Intelligent Transport Systems deliver?. In Proceedings of the 2014 Conference on Railway Excellence: Rail Transport for Vital Economy (pp. 1-12). Railway Technical Society of Australasia (RTSA).

Yeh, M., & Multer, J. (2008). Driver behavior at highway-railroad grade crossings: A literature review from 1990-2006. Report published by the Federal Railroad Administration.

Young, K., Regan, M. and Hammer, M., 2007. Driver distraction: A review of the literature. Distracted driving, 2007, pp.379-405.

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:

• Office of the National Rail Safety Regulator
• Australian Standards technical committee MS-012
• Australian Level Crossing Assessment Model committee
• National Heavy Vehicle Regulator
• Austroads
• Rail Industry Safety and Standards Board

Submissions were received from:

• Office of the National Rail Safety Regulator
• Australian Level Crossing Assessment Model committee
• Austroads
• Rail Industry Safety and Standards Board

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

Glossary

Appendices

Appendix A – Descriptive statistics for level crossing collisions recorded in National rail safety database

Collisions

Appendix B – Summary of sighting distance provisions in Australian Standard AS1742.7:2016

The design criteria for level crossings in Australia are specified the Australian Standard AS 1742.7:2016, Manual of uniform traffic control devices. Part 7: Railway crossings. This standard prescribes the sighting distances required for drivers approaching a level crossing, to enable them to safely identify if it is safe to enter and proceed through a crossing.

The standard seeks to ensure that the driver of a design vehicle will have sufficient time and distance to identify a crossing ahead (S₁) and stop after identifying a requirement to stop. In the case of active control crossings, this would be the detection of flashing lights and/or a lowered boom gate, whereas in the case of a Stop sign controlled crossing, this would be the presence of the Stop sign.

The standard also specifies the minimum distance for a driver to approach the crossing, look along the rail line to check for trains, and then continue and cross safely ahead of any undetected trains (S₂). This is applicable in the case of Give way controlled crossings, where the driver is not required to stop unless a train is detected.

For vehicles stopped at a level crossing, the standard specifies the distance required to be visible along the track from the ‘stop line’. The distance (S₃) is required to be sufficient for the vehicle to start up and clear the crossing prior to the arrival of any train beyond the visible distance.

These distances are illustrated in Figure 6.

Figure 6: Illustration of level crossing sighting distances specified by AS 1742.7:2016

These are expressed in the standards as

These formulae contain variables relevant to the physical characteristics of heavy vehicles. Specifically:

In the case of heavy vehicles, AS 1742.7 specifies values for design vehicles for defined categories of heavy vehicles, with greater vehicle length, slower braking and acceleration, and longer reaction times expected for heavier vehicles.[31] The effect of this is that for crossings expected to carry heavy vehicles, AS 1742.7 requires longer sighting distances are provided.

The ATSB modelled the distances required by AS 1742.7 using a standard case of a crossing on a flat, sealed road, where the road speed was 100 km/h,[32] the track speed was 110 km/h and the road to track angle was 90°. The calculated stopping distances for different types of heavy vehicles are shown in Table 18.

Table 18: AS1742.7 calculated stopping distances for heavy vehicle types

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 Coat of Arms, ATSB logo, and photos and graphics in which a third party holds copyright, this publication is licensed under a Creative Commons Attribution 3.0 Australia licence. Creative Commons Attribution 3.0 Australia Licence is a standard form licence agreement that allows you to copy, distribute, transmit and adapt this publication provided that you attribute the work. The ATSB’s preference is that you attribute this publication (and any material sourced from it) using the following wording: Source: 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.

Executive summary

What happened

In August 2021, during the COVID-19 pandemic, Formosabulk Clement diverted to Brisbane, Queensland for crew change, to be completed to/from a launch via the combination pilot ladder. At the time, Queensland was the easiest jurisdiction in Australia for ships to conduct crew change due to the State’s extensive quarantine arrangements for transporting and accommodating seafarers. By 1833 on 9 August, 8 persons and their luggage had been successfully transferred from the launch PT Transporter to the ship located at an outer anchorage. The launch skipper then drove the launch a short distance off the ship for a short break.

By this time, the ship had turned about the anchor and the boarding side of the ship came under the influence of the forces of wind and sea. After receiving supervisor advice to attempt the transfer again, the launch skipper took the launch back toward the boarding area to assess whether the conditions were suitable. On board Formosabulk Clement, the departing chief mate (dCM) saw the approaching launch and went down the access ladders and onto the vertical pilot ladder in anticipation of boarding. At about 1838, as the dCM waited on the ladder and the launch came close alongside, a large wave passed down the side of the ship and lifted the launch higher than expected. The dCM was struck by the launch and knocked into the sea. Despite being quickly recovered on board, they received fatal injuries.

What the ATSB found

The investigation found that the open water conditions at the anchorage were marginal but, within a lee created by the ship, they were suitable for the transfer to occur. However, as the skipper brought the launch alongside to assess conditions, the transfer area was no longer in a lee and waves were acting against the ship’s side, increasing in height as they passed along it.

Coincident breakdowns in communications between, and within, the ship and the launch resulted in the ship’s master, the dCM, the launch skipper and the launch deckhand having different understandings of the significance and intent of the launch returning alongside the ship.

In addition, the dCM went down the boarding ladders in preparation to board the launch without the knowledge or agreement of either the ship’s master or the launch skipper. The dCM was then in a vulnerable position when the launch was lifted on the higher than expected wave.

The investigation also found that plans and details of how the transfer was to be controlled and progress when the launch was alongside were not adequately shared between all parties in the time available as the launch approached the ship. Neither the shipping company nor the launch company had in place procedures to ensure that such information was shared before the ship arrived at the anchorage.

As a consequence, different interpretations and misunderstandings of the plan and expectations of the people involved in the transfer, particularly those in positions of influence over the progress of the transfer, were not identified and addressed.

What has been done as a result

Following the accident:

• the launch service provider (Pacific Tug) ceased crew transfer operations until an investigation into the accident and assessment of transfer operations were completed
• Maritime Safety Queensland (the State’s regulator) amended the COVID crew change procedure for vessels at anchor including limiting crew changes to risk assessed conditions and daylight hours only
• Formosabulk Clement’s operator, Formosa Plastics Marine Corporation, completed investigations and held multiple safety meetings and training exercises to share details of, and lessons learned from, the accident.

Subsequently, Pacific Tug updated personnel transfer arrangements and procedures. This included developing and implementing a traffic light system (red, amber and green statuses) for operational assessment and control. The system set operational requirements and limits, including weather, against which the conduct of the transfer task was continually assessed. Light sets, mounted on the launch, are used to show the status of the transfer to all involved, on the launch and the ship. A red status requires cessation of the transfer operation and complete reassessment prior to any further attempt to complete the task.

Safety message

All parties are reminded of the importance of maintaining active and adaptable communications before and during the planning, co-ordination and control of a complicated task.

Further, where possible, all parties should share plans and information well before undertaking a task so as to allow all involved to have a common and complete understanding of the planned activity.

The occurrence

During the COVID-19 pandemic, the international maritime industry struggled to comply with requirements for the maximum continuous period that a seafarer could serve on board a ship without taking leave.[1] Following a period of grace during which the continuous service period was relaxed to 14 months, in November 2020, the Australian Maritime Safety Authority (AMSA) notified ship owners, operators and seafarers that it would resume enforcing the 11‑month continuous service limitation from 28 February 2021.

On 25 July 2021, Formosabulk Clement (Figure 1) departed Mailiao, Taiwan, in ballast, bound for Newcastle, New South Wales to load coal. The ship’s trade routes for the preceding 6 months had provided no opportunity for change of crew and 14 of the 25 crew on board had been on board for more than 11 months.

At that time, pandemic-related restrictions on crew changes varied considerably between different States in Australia with Queensland offering the most convenient opportunities. Hence, in early July, Formosabulk Clement’s manager, Formosa Plastics Marine Corporation (FPMC), engaged its local shipping agents in Australia to arrange a crew change off Brisbane, Queensland. The arrangements would involve the ship calling at Brisbane anchorage to rendezvous with a crew change vessel (launch). Eleven crewmembers were to join the ship and 15 were leaving.

Figure 1: Formosabulk Clement at anchor, PT Transporter alongside

Source: AMSA, Pacific Tug with modifications and annotations by ATSB

Arrival to anchor and preparations

At 0912[2] on 9 August, Formosabulk Clement anchored in the port of Brisbane outer anchorage, about 6 miles[3] east of Point Cartwright. Ashore, pandemic protocols had been completed for the joining crew and arrangements for transfer by launch had been made. The launch service provider (Pacific Tug) notified all parties that, to satisfy requirements, the transfer would require 2 trips by the launch PT Transporter. The launch was to collect the passengers and depart from the quarantine facility at Brisbane Rivergate marina and shipyard.[4]

Figure 2: Starboard side boarding site and ladder arrangement

Source: AMSA, FPMC and Pacific Tug annotated by ATSB

Pacific Tug requested that Formosabulk Clement be relocated further south in the anchorage area to assist the transfer, and with Brisbane vessel traffic service (VTS) approval, the ship was moved south. During this same information exchange, the ship’s master was requested to have the boarding arrangements (Figure 2) located on the opposite side of the ship to the deployed anchor and to provide a good lee for the transfer.

At 1436, Formosabulk Clement’s port anchor was let go and at 1500 ‘finished with engines’[5] was rung. The ship was now anchored about 5 miles east of Caloundra Head, about 7 miles south of the original anchor location. It was riding easy on the anchor, 7 shackles in the water in a fresh breeze and moderate seas, from the south-east.[6] Throughout the afternoon, the ship yawed about the anchor, at the whim of the conditions, its heading varying between about 140° and 80° in about 30-minute cycles.

At 1448, PT Transporter departed the Rivergate marina and shipyard with 2 crewmembers (skipper and deckhand) and 8 Formosabulk Clement joining crewmembers on board. At 1530, it passed to the west of the entrance beacons, 30 miles from the ship, heading north-east making better than 15 knots across Moreton Bay. The passage across the bay toward the ship remained uneventful other than for several of the passengers becoming seasick in the choppy conditions.

At 1600, on board Formosabulk Clement, the departing chief mate (dCM) completed handover of rank to the trainee chief mate who had been on board since early May and was now taking over the position. The new chief mate then took the navigation watch from the second mate.

The dCM was to leave the ship on the first launch after supervising the embarkation of the 8 joining crew (and their luggage) and the disembarkation of 7 other crewmembers. In agreement with the master, the plan was for the dCM to be the last person to disembark. The dCM and the master (on the bridge) would be using handheld UHF radios to communicate.

After the evening meal, crewmembers assembled and arrangements for the departure were finalised. At the boarding area on the starboard side, the bosun, one ordinary seafarer and one deck cadet were to be assisted with conducting the transfers by the departing, and then joining, crewmembers, overseen by the dCM. The personnel gathered on deck, with their luggage, and prepared to assist the boarding crewmembers.

At 1725, as PT Transporter approached Formosabulk Clement, the skipper called the ship on the port’s working channel (VHF channel 12) and advised that the launch was 15 minutes away. The skipper confirmed the transfer would be done on the ship’s starboard side and requested that a lee be provided. The ship’s new chief mate confirmed the boarding side and the lee. Subsequent communications generated confusion between the launch and the ship regarding the lee. The launch skipper then requested assistance from the launch passengers and from 1730 communications between the launch and the ship commenced in Chinese. At about 1740, the launch slowed and manoeuvred along the starboard side of the ship.

Transfer to the ship

At 1745,[7] Formosabulk Clement was at one extreme of yaw about the anchor with the starboard side exposed to the weather (heading 089° with weather from 100°— Figure 3). The launch skipper recalled there was a little bit of sea from the east-south-east and that it ‘wasn’t the calmest of conditions’ (noting that several of the launch passengers had been sick during the journey). The ship’s master recalled observing similar conditions at the time with 2 to 2.5 m wave heights. The ship’s stern began to swing to port and conditions along the starboard side began to ease.

Nearing 1800, the ship was at the other extreme yaw angle about the anchor and was on heading 142° with the boarding area sheltered. The master radioed the launch and informed the skipper that the wind was now from the port side and it was suitable to ‘try once again'.

Figure 3: Ship yaw leading up to the accident

Source: ATSB analysis of AIS data from AMSA

On board the launch, the skipper manoeuvred alongside, keeping parallel to the ship’s side with little if any force applied to the hull while keeping the boarding ladder in position on the foredeck of the launch. As the conditions allowed, the transfer commenced with each person guided, one at a time, forward and onto the ladder by the deckhand. Baggage was transferred by line from the after deck.

The weather conditions began pushing the ship’s stern to starboard and the master requested standby on the main engine and at 1806 the engine was ready for manoeuvring. The stern continued to swing to starboard and by 1810 the ship had swung back to heading 095°. The master used dead slow ahead commands in conjunction with rudder movements to drive the stern around and turn the port side of the ship against the weather, thereby providing shelter to the boarding area.

By 1825, the ship had been manoeuvred to heading 167°, creating calm conditions at the transfer site. The transfer continued. At about 1820, the ship’s main engine was stopped and the ship continued to yaw under the influence of the weather. The main engine was used again for a short period and by 1833 the transfer on board was complete and the ship had swung back to 097°.

At the completion of the transfer to the ship, the launch skipper drove the launch away from the ship’s side to rest and reassess the conditions. The launch was stood off a short distance for several minutes. The skipper reported feeling that conditions were marginal and of having contacted the Pacific Tug supervisor for advice. The advice received was to ‘go and have another look and request a lee’. The skipper recalled advising the ship’s master to ‘stand by’ and to provide a lee. The deck hand was also advised to ‘stand by’. The view from the launch’s wheelhouse was somewhat restricted, especially up the side of the ship, and while standing off, the skipper used the launch’s spotlight to view the boarding area. The skipper recalled seeing one person leaning on the rail about midway down the accommodation ladder, and no-one on the boarding ladder.

The skipper then manoeuvred PT Transporter alongside Formosabulk Clement with the intention of assessing the suitability of conditions at the boarding ladder for continued operations.

The accident

Following the brief rest period, the launch was taken back in toward the boarding ladder. The skipper intended to see how the launch sat alongside as the ship yawed and the boarding area became more exposed to the conditions. The skipper also expected to provide approval to the ship before anyone left the deck and did not anticipate that someone would be on the ladder.

Onboard Formosabulk Clement, the master confirmed via radio with the dCM that the boarding phase was completed. At this time, the master could see the dCM on deck and expected that disembarkation would not commence until approval was given. The boarding area on the starboard side was now exposed to the weather and the master was aware that a lee would have to be made on that side. This required the master to manoeuvre the ship to make a lee and then convey this plan to the launch and to the dCM. However, while attempting to communicate with the launch, the master could see it returning.

As the launch approached, the dCM went down the accommodation ladder and onto the pilot ladder. One of the departing able seafarers (AB1) followed onto the accommodation ladder. The dCM was wearing multiple layers of clothes, including a jumper, along with a small document wallet type backpack under a self-inflating buoyancy vest and was carrying a ship’s hand-held radio.

As the launch closed in, the dCM began climbing down the pilot ladder. The master was preoccupied with contacting the launch and remained unaware that the dCM (or AB1) had left the deck and gone onto the ladder(s).

As the launch closed on the ship, the deckhand saw the dCM several rungs down the pilot ladder and descending. As the launch was not yet safely alongside, the deckhand directed the dCM to climb back up. The dCM climbed up until adjacent to the accommodation ladder lower platform.

Soon thereafter, the launch came close alongside and the dCM began to descend the pilot ladder a second time. The deckhand felt the conditions were suitable and did not direct the dCM away. The launch skipper was unaware of the presence of the dCM on the pilot ladder or of the interactions between the deckhand and the dCM.

At about 1838, suddenly a larger wave lifted the launch higher than had to that time been experienced (Figure 4). The skipper sighted the dCM on the ladder as the launch rose, but with insufficient time to take any avoiding action. The deckhand shouted a warning for the dCM to get clear, however, the launch rose up, trapping the dCM between it and the side of the ship. As the launch then came down on the wave, the dCM fell into the water. The dCM’s lifejacket inflated, keeping them afloat.

On board Formosabulk Clement, the master was still attempting to communicate with the launch when notified by the bosun that the dCM had fallen into the sea.

The skipper manoeuvred the launch clear of the ship and the dCM. Once clear, the skipper and deckhand went about rescuing the dCM who was soon retrieved on board PT Transporter. The dCM was seriously injured and the launch crew administered first aid, including CPR. The skipper notified the ship’s master that the dCM had been retrieved but was unconscious.

In the following minutes authorities including VTS, the harbour master and the shipping agents were notified. The emergency authorities were called and police and ambulance personnel were deployed. The launch was directed to take the dCM to Mooloolaba, about 10 miles to the north-west, to meet with the emergency services.

At 1905 PT Transporter departed the area for Mooloolaba. The launch arrived there at 1955 and met paramedics. The dCM was examined and provided treatment but was pronounced deceased soon thereafter.

Figure 4: Approximate positioning of the launch and personnel at 1838

Source: AMSA, Pacific Tug with modifications and annotations by ATSB

Post-accident

All further operations were ceased and both vessels were directed by AMSA to remain in their existing locations pending an accident investigation. During the following days, Formosabulk Clement was detained by AMSA on the grounds that the required boarding arrangements for safe means of access were defective and did not comply with regulations. This notice, along with action by the port authority, prevented personnel from boarding the ship from the water.

Consequently, alternative arrangements were made to complete the still outstanding crew change, which was completed by helicopter on 13 August. The detention order was subsequently lifted and the direction notice closed. At 1530 that day, standby was called and Formosabulk Clement departed the anchorage and continued its onward voyage to Newcastle.

Context

Formosa Plastics Marine Corporation

Formosa Plastics Marine Corporation (FPMC) commenced operations in 1980 with 2 chemical tankers to service the needs of its parent company Formosa Plastics Group. By 2022, FPMC operated a fleet of 48 vessels, including 18 bulk carriers, with total deadweight of 5.9 million tonnes. FPMC ships regularly traded to Australia, calling at ports including Brisbane and Newcastle.

Formosabulk Clement

At the time of the accident, Formosabulk Clement was owned by Pilot Maritime Company (Liberia) operated by the Formosa Plastics Marine Corporation (FPMC, Taiwan) and registered in Liberia. The ship was classed with Bureau Veritas.

Formosabulk Clement had regularly visited Newcastle, with 5 port calls in the 12 months prior to trading between Taiwan and Russia between the period of April 2021 and the accident voyage. The ship had completed one previous crew transfer off Brisbane, by launch, in December 2020. On that occasion, 4 persons had disembarked (none boarded).

Following this accident, Formosabulk Clement was sold and renamed Goody.

Crew details

Formosabulk Clement had a multi-national crew of 25, including the relieving chief mate (scheduled to take over during this crew change), 3 deck cadets and 2 engineer cadets. The master and departing chief mate (dCM) were from China and the remaining crewmembers were from China, Taiwan and India.

The master first went to sea in 1986 and obtained their master’s seagoing qualifications in 1999. They had worked for FPMC since 2003 as master within the bulk carrier fleet. This was the master’s first time on board Formosabulk Clement, having joined in June 2020.

The departing chief mate (dCM) first went to sea as a deck cadet with FPMC in 2007. This was their second time on board Formosabulk Clement, both times as chief mate. They joined the ship in June 2020.

Of the 15 crewmembers scheduled to depart Formosabulk Clement during the call to Brisbane, 14 had joined the ship in June 2020 and had been on board for 409 days.

All crewmembers interviewed reported having used a pilot ladder previously. The least experienced had joined the ship by pilot ladder and stated that more experienced crewmembers gave advice and reassurance when preparing on this occasion. More experienced persons had used pilot ladders multiple times on many different ships. All stated that training was given and discussions held prior to the transfer. Training was also provided at shore-based marine schools.

Crew change procedures

The FPMC fleetwide safety management system (SMS) included several procedures which related to conducting personnel transfer between a ship and another vessel while at anchor in open waters.

Personnel transfer at sea procedure

The ‘Personnel transfers procedure at sea’ advised that crew change at sea was to be undertaken as far as possible in favourable weather conditions, with good visibility, and preferably during daylight. Transfer activities were to be suspended and reported to the superintendent if the weather deteriorated.

The procedure required that, prior to the transfer, a risk assessment and a briefing meeting for the task were to be conducted. The briefing meeting was to include discussion of:

• the launch company’s risk assessment for personnel transfers. The documentation to be received was to include details of the suitability of the launch, details of its life-saving equipment and confirmation of the responsible officer on board the launch
• limiting weather conditions: wind <17 knots, swell <1.25 m, visibility >5 miles
• the requirement for approval by the masters or skippers of both vessels and the persons being transferred, for the transfer to take place.

Personnel to be transferred were to be fully aware of and understand the sequence of events including that only one person was allowed on the ladder at a time. They were also to understand the requirements for and use of any equipment including personal protective equipment (PPE) and clothing, including lifejacket, footwear and safety harness.

The procedure also advised that the responsible officer should not be one of the departing or joining crew.

Furthermore, the procedure advised that good communications be established with the launch and that during the transfer ‘There should also be a clear understanding of the meaning of all terms used in the transfer’.

Risk assessment

A risk assessment for ‘Transfer of personnel to and from small vessels’, involving 14 crewmembers, was authorised and dated 9 August 2021.

The assessment included the risk of personal injury due to a fall. This was mitigated by measures including not transferring in heavy weather (not defined in this document), that all crew were to be briefed on the transfer with the sequence explained, step-by-step, and ensuring good communication with the launch to co-ordinate the task.

The risk of crush injury due to the launch or other equipment was to be mitigated by briefing and training of the crew including confirmation of the sequence of activities.

Also identified was the risk that the gap between the ladder and the launch combined with the motion of the launch may lead to a person falling into the sea. This risk was to be addressed by good communications and co-ordination of the task along with sufficient lighting and personnel wearing the required PPE.

The assessment concluded that the level of risk was tolerable and no ‘additional measures’ to reduce risk or further assessment were required.

Crewmember training—pre-transfer briefing

A record (on FPMC form titled ‘Training Record’) was kept of a meeting held a week before arrival to discuss the transfer. Dated 2 August and signed by all crewmembers, the subject of the training record was ‘Personnel transfers procedure at sea’.

This one-page record included some of the points raised in the personnel transfer at sea procedure, such as providing a lee, having clear communications, completing a risk assessment, and details of lighting and safety equipment. However, no mention was made of the launch company’s documentation or expectations, limiting weather conditions or of signals and terms to be used.

Permit to work overside

A permit to work overside was completed on 9 August 2021, valid from 1530 that day. The permit identified the dCM as the person assigned to the work with the bosun as the team leader at the worksite. Twelve others, including those scheduled to depart, were signed-on to the permit as other persons.

This permit contained general applicable items such as PPE, communications, lighting and consideration of the weather conditions. The permit indicated that all requirements were in place, including that the weather was considered suitable for safe work. However, it did not make specific mention of limiting conditions, personnel transfer activities or communication and signalling protocols.

FPMC risk assessment procedures

The FPMC SMS risk assessment procedures document was to be a source of information and guidance for the use and implementation of risk assessments. The document outlined risk assessment philosophy and the company process. Guidance and suggestions were provided for each stage of the process. The procedure also advised that any person had the authority to stop the work should a condition or behaviour be perceived as unsafe.

The procedure required that all personnel performing a risk assessment on board were to be appropriately trained in the process.

FPMC permit to work procedures

The FPMC ‘Permit to work procedures’ required that, as part of the risk assessment process, a permit to work (PTW) was to be issued before any work was commenced. FPMC had a suite of 13 PTWs for use on board. A permit for work overside was required for any work which required crewmembers to work outboard of the ship’s railings, including preparation of the combination pilot ladder.

The PTW procedures also required that a toolbox talk be completed for all tasks. This was to be held by the person in charge, at the worksite, with all involved, before the work began.

FPMC crew change logistics

The ATSB sought advice from FPMC regarding crew change issues and requirements during the pandemic and the plan to change crew off Brisbane. The response received included:

• Joining personnel were to complete any special requirements before departing crew ended their contracts. This included visa requirements as well as pandemic requirements such as PCR[8] testing and quarantining.
• COVID-19 pandemic restrictions meant that, in the preceding 6 months, crew were denied entry into Russia or Taiwan. AMSA enforcement of MLC time on board limits meant that crew change needed to occur prior to or at Newcastle.
• Approval for crew change was not obtained from New South Wales (NSW) Health, and, FPMC had found, on previous occasions, that crew change in NSW was difficult.
• Queensland had put in place a COVID-safe plan which allowed crew changes to be completed and Brisbane was selected for the crew change on this basis.
• Crew change arrangements were made by a local Brisbane agent. At that time, arrangements included COVID-safe transport, accommodation and testing, in addition to the usual logistics of launch provision, notification of the required agencies for port, customs, immigration and biosecurity requirements.
• Earlier crew change operations for FPMC vessels had showed launch transfers to be significantly more cost-effective than other means, such as helicopters. In addition, the need to disembark 15 persons and embark 11 made launch transfer the preferred choice for Formosabulk Clement.
• FPMC also advised, that at the time, there were no restrictions on timing for the crew change with limitations only due to weather. The vessel position was guided by the port and advice from the launch company.

Boarding arrangement

At the time of the accident, freeboard at the boarding location was about 16 m (Figure 2). The combination pilot ladder was arranged with the accommodation ladder section at maximum angle leaving more than 5 m from the bottom of the ladder to the waterline. The vertical, rope pilot ladder was deployed from deck with the bottom step positioned about 1.4 m above the water.

Following the accident, PSC inspection found the pilot ladder defective, as, among other issues, the lower spreader was cracked (possibly due to contact from the launch during the transfer). The ship was detained by AMSA under the Navigation Act 2012, in part, due to deficient safe means of access to the vessel. The rope pilot ladder was replaced prior to the ship sailing from Australia and the deficiency was closed.

Lighting for the boarding area was provided by a portable floodlight mounted over the side at deck level and floodlights mounted on the ship’s starboard bridge wing (Figure 5). Deck lighting was also used to illuminate the work area on deck. Sunset was at 1723, and the ship’s crewmembers reported that there was sufficient lighting for the task.

Figure 5: Ship lighting for the starboard boarding area

Source: AMSA with annotations by ATSB

Recorded data

Formosabulk Clement was fitted with a simplified voyage data recorder (S-VDR)^[9] designed to collect and store data from various shipboard systems in compliance with regulations.

This system required crew action to ensure the data was saved following an incident where power was not lost. However, this procedure was not followed on this occasion and consequently the recorded data was not available to the ATSB. Bridge audio, radar data and rudder orders were not recorded.

Engine orders were recorded via the telegraph logger and printer fitted in the bridge console.

Pacific Tug

Pacific Tug (Aust) was a family owned and operated Australian registered company based in Victoria Point, Queensland, with operations in the Brisbane River and around Australia. The company began in 1965 and provided marine services including towage, salvage and personnel transfer. In 2022, the company had a fleet of 25 tugs, barges, support and crew transfer vessels.

Procedures

Pacific Tug maintained a company-wide operations manual within an integrated management system. This system included multiple procedures relevant to personnel transfer between vessels.

Risk assessment

The purpose of the risk assessment document was to provide clear and defined processes for the preparation and execution of a risk assessment at Pacific Tug workplaces. All workers involved in a task were required to actively participate in a risk assessment.

Risk assessments were required for activities which included new or non-routine tasks and prior to any complex task as required by legislation, regulations, standards and codes of practice.

Two types of risk assessments were described:

• A job safety environment analysis (JSEA) which detailed step-by-step how a task was to be carried out safely. The analysis considered:
  • tasks—a step-by-step list of the basic activities of the task
  • hazards—a list of potential hazards at each step of the task
  • control measures—step-by-step instruction on how to safely carry out the task by controlling each identified hazard.
• A safe work method statement (SWMS) was also available. This statement detailed step‑by‑step how a task was to be carried out safely. It differed from a JSEA in that the SWMS was required, under Australian safety regulations, for all high-risk construction work.

In addition to other activities, Pacific Tug utilised PT Transporter for construction activities and both types of risk assessment were completed for the vessel. Personnel transfer between vessels was an activity for which a risk assessment was required.

Vessel transfers

The Pacific Tug fleet operations vessel transfers procedure aimed to provide clear and defined processes for boarding and disembarking from a vessel. The procedure outlined the master’s responsibilities and then provided guidance on transfer of personnel, including the importance of maintaining 3 points of contact and movement of personnel between a vessel and fixed wharf.

The master (skipper) was to ensure the safety of personnel during the transfer to and from the vessel. If the transfer was between vessels, the masters were to discuss and reach agreement beforehand regarding the person in charge of the activity (and on which vessel), with one person in charge at any one time.

Workers being transferred were to be briefed prior to the transfer and were to be familiar with the method of transfer. The procedure also advised, among other things, that:

• Before the transfer operation, the master was to designate who was responsible for the transfer. This designated crewmember was to be in charge of the actual transfer of the worker. No worker was to transfer onto, or off, the vessel until that crewmember considered it safe to do so and indicated that transfers may take place. Any doubt was to be resolved in consultation with the master before the transfer proceeded.
• A clear understanding of the meaning of all terminology used in the transfer should be established between all those taking part.
• Clear communications were to be established and maintained between the person conning the vessel and those assisting in the transfer.
• Only one person was permitted to transfer at a time.
• The vessels (where appropriate) should be manoeuvred to create a suitable lee and provide protection to the transfer site.

Permit to work

The Pacific Tug permit to work (PTW) procedure was to provide clear and defined processes for each type of activity requiring a PTW. A PTW was required for any non-routine activities and any activity deemed high risk (defined in the procedure as ‘An activity that is deemed to have potential risk or has hazards that may injure a worker or the potential to injure workers.’). Permits existed for work aloft or over the side and for lifting people, but not specifically for personnel transfer.

The PTW system was to identify, plan and control any hazardous tasks and promote worker accountability and responsibility for the task to be undertaken. Each PTW was to describe the high-risk work to be done and the precautions to be taken while doing it. It also highlighted the conditions to be met for the PTW and hence for the work to proceed.

The risk assessment completed for personnel transfer by launch did not identify that a separate permit to work was required.

COVID safe plan

In compliance with relevant legislation, Pacific Tug completed requirements put in place to respond to the COVID-19 pandemic. This included being a Queensland Government endorsed transport provider with an approved transport plan in place and Statement of Compliance for PT Transporter.

Customer interaction

Pacific Tug had processes in place for corresponding with customers (in this case this was via the agent) which included providing a package of standard information. This package included a schedule of rates, vessel specifications sheet and pandemic response requirements. It did not include Pacific Tug procedures or operational requirements relating to details of the transfer of persons from one vessel to another.

PT Transporter

PT Transporter (Figure 6) was an aluminium catamaran, built in 1996 and refurbished in 2019 with machinery and fit out changes and upgrades, including main engines and generator. The vessel had a cruising speed of 20 knots and was 11.9 m long, with a breadth of 4.80 m and draught of 1.20 m. It was an Australian domestic commercial vessel (DCV) regulated under the Marine Safety (Domestic Commercial Vessel) National Law Act 2012.

PT Transporter was in survey to class 1C,[10] certified to carry 2 crew and 23 passengers. It was designed, and predominantly used, for personnel transfer with the foredeck purpose-built for such. Since October 2019, PT Transporter had completed 51 personnel transfer operations, 39 of which were to outer anchorages.

Figure 6: PT Transporter

Source: Pacific Tug, annotated by ATSB

The vessel’s helm position was located to starboard of the centreline. The cabin layout combined with the foredeck awning limited the skipper’s view upwards, especially in close quarters such as when the launch was alongside the ship (Figure 7).

Crew

PT Transporter was certified with a complement of 2: master with minimum coxswain grade 1 certificate and an uncertificated deckhand.

PT Transporter’s master (skipper) had more than 15 years’ small boat experience in various locations around the coast of Australia. They had worked on several major construction projects as a crew boat skipper. At the time of this accident, they held valid certification as master <35 m near coastal and marine engine driver grade 2 NC (near coastal).

The skipper was employed with Pacific Tug on a contract/casual basis beginning in 2010 as a general-purpose deckhand with short periods as master of PT Transporter. In 2019, after time away, they were re‑employed as master of PT Transporter with stints in other vessels as deckhand and mate.

The deckhand commenced casual employment with Pacific Tug in June 2021 and had served in 4 vessels (total 8 days service), including the day of the accident. On 3 previous occasions the tasks had involved personnel transfer, all on board PT Transporter. The deckhand held a valid AMSA certificate of proficiency as an integrated rating.

Both crewmembers held a valid AMSA certificate of safety training. This was the first time they had worked together.

Figure 7: The view from PT Transporter’s skipper’s position at the helm

Source: Pacific Tug annotated by ATSB

PT Transporter specific procedures

The Pacific Tug operations manual included several procedures specific to the operation of PT Transporter.

Transfer of personnel job safety and environment analysis (JSEA)

A job safety environment analysis (JSEA) was completed in April 2020 for PT Transporter for the task of personnel transfer. This document identified that no permit to work was required and then identified hazards and risk control measures for steps within the task. Users were directed to the fleet procedure relating to personnel transfer.

Identified risk control measures included:

• clear communications
• assess vessel movement before commencing a transfer
• ensure sufficient lighting
• maintain communication between vessels at all times
• both launch personnel to monitor the sea conditions at all times
• the transfer was to proceed only after the skipper had given the ‘all clear’
• a lee was to be provided as necessary.

Both the skipper and deckhand of PT Transporter had signed multiple times that they had read and understood this document, with the most recent sign-offs less than 2 weeks before the accident.

General vessel operations job safety and environment analysis

This JSEA was for PT Transporter general operational activities. The risk assessment covered 14 identified activities, several of which were relevant on 9 August. In addition to general considerations, this included arrive at / depart another vessel, transiting, pushing up to another vessel, and transfer of personnel and belongings.

Hazards for each activity were identified, as were control measures. Common to the control measures were ensuring effective communication and ensuring all personnel understood the activity and their role in it.

The JSEA advised that for all vessel activities an ‘Operational Risk Assessment’ was required.

Safe work method statement (SWMS) for letting go / making fast—wharf or vessel

A SWMS risk assessment was completed in April 2020, for PT Transporter operations associated with construction activities identifying hazards and control measures related to the launch making fast to, or letting go from, another vessel. While not expressly applicable to the launch being manoeuvred alongside the ship, this risk assessment showed that relevant risk control measures were considered and applied to operations of this vessel. These included ensuring clear communication and instruction on the task sequence and between vessels, consideration of weather and sea conditions, and proceeding with the task under instruction of the master (skipper) and as per the planned sequence.

PT Transporter vessel safety training manual

This manual focussed on vessel‑specific safety training, equipment and firefighting arrangements with few references to operational procedures. Users were advised that this manual was in addition to the fleet operations manual which contained fleet standards and operating procedures including permits to work, work planning and hazard analysis.

PT Transporter Wellbeing Plan

This document outlined actions Pacific Tug would take to provide a consistent approach for the management of Safety, Health and Environmental (SHE) requirements. The document included details of crewmember responsibilities along with guidance on addressing the COVID-19 requirements at the time. To comply with Queensland Health guidelines, PT Transporter could accommodate 9 passengers and 2 crew.

PT Transporter COVID safe transport of quarantined persons plan

The Queensland Government’s Public Health Directions required that quarantining persons were moved to and from nominated quarantine premises by dedicated modes of transport. This required that the transport provider was endorsed by a government authority, with a Transport Plan in the form approved by the State’s Chief Health Officer. This was in addition to the in-house PT Transporter wellbeing plan.

Under PT Transporter’s approved transport plan, the vessel could safely transport 2 crew (housed in the cabin) and 13 passengers separately accommodated in the after deck seating area.

Daily pre-shift work plan

A daily pre-shift work plan for the vessel was completed on 9 August for a ‘crew run at the outer anchorage’. The work planned was for 8 persons on, 8 off Formosabulk Clement at the outer anchorage, departing from Rivergate Marina. This plan did not expand on this information or make mention of any related risk assessments or permits to work in the columns provided for such. Both crewmembers had signed the plan.

Provisional Pacific Tug procedures

As part of the company continuous improvement programme, at the time of the accident, an operational risk assessment procedure was being trialled for PT Transporter general operations. This provisional procedure included:

• A ‘Traffic light assessment’ system and criteria for operations. This system set the conditions used to determine the state of operations—safe, stop and assess, stop immediately and for re‑starting. This included environmental limitations:

• Operational requirements, which were separated into when the launch was approaching the ship and when alongside. Both situations mentioned communications between vessels and personnel. While approaching the ship the transfer process was to be communicated to and understood by transferring persons.

Weather

From arrival off Brisbane and throughout the day and into the evening of 9 August, the ship’s navigation watch officers recorded weather conditions as winds from the south-east at force 5[11] (17 to 21 knots, fresh breeze), moderate seas (1.25 to 2.5 m) on 3 m swell with good visibility and overcast conditions.

At the time of the transfer, the master observed that conditions were as above with waves about 2 to 2.5 m and a lee would have to be made for a safe transfer. Testimony of crewmembers on deck at the time was that conditions were windy and that the launch was noticeably moving.

Earlier in the day, as the launch skipper took command, the departing skipper reported that conditions on the bay were ‘not very nice’. Transit across the bay to the ship was made at better than 15 knots and while several of the passengers were sick, the launch crew were not and attributed this to the passengers’ being less experienced in small boats.

Once at the ship, the skipper observed that, with the boarding area exposed to the weather, the conditions were choppy and uncomfortable. The skipper requested the ship (master) provide a lee for the area. As the ship moved and when manoeuvred about the anchor, the conditions at the boarding site improved dramatically. The skipper was satisfied that the conditions, though not dead calm, were sufficiently smooth and comfortable to allow persons to board the ladder and went ahead with the personnel and luggage transfer onto the ship.

The deckhand stated that while choppy, the conditions on the bay and at the ship were similar to other times they had been out and that once in the shelter of the ship, the seas calmed and the launch moved up and down comfortably less than a metre.

Several government agencies, and others, maintain weather and sea monitoring equipment in the area around Moreton Bay and the port of Brisbane. Recordings obtained for locations nearby the anchorage position showed no abnormally high waves were recorded at this time. In the hour preceding the accident, significant wave heights varied from 1.1 m to 1.3 m with maximum wave height of less than 2.4 m recorded. The wave direction varied from 97° to 103°.

Winds were recorded as 10 to 20 knots from about 100° (varying from 70° to 120°).

COVID-19 pandemic

The first human case of a new (novel) coronavirus was identified in December 2019 and on 20 January 2020 the World Health Organization (WHO) issued the first situation report in relation to what was to become known as COVID-19. In response, countries implemented strict measures on the movement of people, including quarantine, isolation and lockdown requirements.

The escalating situation had a profound effect on the maritime industry and seafarers. By July 2021, the IMO estimated that some 250,000 seafarers remained on board commercial vessels, unable to be repatriated and past the expiry of their shipboard contracts. A similar number of seafarers urgently needed to join ships to replace them.[12]

Requirements of the MLC[13] limited the maximum continuous period of service on board ship to 11 months. However, as a consequence of the pandemic, in June 2020 AMSA issued a marine notice[14] allowing for extension of this period of service to 14 months maximum.

In November 2020, AMSA notified vessel owners, operators and seafarers that this extension would end on 28 February 2021.[15] At this time, crew change hubs were available in several major Asian ports including Hong Kong and Singapore. AMSA considered that industry had had sufficient time to adjust to the challenges of repatriation, and did not consider difficulties in finding a flight as an appropriate reason for an extension of a seafarer’s service onboard. Consequently, AMSA advised that from that date, the 11-month continuous service limitation as outlined in Marine Notice 17/2016 would be enforced.

The first case of the virus in Australia was reported on 25 January 2020, and on 20 March the international borders were closed to all non-residents and non-citizens and tight restrictions applied to movement of citizens.

Queensland

On 29 January 2020, Queensland recorded its first COVID-19 case. From 30 January, Maritime Safety Queensland (MSQ), via the vessel traffic service (VTS), increased vetting of vessels prior to pilot boarding. Then, on 25 March, border access restrictions were imposed which limited movement between States. In accordance with emergency powers arising from the declared public health emergency, the Queensland Chief Health Officer issued the first Border Restrictions Direction. Under these restrictions, anyone arriving into Queensland from another State or Territory had to self-quarantine for 14 days, unless they were an exempt person.

In July 2020, Queensland implemented the requirement for a maritime crewmember to comply with a comprehensive ‘Protocol for maritime crewmembers joining or signing off a vessel in Queensland’. The protocol contained instructions and requirements for crew joining or leaving a vessel including testing, quarantine and transport. The protocol required that a crew changeover checklist, endorsed by the regional harbour master (RHM), be completed for the joining crew and another for the departing crew. By June 2021 the protocol had been amended and updated to be version 10.

Also, in July 2020, MSQ established a liaison role to work with the Queensland State Health Emergency Co-ordination Centre to manage, among other roles, interaction between the health and shipping requirements. By mid-2021, a network of persons to assist and oversee crew change on behalf of RHMs and health authorities was in place. This included 2 full-time crew change assistants (CCA) based in Brisbane, 2 in regional ports and a hotel liaison officer based in Brisbane to manage accommodation and quarantining requirements.

In Brisbane, a 230-room hotel was designated for the primary use of maritime crew quarantine. From May 2020 to the end of July 2021, more than 9,000 mariners, from almost 1,500 ships, had changed through Queensland ports. About 4,500 had been through the port of Brisbane with less than 500 changed by launch to vessels at anchor.

At the time, a major Australia-based shipping industry representative body considered that crew changes were most easily conducted in Queensland and had provided that advice to members.

Quarantine corridor—crew changes in Queensland

To complete a crew change in Queensland, the shipping company’s representative (usually the local agent) applied to Queensland Health via a ‘Crew changeover checklist’ form. The form was submitted to the RHM who was acting on behalf of health authorities in this regard. In Brisbane, the CCA then assessed the application and the RHM approved the plan. The CCA responded to the applicant with information outlining the requirements to be met prior to, during and after the crew change.

Following the response from the CCA, communication followed in which the details of the crew change were revised, updated and approved as circumstances required. This included all travel and accommodation arrangements regardless of when these occurred—the first joining personnel for Formosabulk Clement arrived into quarantine in Brisbane on 13 July, almost 4 weeks before the ship arrived.

In this way, a ‘quarantine corridor’ was set up and maintained about any ship personnel travelling through Queensland. The CCAs kept oversight of the planning and progress. They ensured that plans were in place which ensured that a person remained quarantined from the point of arrival into Queensland until they boarded the ship, and vice versa. All steps within the corridor were assessed and approved. This extended to the CCA ensuring that all requirements were in place (for example, PCR testing and quarantine periods) to enable a departing person entry into their destination country.

This system thereby provided the flexibility and security to enable continuing crew changes within the dynamic circumstances created by the pandemic.

COVID safe transport plan

Queensland Government’s Public Health Directions required that quarantining persons were moved to and from a nominated quarantine premises by dedicated modes of transport. In certain circumstances, quarantined persons were required to travel with a transport provider endorsed by a government authority, with a Transport Plan in the form approved by the Chief Health Officer. To fulfill this requirement, transport providers developed a Transport Plan using a template and sought endorsement to operate from the Department of Transport and Main Roads (DTMR).

The purpose of the Transport Plan was to document the practices that would be applied by transport providers to keep themselves and the community ‘COVID safe’. It contained the minimum standards of practice that must be met by transport providers when offering a service to quarantined persons.

Pacific Tug was a Queensland Government endorsed transport provider with an approved transport plan in place and Statement of Compliance for PT Transporter.

Safety analysis

Introduction

In 2021, the COVID-19 pandemic and the response of international shipping to manage it created challenges for ship crew changes. In the case of Formosabulk Clement, this resulted in the ship diverting to Brisbane to conduct a change of long‑serving crewmembers on 9 August 2021. With the ship at anchor, the crew change was being conducted via a combination pilot ladder from the crew transfer launch, a much smaller vessel about 16 m below the ship’s deck.

The transfer was conducted in the early evening on the ship’s starboard side, the leeward side from time to time as the ship yawed at anchor in the rough sea conditions. The ship’s main engine was also used to maintain a lee and, by 1833, the 8 joining crewmembers (and their luggage) had safely boarded the ship.

The launch then moved clear of the ship to wait for the transfer of crew from the ship. While waiting, the launch skipper felt the seas become rougher and, after conferring with shore management, returned the launch alongside. The skipper reported that the deckhand and ship’s master were advised to stand by while they (the skipper) assessed the feasibility of continuing in the conditions. The skipper did not expect to embark anyone at that time and believed permission would be required before anyone attempted to use the ladder.

Meanwhile, the ship’s master was expecting the launch, having moved off, to wait until the ladder was in the ship’s lee again and the master had asked the launch to return. The master knew the starboard side was exposed to the weather and that making a lee using the ship’s engine would take some time and planned to resume the crew transfers once there was a good lee. However, the language difficulties experienced combined with other tasks in this busy period resulted in the master not being able to communicate this plan to the skipper of the returning launch.

At the same time, the departing chief mate (dCM) climbed onto and down the pilot ladder in anticipation of boarding the launch. Neither the ship’s master nor the launch skipper was aware that the dCM had descended the ladder. In addition, the deckhand did not know that the skipper only intended to assess conditions, not to embark anyone. The deckhand saw the dCM on the ladder, directed them clear of the approaching launch and anticipated signalling readiness when it was safe to board.

However, neither the skipper nor the deckhand nor the dCM recognised the significance of the rough seas and swell running along and against the ship’s starboard side resulting in larger and higher waves. Consequently, they did not expect the launch to be lifted higher than previously experienced while alongside. A large wave, however, lifted the launch high and it struck the dCM, who fell into the water. The dCM was recovered but had been fatally injured.

Weather and making a lee

While the weather conditions at the time with rough, choppy seas exceeded limits defined in both ship and launch procedures for operations in open waters, conditions in the lee of the large ship were sufficiently benign to allow safe personnel transfers. The ship’s main engine was also used to effect to make a lee on the starboard side as the anchored ship’s heading swung through its natural yaw. This provided 15-minute windows for safe transfers, and the accident occurred outside one of these windows while the master was attempting to make a lee.

Communications

Clear, unambiguous communications within and between workgroups are essential for the safe completion of any task. Specifically, this ensures that all participants:

• have a shared understanding of the task, of what is expected and what is about to happen
• know and understand their input into that task as well as their individual, and team, roles and responsibilities.

Formosabulk Clement’s personnel transfer required active and adaptable communications at all stages of the planning, co-ordination and control of the high-risk task. Difficulties maintaining optimal communications can arise with personnel who have not previously worked together, increased workgroup size, multiple workgroups of differing skill sets, multiple worksites, and different languages. These factors complicated the task on 9 August.

The personnel transfer operation comprised crew with roles and responsibilities in 4 closely linked, but distinct locations:

• ship’s bridge
• ship’s deck
• launch cabin
• launch deck

The crew in these locations were central to the co-ordination and safe management of the transfer, and safe completion of the task required clear communication between all of them. They were all aware of the overall objective and sequencing for the transfer but details of their individual and common plans for each step were not shared and agreed.

Multiple communication breakdowns occurred, resulting in the responsible crew at each location being unaware of the intentions or expectations of those at the other locations.

For example:

• On the ship’s bridge—the master intended to manoeuvre the ship to re-create the lee and did not expect the launch to return alongside and did not know its skipper’s intentions.
• On the ship’s deck—the dCM changed the departure plan, without informing others, including the master, and left the deck first. The intentions and expectations of the dCM were not known to anyone on the ship or the launch.
• In the launch cabin—the skipper planned to bring the launch alongside and assess conditions before allowing anyone to board. They did not expect anyone to be on the pilot ladder.
• On the deck of the launch—the deckhand was unaware of the skipper’s intention to only assess conditions. The deckhand expected boarding to take place once the launch was alongside and stable.

The opportunity to discuss the next phase of the operation while the launch was stood off from the ship was not taken. Had that time been utilised by the various participants to discuss and agree the next phase, the conditions would likely then have been favourable due to a lee and the dCM would have probably not been in the position they were at the time of the accident.

On this occasion, key personnel misinterpreted the actions and intentions of other parties. Both ship and launch crew assumed those on the other vessel understood what was happening and about to happen and were experienced in the transfer operation from each other’s perspective.

Pre-task information sharing and task understanding

The decision to conduct the personnel transfer was made several weeks before the event. There were explicit procedural requirements to seek and discuss operational requirements prior to the transfer, at least on the part of Formosabulk Clement. However, neither Formosa Plastics Marine Corporation (FPMC) nor Pacific Tug (PT) had specific guidance in place regarding the content of, or timing for, information to share with third parties regarding the specifics involved in the transfer of personnel to or from company vessels.

As a consequence, other than information related to the procedures and protocols required by shore agencies to prevent transmission of the COVID‑19 virus, only generic particulars such as time, location and boarding ladder positioning were exchanged in the lead up to the event. This in turn led to a situation where misunderstandings in, and differences between, plans could only be clarified while the transfer task was underway.

English is the international language for ship-to-shore communications[16] and masters and mates are required to demonstrate competence in its use.[17] Problems with language, translation and interpretation are known risks in international shipping that were not adequately mitigated on this occasion. Limitations of the ship’s officers’ English language ability were exposed when attempting to communicate complicated details of a dynamic situation, over radio, in a busy, noisy environment. In this case, the launch skipper resorted to enlisting the passengers to assist communications with the ship’s master in their native language. However, this removed the skipper from the communication loop and from developing a proper understanding about what the ship’s master and crew would do.

In general, standard procedures on both the ship and the launch emphasised the importance of establishing and maintaining effective communications, including communicating plans, requirements and expectations for the activity. However, the evidence indicates that there was no common or complete understanding of how the transfer would be conducted either within, or between, the crew of either vessel.

Formosabulk Clement

Multiple procedures on board Formosabulk Clement were in place for crew change operations. Crewmembers completed risk assessments and held meetings to share details, roles and responsibilities of the task. This provided the opportunity to reduce misunderstanding or miscommunication between those involved on board the ship.

Procedures also required a toolbox talk prior to the task and the crewmembers involved in the transfer recalled discussions prior to arrival and commencing the task. However, their different versions and details of those discussions did not indicate that the required distinct, formalised toolbox talk took place, potentially reducing its effectiveness.

Procedures also required that the launch’s risk assessment be obtained and discussed on board but this was not done. Consequently, the plans and discussions relating to the task were imprecise and did not include detail of how an individual was to safely transfer between the vessels and how this was to be controlled and co-ordinated. This contributed to the master losing understanding of what was happening on the ship’s deck and what the launch was doing and why. Most significantly, the dCM left the deck first and climbed down the pilot ladder without the master’s knowledge or approval.

Pacific Tug

Pacific Tug (PT) had considerable experience in conducting personnel transfers with procedures to support these operations. However, PT had not ensured that all involved personnel, within the company and external parties (such as ship masters), had an agreed, clear and complete knowledge and understanding of how this process would occur and progress.

As a consequence, when the launch skipper returned alongside the ship with the intent to assess conditions, their actions and plans were not clearly understood by all. Therefore, their intentions were not understood on board the ship (bridge or deck) or by the launch deckhand. Specifically:

• the approaching launch was probably mistaken by the dCM as a signal to climb down the ladder in preparation to disembark
• the master did not understand why the launch was returning before a lee had been provided and before being requested to return
• the deckhand took it as confirmation that crew transfers were resuming.

Findings

The following findings are made with respect to a crewmember’s fatal fall from Formosabulk Clement’s pilot ladder during crew transfer operations with the launch PT Transporter on 9 August 2021.

Contributing factors

• Formosabulk Clement’s departing chief mate climbed down the vertical pilot ladder intending to board the crew transfer launch. They did this without the knowledge of the ship’s master or the skipper of the launch.
• After the ship's master had manoeuvred the ship to create a lee and embark the joining crew from the launch, the ship yawed about the anchor, which exposed the transfer area to the prevailing weather.
• Immediately before the accident, there were difficulties in communicating between the ship’s bridge and the launch. As a consequence, the main engine was not used to re-create a lee before the launch came back alongside the ship, and the transfer location was exposed to the weather.
• A wave, larger than previously encountered, lifted the transfer launch higher than expected and sufficient to make contact with the departing chief mate, knocking them into the water.
• On board Formosabulk Clement, communications plans and protocols for the crew transfer operation had not been sufficiently well implemented and agreed to ensure that deck and bridge personnel maintained awareness and understanding of what was occurring and what was about to occur.
• On board PT Transporter, communications arrangements had not been sufficiently well implemented to ensure that events occurring as the crew transfer process progressed were clearly understood and agreed between its skipper and deckhand.
• Neither vessel’s managers had ensured that the involved personnel had a common and complete understanding of how the personnel transfer would be conducted with respect to aspects including:
  • outlining task steps, limits and triggers such as agreed permissions for vessel movements
  • setting operational limits
  • defining terminology for key transfer-related communications.

The opportunity was not taken to share information well beforehand, which would have allowed differences to have been addressed and a single plan developed and agreed before the task was underway.

Safety action

Pacific Tug

Pacific Tug (PT) informed the ATSB that the company immediately ceased personnel transfer operations after this accident until a comprehensive review of such operations was conducted. PT sought a system of communication less constrained by language and amenable to being shared beforehand to assist in achieving the shared mental model of the task among all participants.

To that end, PT designed and implemented a traffic light system (red, amber and green states) for control of 2 separate aspects of the transfer process:

• operational status—through the updated ‘operational risk assessment for personnel transfers to/from a PT vessel’ a range of preconditions were set which need to be met before the planned operation could commence. These preconditions included the setting of weather limits for each state as well as requirements on all involved vessels and crewmembers. The weather parameters are set prior to the launch leaving the wharf and then assessed against at the transfer location.
• transfer status—indication of the status of the transfer as it is occurring and controlled by the crew of the transfer launch. 
  • Green—transfers can occur as guided by launch crew
  • Amber—transfers to stop, personnel make safe, reassess conditions to determine next steps
  • Red—the transfer operation is stopped and not to recommence until the operational status preconditions are reassessed and met.

Transfer of personnel can only commence when both statuses are green, and masters of both vessels have agreed on the process and given permission.

PT procedures required that the operational risk assessment and a ‘personnel transfer procedure poster’ were shared with the vessel requiring the transfer of personnel. The poster outlined the traffic light system and actions to take in each situation. The masters of involved vessels, all personnel transferring and anyone assisting in the transfer were to provide confirmation that they understand and agree to the transfer protocols.

It was also emphasised that the master of the crew change vessel was to be in ultimate control for all transfers.

These procedural changes were supported by modifications to the launch (PT Transporter) through the fitting of red, amber and green indicator light sets: one in the passenger area and a second on the main mast. These light sets were designed to show the status of the transfer to persons on board both vessels. The launch crew (master and deckhand) each have remote controls for the lights.

During an extended transfer (many persons), the status was able to alternate between amber and green only. Red required a complete termination of the operation.

This personnel transfer process was further supported by updating of procedures including those for:

• Risk assessment
• Operational risk assessment for personnel transfers to/from a PT vessel
• Personnel transfers

Formosa Plastics Marine Corporation

Following this accident, Formosa Plastics Marine Corporation (FPMC) notified the ATSB that the company had:

• completed on board training and reassessment of risks relating to transfer of personnel to and from small vessels, including pilots
• conducted investigations via a third party and internally
• shared lessons learned throughout the FPMC fleet including a fleet circular outlining precautions for crew change at anchorage
• amended the ‘personnel transfers procedure at sea’
• held company safety meetings, including to senior shore management, to discuss the accident and outcomes.

Maritime Safety Queensland

Maritime Safety Queensland (MSQ) amended the COVID crew change procedure for vessels at anchor following the accident. The changes included that:

• all crew transfers were to be conducted during daylight hours only, commencing after sunrise, completing before sunset
• weather conditions were to be appropriately risk assessed by those involved prior to conducting the transfer 
• both the ship and the transfer vessel were to be provided a copy of the Maritime Safety Awareness Bulletin September 2019 – Safe Access to Vessels.  
• the transfer operations were to be appropriately risk assessed considering the competency and ability of all involved personnel (ship and launch) to complete transfers between vessels.

These conditions remained in place until COVID restrictions were eased in 2022.

Glossary

Sources and submissions

Sources of information

The sources of information during the investigation included:

• the master and crewmembers of Formosabulk Clement
• the skipper and deckhand of PT Transporter
• Formosa Plastics Marine Corporation
• Pacific Tug
• Wave Shipping
• LBH Australia
• Maritime Safety Queensland
• Australian Bureau of Meteorology
• Queensland Police Service
• Australian Maritime Safety Authority
• Marine Investigation Department of the Liberian Registry
• Queensland Department of Environment and Science
• Aus Ship P&I
• Wilhelmsen Ships Service
• Bhagwan Marine
• Citizens Radio Emergency Service Teams

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:

• the skipper and deckhand of PT Transporter
• Pacific Tug
• Formosa Plastics Marine Corporation
• the master of Formosabulk Clement
• Maritime Safety Queensland
• Australian Maritime Safety Authority
• Marine Investigation Department of the Liberian Registry

Submissions were received from:

• the skipper of PT Transporter
• Pacific Tug
• Formosa Plastics Marine Corporation
• Maritime Safety Queensland
• Australian Maritime Safety Authority
• Marine Investigation Department of the Liberian Registry

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

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 2023 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 Coat of Arms, ATSB logo, and photos and graphics in which a third party holds copyright, this publication is licensed under a Creative Commons Attribution 3.0 Australia licence. Creative Commons Attribution 3.0 Australia Licence is a standard form licence agreement that allows you to copy, distribute, transmit and adapt this publication provided that you attribute the work. The ATSB’s preference is that you attribute this publication (and any material sourced from it) using the following wording: Source: 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.

Executive summary

What happened

On 18 August 2021, an amateur-built Stolp Acroduster II SA-750, registered VH-YEL, departed Caboolture Airfield, Queensland for an aerobatic flight, with the pilot being the sole occupant. A short time later, the aircraft sustained an in-flight break-up. The pilot was fatally injured and the aircraft was destroyed.

What the ATSB found

The ATSB found that the centre section of the upper wing was located away from the main aircraft wreckage. Technical examination of the attachment points identified fatigue cracking on the eye bolts located in the upper wing forward position on the left and right cabane struts. The fatigue cracking had initiated in the thread root of each eye bolt at its termination into the cabane strut. The left eye bolt fracture surfaces were corroded, that indicated the left eye bolt had failed some time before the accident flight. Fatigue cracking was also identified in the left roll brace where it secured to the same attachment points. The weakened eye bolts and roll brace subsequently fractured, which led to structural instability of the centre wing section, and an in-flight break-up of the aircraft wing structure.

The fatigue cracks were located in areas that would not be easily identifiable during standard maintenance inspections without disassembly of the relevant wing attachment points. Further, fatigue cracks were identified on the same aircraft type, in the same eye bolt and roll brace positions 2 decades before this accident. Despite this, there was no provision for product improvements to identify emerging design issues and to implement design change or additional maintenance inspections, nor was this required for aircraft in the experimental category. 

What has been done as a result

In August 2021, the ATSB notified the Civil Aviation Safety Authority, the United States (US) National Transportation Safety Board, the US Federal Aviation Administration, the US Experimental Aircraft Association (EAA), and the aircraft design owner of the initial finding that fatigue cracking had been identified within the upper wing centre section attachment eye bolts. The ATSB also contacted the owner of the only other Stolp Acroduster II SA-750 on the Australian civil aircraft register and informed them of the issue.

In November 2021, the ATSB issued a preliminary report and safety advisory notice (SAN) AO‑2021-032-001 to inform aircraft type owners of the circumstances of the accident and the fatigue crack issue. The FAA, EAA and the design owner also provided information to aircraft owners with a link to the ATSB’s preliminary report and SAN.

Safety message

Aircraft owners and maintainers of experimental amateur-built aircraft should consider conducting additional detailed inspections that exceed the minimum standards in areas of the aircraft that are critical to the safety of flight. Further, Acroduster service history has shown that elements of the upper centre wing attachment joint have a predisposition to cracking. Care should be taken during pre-flight and scheduled inspections, being particularly attentive for the identification of cracks/corrosion/fretting.

The investigation

The occurrence

On 18 August 2021, at 0852 local time, an experimental amateur-built Stolp Acroduster II SA-750 aircraft, registered VH-YEL, departed the Caboolture airfield, Queensland, for a local aerobatic flight. The pilot was the sole occupant.

Shortly after, the crew of a helicopter operating in the vicinity, witnessed red and white debris falling from the sky. After searching the immediate area, the helicopter crew identified the main wreckage of VH-YEL.

The majority of the wreckage came to rest inverted in tidal wetlands, close to the mainland shoreline of Pumicestone Passage, adjacent to Bribie Island (Figure 1). The aircraft was destroyed, and the pilot was fatally injured.

Initial assessment of the aircraft wreckage distribution by the ATSB at the accident site, indicated that the aircraft had sustained an in-flight break-up. The main aircraft wreckage and other components were recovered from the accident site and taken to a secure facility for detailed examination by the ATSB.

Figure 1: Aircraft departure point and accident location

Source: Google Earth, annotated by the ATSB

Context

Pilot information

The pilot held a valid Private Pilot Licence (Aeroplane) with 674.2 hours total flight time, including 108 flight hours conducted in VH-YEL. They also held a Private Pilot Licence (Helicopters) with an additional 102.9 hours flight experience. The pilot held the following ratings and endorsements:

• single-engine aeroplane class rating
• aerobatics and spinning flight activity endorsement
• manual pitch propeller control and tail wheel undercarriage design feature endorsements.

A review of the pilot’s logbook indicated that they had predominately focussed nearly all their flying since June 2016 on the conduct of aerobatics in a variety of different higher-performing single‑engine aircraft, including the accident aircraft.

The pilot held a valid Class 2 Aviation Medical Certificate with no restrictions.

Aircraft information

Experimental amateur-built aircraft

An amateur-built aircraft is an aircraft, the major portion (more than 50%) of which has been fabricated and assembled by a person who undertook the construction project solely for their own education or recreation (Advisory Circular AC-21.4(2) Amateur-built Experimental Aircraft – Certification). An amateur-built aircraft can be built from scratch based on an original design or established plans, or from a kit. Pilots and passengers of these aircraft operate under the premise of informed participation and accept the risk that these aircraft may not meet the same airworthiness safety standards as certified aircraft.

General information

The Stolp Acroduster II SA-750 is a 2-seat amateur-built aerobatic bi-plane rated to +/- 9 g.[1] The aircraft was first introduced in 1971 as a plans-built aircraft, available from the Starduster Corporation in the United States. The current design owner (Aircraft Spruce) advised the ATSB that the plans for all Starduster Corporation aircraft had not been modified since first produced. The aircraft fuselage is constructed of welded steel tube and the wings incorporate wooden structure with steel tube internal supports. The majority of the aircraft is covered in fabric. It is powered by a 4-cylinder piston engine driving a 2-blade fixed pitch propeller (Figure 2).

Figure 2: VH-YEL Stolp Acroduster II SA-750  

Source: Supplied

There are several other plans-built Starduster Corporation aircraft variants developed by the same designer that share similar design features, such as the Starduster and single seat Acroduster (SA-700). However, only the Acroduster series aircraft share the same upper wing attachment points. There were about 130 Acroduster SA-700/750 aircraft that were completed with 2 of them being on the Australian register. Some components for the aircraft were able to be purchased as kits or raw materials.

VH-YEL details

The aircraft’s plans were supplied from the Stolp Starduster Corporation in 1976 and the aircraft was constructed in the United States with a serial number of T-02. The aircraft was first registered as N97177 and first flew in 1981. It was disassembled and shipped to Australia in 2003. At that time, the aircraft had accumulated about 430 flight hours.

The aircraft was re-registered in Australia in 2007 as VH-YEL and was operated under the experimental category.[2] The aircraft was purchased by the accident pilot on 9 July 2019. It was maintained in accordance with the Civil Aviation Safety Authority maintenance schedule 5[3] and, at the time of the accident, had accumulated about 717 flight hours. The last periodic maintenance inspection was conducted on 27 August 2020 at 687.3 hours total time‑in‑service, with the next inspection due 9 days after the accident.

Wing attachment and bracing description

The lower wings of the Acroduster SA-700/750 were directly attached to the lower fuselage at 2 wing root attachment points on each side. The upper wings incorporated the left, right and centre sections (Figure 3). The centre section was attached to the fuselage by eye bolts (yellow) that then connected to the forward cabane struts (green) and roll braces (light orange). The wings were externally braced against each other using flying wires (red), landing wires (dark orange), and interplane struts (blue). In-flight, the lower wing outboard sections were supported by the upper wing outboard sections through the interplane struts and flying wires.

Figure 3: Wing bracing and attachment

Source: Starduster, annotated by the ATSB

The cabane struts and roll braces were provided to support and attach the centre section of the upper wing to the fuselage. Eye bolts had been manufactured to thread into the upper portion of each cabane strut and completed the attachment between the upper wing centre section and the struts (Figure 4). The eye bolts were adjustable in length through their threaded section.

Figure 4: Upper wing centre section attachment

Source: Supplied, annotated by the ATSB

Technical support and eye bolt inspection requirements

There was no in-service technical support for the aircraft type provided by the design owner (Aircraft Spruce), nor was there required to be by regulations for aircraft in the experimental category. Such support would typically include:

• the communication and analysis of service difficulties, defects, design issues and or technical failures of components
• detailing specific type maintenance inspection requirements based on known design issues
• providing technical assistance for product improvements to identify emerging design issues and to implement design change.

The only method to identify and mitigate design issues with the aircraft was via word of mouth, magazine articles, amateur builder associations, or on-line aircraft forums where owners could discuss safety issues and design problems among themselves.

Apart from the requirement to complete a general inspection of the wing attachment points and wing structure periodically for condition and security, the aircraft type did not have a specific detailed inspection schedule for the eye bolts or roll braces to ensure their ongoing airworthiness.

In addition, the aircraft type did not have a specified life‑limit to replace the eye bolts at set flight hours and/or cycles, rather, they were an on‑condition replacement.

Meteorological information

Closed circuit television footage at Caboolture airfield showed there was little to no wind, good visibility, and broken[4] cloud with no rain at the time the aircraft departed. This was consistent with the Bureau of Meteorology graphical area forecast, which further indicated wind speeds of 17‑20 kt from the south, south‑south-east between 1,000-5,000 ft. Therefore, weather was not considered to be a contributing factor to the accident.

Aircraft radar surveillance information

Recorded surveillance data was obtained from Airservices Australia. As the aircraft was not fitted with a transponder, only primary radar returns[5] were available. The coverage will detect aircraft at, or above heights depending on the aircraft’s proximity to radar coverage and terrain shielding. Due to the limitations of primary radar and terrain shielding, there were gaps in the radar recording and the aircraft’s altitude, speed and timing data was not available.

However, there was limited radar data in respect to an aircraft departing Caboolture and a flight path consistent with VH-YEL. This included termination of the flight in the immediate vicinity of the accident site. The ATSB’s analysis of the recorded flight path data showed that the aircraft was conducting several minutes of aerobatic manoeuvring leading up to and likely during the in-flight break‑up sequence.

Wreckage and impact information

The aircraft wreckage was located on tidal mud flats and farmland, it was spread over a distance of about 2.4 km. The trail was oriented in a south-east to north-west direction consistent with wind direction on the day.

The main wreckage consisted of the fuselage, tail, and outer-wing sections still attached to the fuselage by the flying and landing wires. The left lower inboard wing section had separated from the aircraft but was located adjacent to the main wreckage. The upper wing centre section was located about 500 m away from the main wreckage. The majority of the remaining scattered debris consisted of light wooden sections of wing structure, wing skin fabric, panels and windshield fragments (Figure 5).

Figure 5: Wreckage trail distribution over mud flats and farmland

Source: Google Earth, annotated by the ATSB

The main wreckage was recovered by barge at the changing of the tide and was transported to a secure storage facility for examination by the ATSB (Figure 6). Reconstruction of the sections of recovered aircraft identified that:

• the majority of the parts were accounted for with the exception of the right interplane strut
• the aircraft flight controls had no pre-impact defects
• the fuselage, wings, flying and landing wires had no pre-impact defects
• the forward cabane strut upper wing attachment point eye bolts to the centre-wing structure had fractured and displayed evidence of pre-existing fatigue cracking
• there was no evidence of propeller rotation at impact, however, the engine most likely stopped from fuel starvation during the in-flight break-up sequence.

Figure 6: VH-YEL as recovered and partially reassembled

Source: ATSB

Upper centre wing attachment points

Technical examination of the upper centre wing attachment – summary

Both forward cabane struts and roll braces, and the attaching hardware were cleaned and treated against corrosion during the wreckage recovery process. They were removed from the aircraft and sent to the ATSB’s technical facilities in Canberra for detailed examination. For further detail refer to the materials analysis report at Appendix A. The examination identified:

• The left and right eye bolts were cadmium plated, had rolled threads, and were likely made of 4140 alloy steel.
• The eye bolts had no observable manufacturing defects.
• There were fatigue cracks in a fore/aft direction through 35% of the left eye bolt threaded cross section with the remaining material fractured in overload.
• There were uniaxial fatigue cracks in a fore/aft direction through 85% of the right eye bolt threaded cross section with the remaining material fractured in overload (Figure 7).
• There were several areas of additional micro-cracking present in the threaded portion of the forward eye bolts.
• The left eye bolt fracture surfaces had significant corrosion when compared to the right eye bolt fracture surfaces. This indicated that the left eye bolt had completely fractured some time before the accident flight (Figure 8).
• Visual examination of both roll braces indicated that they had fractured at their upper connection.
• The left roll brace fracture surface that was coincident with the position of the upper cabane strut eye bolts had fatigue cracks through 60% of its cross-section.
• Due to the development of corrosion damage to the fracture surfaces, it could not be ascertained if the right roll brace had also developed fatigue cracking.

Figure 7: Right eye bolt, cabane strut and fracture surface

Source: ATSB

Figure 8: Corrosion level differences between left and right eye bolt fracture surfaces

Source: ATSB

The location of the crack on the right forward cabane strut eye bolt was directly under the securing nut, making it imposssible to identify the fatigue crack without dissassembling the attachment point and removal of the eye bolt. The location of the crack on the left forward cabane strut was directly above the securing nut and would have been difficult to identify while it was forming in the threaded section without dissassembly and removal of the eye bolt from the cabane strut.

Eye bolt history

A subject matter expert familiar with the early design and development of the aircraft type indicated that there has been a history of cabane strut eye bolt cracking in the same area of the threaded section as occurred with VH-YEL. They provided the ATSB with an image taken 2 decades before this incident of an Acroduster SA-750 where a left forward upper wing cabane strut eye bolt fractured completely through its threaded section. That fracture was in the same location as VH-YEL. The left roll brace was the only remaining structure providing support to that section of the wing (Figure 9). It could not be determined how long prior, or for how many hours the aircraft had flown with the failed eye bolt.

Figure 9: Previous Acroduster SA-750 with fractured wing attachment eye bolt

Source: Supplied, annotated by the ATSB

The ATSB was further advised that some owners of Acroduster aircraft made modifications to the upper wing attachment points. The modification removed the eye bolts and replaced them with a welded structure, which attached directly to the upper wing attachment point, effectively strengthening the attachment points (Figure 10). The ATSB notes that the modification shown was not approved by the aircraft design owner and had been added by experimental amateur-built owners. Advice on such modifications should be sought from a suitably qualified aeronautical engineer on any design changes undertaken.

Figure 10: Right-side upper wing centre section attachment point with a standard eye bolt attachment (left) and a modified attachment (right)

Source: Supplied, annotated by the ATSB

Roll brace history

The Starduster Magazine had been produced as a technical document to assist amateur builders of Starduster company aircraft including the Acroduster aircraft. In the April 1980 edition, the author reported that while completing an aircraft inspection, cracks were identified in the roll brace and the cabane strut. Notably:

The most significant repair was on the Cabane struts, sheet 5, items 4 and 5. One of them was cracked, and the brace strut [roll brace] on that side was carrying no load.

As such, the author provided a cautionary note:

If you are already flying your machine, we recommend that you keep a careful eye on these sway [roll] braces, and carefully inspect them both visually and by shaking, before your first flight of the day. You should change them at the next annual inspection.

The article stated that a modification had been made to replace the steel tube roll braces with stainless steel material that was the same type as the flying wires. The last edition of Starduster Magazine was produced in October 1985.

Safety analysis

Introduction

While conducting an aerobatic flight, the aircraft sustained an in-flight break-up and a subsequent collision with the terrain. The pilot was fatally injured and the aircraft was destroyed. Adverse weather was not identified as a factor in the accident flight.

The following analysis examines what contributed to the in-flight break-up sequence and technical support, design, and maintenance inspection considerations.

In-flight break-up sequence

The aircraft wreckage was distributed over 2.4 km, with the centre wing section found about 500 m away from the main wreckage. The ATSB’s metallurgical examination of the forward left and right upper wing centre section attachment eye bolts (eye bolts) from VH-YEL did not identify any manufacturing defects or anomalies outside of the limited design specifications provided. However, the ATSB’s inspection found areas of significant pre-existing fatigue cracking in the threaded portions of the eye bolts. There was also significant corrosion on the fracture surfaces of the left eye bolt, when compared to the right eye bolt. This indicated the left eye bolt had failed some time before the accident flight.

Further, inspection of the left roll brace also identified areas of fatigue cracking. It was likely that, when the left eye bolt failed, the flight loads transferred to the adjacent structure, which was the left roll brace. This increased the fatigue inducing flight loads acting on the roll brace. A previous failure of an eye bolt that occurred on another Acroduster aircraft showed that aircraft flight loads could be supported by the respective roll brace.

Subsequently, it was likely that the right eye bolt and left roll brace fatigue cracks propagated to a critical size and that the remaining material was no longer able to sustain normal flight loads and consequently failed in overstress. They, along with the previously failed left eye bolt initiated the in‑flight break-up sequence.

Detection of the cracks

In this accident, the fatigue cracks on the cabane struts formed in an area of the eye bolts that were obscured by the securing nuts and threads of the eye bolts. Therefore, the cracks would not have likely been readily identifiable during standard maintenance inspections without disassembling the attachment points. However, as the last periodic inspection occurred about 1 year prior to the accident, it could not be determined if the cracks would have been present at that time.

Aircraft design and maintenance

As discussed above, the ATSB’s examination found fatigue cracking in both eye bolts and the left roll brace. An expert familiar with the Acroduster design had stated that there was a history of eye bolt cracking as previously evidenced on another aircraft. Similarly, the Starduster Magazine, which had ceased publishing in 1985, had mentioned incidents of roll brace cracking.

Consequently, some Acroduster owners had modified the attachment points to remove the eye bolts and replaced the roll braces with bracing wires. Those design improvements were made at a local level by individual aircraft owners, rather than through direct and broad reaching technical support provided by the design owner. Without design owner tracking and analysis of historic in‑service component and structural design issues:

• broad reaching structural design improvements will not be made
• additional maintenance inspections to identify defects in areas with known issues will not be implemented
• time/cycle component retirement life assessments will not be conducted.

In this case, the aircraft was provided to the builder in a basic kit form, with a set of design plans that had not changed since they were originally produced about 50 years ago. Further, there was no explicit requirement to inspect the eye bolts or roll braces in detail, despite being a known area for fatigue cracking and on-condition component with no specified replacement interval.

Therefore, without a mechanism for identifying product improvements, Acroduster owners may not be aware of possible design issues and enhancements made by others.

Findings

From the evidence available, the following findings are made with respect to the in-flight break-up accident involving a Stolp Acroduster II SA750, registered VH-YEL, on 18 August 2021.

Contributing factors

• The left eye bolt that supported the upper wing centre section failed prior to the accident flight due to fatigue cracking. This, in combination with pre-existing fatigue cracking and subsequent overstress of the right eye bolt and left roll brace during the flight, led to the in‑flight break-up.

Other factors that increased risk

• The eye bolt fatigue cracks formed in areas that would not be easily identifiable during the periodic maintenance inspections without the requirement for removal of the eye bolts from the cabane struts. This reduced the opportunity to identify fatigue cracks before they propagated to a critical size.
• There was no provision for product improvements to identify emerging design issues and to implement design change or additional maintenance inspections, nor was this required for aircraft in the experimental category.  

Safety actions

Safety action by the ATSB

Shortly after the accident, the ATSB notified the Civil Aviation Safety Authority, the United States National Transportation Safety Board and Federal Aviation Administration, the kit plane design/material provider and the Experimental Aircraft Association that fatigue cracking had been identified within the upper wing centre section attachment eye bolts. The ATSB also contacted the owner of the only other Stolp Acroduster II SA-750 on the Australian civil aircraft register and informed them of the fatigue cracking.

Further, the ATSB released a safety advisory notice (AO-2021-032-SAN-001) with the preliminary report to inform aircraft type owners of the circumstances of the accident and the fatigue crack issue. The notice stated that:

The Australian Transport Safety Bureau advises all owners, operators and maintainers of Stolp Acroduster SA‑700/750 aircraft to consider the safety implications of the initial findings of this investigation regarding the fatigue cracking on forward cabane strut upper-wing centre-section attachment eye bolts, and take action where considered appropriate to ensure that their aircraft remain airworthy.

Safety action by the Experimental Aircraft Association

In November 2021, the United States Experimental Aircraft Association released a notice on its website informing its members of the Acroduster accident and linking the notice to the ATSB preliminary report and safety advisory notice.

Safety action by design owner (Aircraft Spruce)

In November 2021, the design owner issued a safety advisory notice to every purchaser of Starduster and Acroduster design plans since 2003. The notice informed customers about the fatigue cracking and provided a link to the ATSB preliminary report.

Safety action by the United States Federal Aviation Administration

In November 2021, the United States Federal Aviation Administration issued a notice to more than 280,000 web subscribers informing them of the wing attachment point fatigue cracks and providing a link to the ATSB preliminary report and safety advisory notice.

Sources and submissions

Sources of information

The sources of information during the investigation included the:

• aircraft design owner (Aircraft Spruce)
• aircraft design subject matter expert
• Experimental Aircraft Association
• United States National Transportation Safety Board
• United States Federal Aviation Administration
• various experimental aircraft forums
• Airservices Australia
• Civil Aviation Safety Authority.

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:

• aircraft design owner (Aircraft Spruce)
• aircraft design subject matter expert
• Experimental Aircraft Association
• United States National Transportation Safety Board
• United States Federal Aviation Administration
• Airservices Australia
• Civil Aviation Safety Authority.