Marcus' Model Railway Journey

Summary

At around 18:43 hrs on 31 October 2021, train reporting number 1L53, the 17:20 hrs South Western Railway passenger service from London Waterloo to Honiton, passed a red signal and collided with the side of train 1F30, the 17:08 hrs Great Western Railway passenger service from Portsmouth Harbour to Bristol Temple Meads.

At the point of collision, train 1L53 was travelling at approximately 52 mph (84 km/h) and train 1F30 at 20 mph (32 km/h).

The collision took place at Salisbury Tunnel Junction, which is on the immediate approach to Fisherton Tunnel, near Salisbury in Wiltshire.

The impact of the collision caused the front two carriages of train 1L53 and the rear two carriages of train 1F30 to derail.

Both trains continued some distance into Fisherton Tunnel before they came to a stop.

Thirteen passengers and one member of railway staff required treatment in hospital as a result of the accident, which also caused significant damage to the trains and railway infrastructure involved.

A potentially far more serious collision between train 1L53 and an earlier train travelling in the opposite direction was avoided by less than a minute.

The causes of the accident were that wheel/rail adhesion was very low in the area where the driver of train 1L53 applied the train’s brakes, that the driver did not apply the train’s brakes sufficiently early on approach to the signal protecting the junction to avoid running on to it, given the prevailing low level of adhesion, and that the braking systems of train 1L53 were unable to mitigate this very low adhesion.

The level of wheel/rail adhesion was very low due to leaf contamination on the railhead, and had been made worse by a band of drizzle that occurred immediately before the passage of train 1L53.

This leaf contamination resulted from the weather conditions on the day of the accident, coupled with an increased density of vegetation in the area which had not been effectively managed by Network Rail’s Wessex route.

Network Rail’s Wessex route had also not effectively managed the contamination on the railhead with either proactive or reactive measures.

RAIB’s investigation found that a probable underlying factor was that Network Rail’s Wessex route did not effectively manage the risks of low adhesion associated with the leaf fall season.

RAIB also found that South Western Railway not effectively preparing its drivers for assessing and reporting low adhesion conditions was a possible underlying factor.

RAIB has also made two safety observations.

These relate to the application of revised design criteria for the Train Protection and Warning System and the assessment of signal overrun risk and how this accounts for high risk of low adhesion sites.

Two issues were found relating to the severity of the consequences.

These were a loss of survival space in the driver’s cab of train 1L53, and the jamming of internal sliding doors, which obstructed passenger evacuation routes.

Since the accident, Network Rail has reviewed its training and competence framework for off track staff at network level, and is also reviewing its adhesion management standards.

Network Rail’s Wessex route is reviewing its arrangements for proactively responding to reports of low adhesion, including how it undertakes railhead treatment.

South Western Railway has made changes relating to training and briefing of its drivers to ensure information on autumn arrangements has been effectively briefed and understood.

Network Rail and South Western Railway have also jointly updated their annual autumn leaf fall working arrangements to ensure that sites at high risk of low adhesion are identified, reassessed, managed and monitored.

The Rail Safety and Standards Board has revised the rail industry standard that provides guidance for the rail industry regarding the management of low adhesion.

Cross-industry working groups have also issued revised guidance regarding low adhesion.

In December 2021, the safety authority for the mainline railway in Great Britain, the Office of Rail and Road, issued an improvement notice to Network Rail’s Wessex route requiring it to improve its vegetation management and its assessment and control of low adhesion risks.

As a result of the investigation, and accounting for the work done by the industry since the accident, RAIB has made ten recommendations.

Seven of these recommendations are made to Network Rail.

These relate to: a review of the processes, standards and guidance documents relating to the management of leaf fall low adhesion risk; the training and competence of staff dealing with vegetation management and seasonal delivery; responses to emerging and potential railhead low adhesion conditions; management of railhead treatment regimes; assessment of the risk of overrun at signals which have a site at high risk of low adhesion on their approach; and a review of the retrospective application of design criteria for the Train Protection and Warning System.

One recommendation is made to South Western Railway to review and improve its arrangements for training and briefing drivers to ensure that they are able to effectively identify areas of low adhesion and report them as appropriate.

One recommendation is made to the Rail Delivery Group in consultation with train operators and the Rail Safety and Standards Board regarding the review of technologies other than sanding systems and wheel slide protection to improve
braking in low adhesion conditions.

One recommendation is made to Porterbrook, Eversholt and Angel Trains regarding the design of the internal sliding doors on class 158 and 159 carriages.

The accident
Summary of the accident

At around 18:43 hrs on Sunday 31 October 2021, train reporting number 1L53, the 17:20 hrs South Western Railway passenger service from London Waterloo to Honiton, collided with the side of train 1F30, the 17:08 hrs Great Western Railway passenger service from Portsmouth Harbour to Bristol Temple Meads.

Train 1L53was travelling at approximately 52 mph (84 km/h)2 and train 1F30 at 20 mph(32 km/h).

The collision occurred at Salisbury Tunnel Junction, which is on the approach to Fisherton Tunnel, near Salisbury in Wiltshire.

Extract from Ordnance Survey map showing the location of the accident at Salisbury TunnelJunction

The movement of train 1F30 across the junction was being protected from trains approaching on the Down Main line by signal SY31, which was displaying a red (danger) aspect.

The signal had previously been protecting another train, train1F27, which passed over the junction in the opposite direction less than a minute before the accident.

Train 1L53 passed this signal by around 200 metres before colliding with train 1F30.

The impact of the collision caused the front two carriages of train 1L53 and the rear two carriages of train 1F30 to derail.

Both trains continued some distance into Fisherton Tunnel following the collision before they came to a stop.

Overview of the location of the accident and geographical relationship of the main features

Diagram of the location showing the direction of train 1F27 that had just passed over SalisburyTunnel Junction and the approach of trains 1L53 and 1F30 to Fisherton Tunnel

The aftermath of the accident

The aftermath of the accident


The accident caused substantial damage to the track, the tunnel and the surrounding railway infrastructure.

The leading carriage of train 1L53 and the rear two carriages of train 1F30 were damaged beyond economic repair.

The rear two carriages of train 1L53 were less seriously damaged and the leading two carriages of train 1F30 were undamaged.

The aftermath of the accident



The driver of train 1L53 was seriously injured in the accident and spent three weeks in hospital.

Thirteen passengers were also taken to hospital, with one suffering serious injuries.

A further ten passengers were treated at the scene for cuts and bruises.

In total, there were 197 passengers and five members of staff on the two trains.

Each train had a driver and a guard on board.

There was also an on-duty driver travelling as a passenger on train 1F30 when the accident occurred.

Context

Location

At Salisbury Tunnel Junction, the Up and Down Main lines, which run to and from London Waterloo via Basingstoke, meet the Up and Down Dean lines, which run to and from Southampton via Romsey.

Beyond the junction is the 405 metres (443 yards) long Fisherton Tunnel and, just over 1 mile (1.6 km) further west, Salisbury station.

The collision occurred around 200 metres after train 1L53 had passed signal SY31, at 82 miles 35 chains from London Waterloo.

The maximum permitted speed for trains approaching the junction on the Down Main line is 90 mph (145 km/h), reducing to 50 mph (80 km/h) approximately700 metres before the junction.

Drivers are informed of the 50 mph (80 km/h) speed restriction by a warning board about 1,400 metres on approach to its start, and the start of the speed restriction itself is indicated by a commencement board.

Due to a sharp curve, the maximum permitted speed on the Down Dean line approaching the junction is 20 mph (32 km/h).

Organisations involved

The railway infrastructure at Salisbury is owned, operated and maintained by Network Rail.

Salisbury Tunnel Junction forms part of Network Rail’s Wessex route which, along with the Kent and Sussex routes, forms part of the Southern region.

Each Network Rail region acts as a devolved management organisation within the company.

Due to the size of the Wessex route, maintenance and inspection of the infrastructure is split between outer and inner maintenance delivery units (MDUs).

The Salisbury line falls under the responsibility of the Wessex outer MDU.

Network Rail employs the signallers at Salisbury signal box, autumn control room staff at the Wessex integrated control centre (WICC) in Basingstoke, the seasons delivery specialist (SDS) and the lineside and drainage senior asset engineer (SAE).

It also employs the MDU engineering and inspection off track staff, who manage the control and inspection of vegetation.

Network Rail Supply Chain Operations, part of Network Rail Route Services, organises the contractual arrangements and route schedules for the Wessex route railhead treatment multi-purpose vehicle (MPV) fleet.

Although Network Rail's Wessex route is the fleet’s owner and entity in charge of maintenance, operation of the MPV fleet and the actual undertaking of the maintenance is contracted out.

Train 1L53 was operated by First MTR South Western Trains Ltd, trading as South Western Railway (SWR).

SWR employed the train driver and the guard of this train.

Train 1F30 was operated by First Greater Western Ltd, trading as Great Western Railway (GWR).

GWR employed the train driver and guard of this train.

Porterbrook Leasing owned the carriages which formed trains 1L53 and 1F30 and leased them to the train operators concerned.

Network Rail and SWR together undertook an internal railway industry investigation into the accident (‘the joint industry investigation’).

All parties freely co-operated with RAIB’s investigation.

Staff involved

The driver of train 1L53 started his railway career in 1962 with British Railways, working as a fireman on steam locomotives and later as a train driver’s assistant.

He started driving trains for British Rail in 1982 and, following privatisation of the industry, was employed by South West Trains and latterly SWR as a driver and driver instructor.

The driver had been based at Salisbury depot for his entire career and had moved to a part-time contract in 2019.

The driver had been deemed competent to drive trains by SWR and all his competence assessments were in date in accordance with the company’s train driver competency management process.

These assessments included defensive driving techniques relating to low adhesion, against which the driver demonstrated his competence between August 2018 and July 2021, as part of a three-year competence assessment cycle, and was due to be reassessed in 2024.

At the time of the accident, the driver was working two days a week, driving class 158 and 159 trains between London Waterloo and Exeter, Southampton and Basingstoke, as well as on the Salisbury to Bristol route.

For the week starting 25 October 2021, the driver worked on Tuesday 26 and Saturday 30 October.

He was then scheduled to work nine-hour shifts on Sunday 31 October, the evening of the accident, and Monday 1 November 2021.

The SDS joined Network Rail in late 2019 on the graduate entry scheme.

They were initially seconded as an assistant supporting the then incumbent Wessex route SDS, during which time they also gained experience as a track patrol assistant.

They later transferred to London Waterloo, returning to their SDS support role in early 2020.

After the incumbent SDS was transferred to another role, they took up the position of Wessex Route SDS on a permanent basis.

Railway systems and infrastructure involved

Signalling in the area is controlled from a signal box at Salisbury station.

On the Down Main line, Salisbury Tunnel Junction is protected by signal SY31, a three- aspect colour light signal.

Signal SY31 is located at 82 miles 25 chains, around 200 metres from the point where the collision occurred.

Although signal SY31 had been passed at danger twice before, in May 2001 and May 2006, neither of the signals passed at danger (SPADs) were documented as being caused by low adhesion.

Before reaching signal SY31, trains approaching the junction on the Down Mainline will first pass signal SY29R (80 miles 30 chains) and then signal SY29 (81 miles 66 chains).

These are approximately 3,320 metres and 980 metres from the point of collision, respectively.

If signal SY31 is displaying a red aspect, signal SY29R will display a double yellow (preliminary caution) aspect and signal SY29 will display a single yellow (caution) aspect.

The gradient beyond signal SY29R is 1 in 169 descending, before it levels out again on approach to signal SY29.

Beyond signal SY29 the gradient is 1 in 733 descending (from 81 miles 67 chains), increasing to 1 in 610 descending (from 82 miles 6 chains).

This latter gradient encompasses the line on approach to and beyond signal SY31 and towards the point of the collision and Fisherton Tunnel (82 miles 37 chains).

Signalling layout, gradient profiles and distances on approach to the point of collision

For trains approaching on the Down Dean line, the protecting signal for the junction is signal SY37, located 120 metres from the point of the collision.

To reduce the risk from signals being passed at danger, Automatic Warning System (AWS) and Train Protection and Warning System (TPWS) equipment is provided in the area controlled by Salisbury signal box.

AWS provides an audible and visual warning to a driver on the approach to signals as well as some other infrastructure features, including certain changes in permitted speeds.

AWS uses track mounted magnets which are detected by receivers fitted to trains.

When a train passes over a magnet, the system on the train sounds a bell or chime, if approaching a signal displaying a green aspect, or a horn if approaching a signal displaying any other aspect.

The system also sounds a horn if approaching a specified reduction in permissible speed.

The driver must acknowledge the horn warning by pressing a button on the driving desk within a specified time, otherwise the system on the train will make an emergency brake application.

The acknowledgement of a horn warning is shown on a visual indicator which changes to a yellow and black display known as a ‘sunflower’.

TPWS was developed in the mid‑1990s and implemented in accordance with the Railway Safety Regulations 1999 to automatically apply the train's brakes and reduce the risk arising from trains passing signals at danger.

It is fitted at signals on passenger lines which protect certain conflicting movements.

The system is also used to enforce reductions in permissible speed on approach to stop signals and speed restrictions and to intervene if trains approach buffer stops at too high a speed.

TPWS is not designed to prevent incidents of speeding or signals passed at danger (SPADs); rather, it is designed to reduce the consequences of such incidents by slowing trains or, in the case of a SPAD, by stopping a train before it reaches the conflict point at a converging junction.

The system is also not designed to fully protect all types of train at all speeds.

TPWS uses radio frequency transmitters (known as ‘loops’) fitted between the rails and receiving equipment on the train.

At signals capable of showing a red aspect, a pair of loops is used in a configuration called a train stop system (TSS).

If the features of the layout mean that a TSS alone would not stop a train before the conflict point, then a second pair of loops can be placed at a specified distance on approach to the signal.

These loops are known as an overspeed sensor system (OSS) and may also be fitted where there are significant reductions in permitted speed (that is, where permissible speed warning indicators are used).

The two OSS loops in a pair are placed a set distance apart according to the speed at which an intervention is required at that location.

The OSS and TSS loops for signals are energised when the signal is displaying a red aspect, whereas loops for speed reductions are permanently energised.

If a train passes over TSS loops when they are energised, the TPWS equipment on the train will demand an emergency brake application.

For OSS loops, the brake demand depends on the train reaching the second loop in the pair before a specified time period has elapsed, thus indicating that the train’s speed is higher than the ‘set speed’ for the TPWS installation.

The driver receives a visual indication that the brake intervention has occurred and must acknowledge it as part of the process of resetting the system before continuing.

Signals SY31 and SY29are fitted with TSS loops; both signals and the 50 mph (80 km/h) permanent speed indicator are fitted with OSS loops.

Trains involved

On the day of the accident, train 1L53 was formed of a single three-carriage class 159 diesel multiple unit (DMU), number 159102.

Although shorter formations are not unusual, this train service would normally have been formed of nine carriages.

On the day of the accident the train length was reduced due to service disruption.

South Western Railway 159102 runs solo through Potbridge working 1L46, the 13.47 Salisbury to London Waterloo service on October 3, 2020

Train 1F30 was formed of two class 158 DMUs, numbers 158762 (leading) and 158763 (trailing).

Each of these units comprised two carriages, forming a four-carriage train.

158762 at Westbury on 22nd November 2021

Class 158 and 159 trains are very similar and are constructed by welding aluminium extrusions together to form the bodyshells.

The carriages which formed trains 1L53 and 1F30 were built between 1989 and 1992 by British Rail Engineering Limited in Derby.

The braking system fitted to class 158 and class 159 units decelerates the train by supplying air to brake cylinders mounted on the train’s bogies.

These brake cylinders apply friction pads to brake discs mounted on the wheelsets.

A driver can apply three levels of braking in normal service.

Step 1 brake provides the lowest level of braking, while step 3 (known as ‘full service braking’) provides the maximum braking effort.

A driver can also make an emergency brake application.

This applies the same level of retardation as step 3 but uses a different control system to normal service braking so the train can still be braked in the event of a failure of that system.

Both trains involved in the accident were fitted with a wheel slide protection (WSP) system.

Analogous to the anti-lock braking system in a car, the WSP system monitors the rotational speed of the train’s wheelsets to determine if any have stopped rotating and are therefore sliding on the rail.

If the WSP system detects that wheel slide is occurring, the system automatically reduces the brakeforce being applied to the sliding wheelsets until the system determines that they are no longer sliding.

The WSP system is designed to optimise the train’s braking in conditions of low friction between the wheel and the rail, while also minimising the potential for the wheels to be damaged by sliding.

The system can also increase the available friction (and hence the available brake force) by injecting sand from nozzles into the wheel/rail interface of some wheelsets.

The sanding system is controlled electrically but uses the train’s air system to blow sand into the wheel / rail interface.

Sand is discharged automatically when the WSP system detects wheel slide in any brake step.

On class 159 units, such as that which formed train 1L53, sand is injected under both wheels of the third wheelset from the front of the train.

A sander system is also provided on the rear carriage of the three-carriage train, but this is inhibited when the carriage is trailing.

This means that only the nozzles on the front carriage of a three-carriage Class 159 unit will discharge sand.

Drivers of class159 units do not have a facility to manually operate the sanding system when the train is braking; a manual sand button is provided for use when the driver is applying traction power.

External circumstances

On the day of the accident, an unusual type of weather front passed through the Salisbury area, resulting in a decrease of air pressure concentrated on a particularly small region of around 30 km in diameter.

This type of weather front, known as a mesoscale low-pressure event, is a relatively rare occurrence in the United Kingdom, with only ten similar events reported between 2009 and the date of the accident.

The mesoscale event in the Salisbury area on the morning of 31 October 2021contributed to the worst recorded weather (in terms of combined wind and rainfall) data recorded by Met Desk and the Met Office so far that year.

The weather front brought localised strong winds and heavy rain to the area, causing major disruption to the Wessex route.

Weather stations located near to Salisbury recorded maximum wind gusts of approximately 73 mph (117 km/h), peaking between 08:00 and 10:00 hrs.

In the 12 hours preceding the accident, around 21 mm of rainfall was recorded, some 36% of the average monthly rainfall for the area.

After the weather front had passed, the area remained windy for the rest of the day with peak wind gusts of approximately 23 to 37 mph (37 to 60 km/h).

Between18:35 and 18:40 hrs, radar imagery showed a band of drizzle passing over the Up and Down Main lines in the area of Salisbury Tunnel Junction, just before the passage of train 1L53.

The sequence of events

Events preceding the accident

At 08:44 hrs on 31 October 2021, a train not involved in the accident, train 1L13, struck a tree that had fallen onto the Down Main line at 80 miles 70 chains, around 250 metres west of Broken Cross bridge (80 miles 58 chains).

The collision damaged the cab of the train, but no injuries were reported.

The driver of 1L13 reported that they came to a stop in the vicinity of the AWS magnet for signal SY29, which was approximately 1350 metres beyond the fallen tree.

On reporting this incident to the signaller, the driver stated that the train’s braking was affected by "slipping" before colliding with the tree, but the conversation did not identify railhead contamination as a factor.

There was no mobile operations manager (MOM)5 available from Network Rail to attend the incident, so the train's driver and guard cut back the branches that were foul of the railway line, leaving the tree in the cess of the Down Main line.

The line was reopened at10:44 hrs.

The incident near Broken Cross bridge was one of fifteen weather-related railway incidents that occurred on Network Rail’s Wessex route that day.

During the morning other incidents involving fallen trees and flooding caused significant disruption, including a tree blocking the Down Main line at the Salisbury (west) portal of Fisherton Tunnel.

The tree struck by train 1L13 (post-collision) on the Down Main line looking east towards London with Broken Cross bridge in the background

The driver of train 1L53 booked on for duty at Salisbury depot at 10:30 hrs.

He was originally due to travel to Honiton to start his driving duties, but the effect of the severe weather conditions on the timetable led to several trains being cancelled.

The driver therefore stayed in the mess room in Salisbury depot for the first part of his shift.

At around 14:50 hrs, the operational resource manager for SWR requested that the driver take train 1L52, the 15:27 hrs service from Salisbury to London Waterloo.

Train 1L52 was formed of two class 159 DMUs, each of three carriages forming a six-carriage train.

The driver walked to the platform at Salisbury to meet the incoming train which would form train 1L52.

Although there is witness evidence that he met another driver on the way who warned him of the adverse weather conditions, the driver stated that such a meeting did not occur.

Later, while the driver was in the train's cab preparing train 1L52 for service to London, another driver contacted him from the train’s rear cab using the cab-to-cab telephone.

This was to warn him about poor rail adhesion on the Down Main line approaching Salisbury, which had been experienced by the driver of the incoming train from London Waterloo.

The driver acknowledged the warning and thanked his colleague.

Train 1L52 departed from Salisbury at 15:27 hrs.

Approaching Broken Crossbridge on the Up Main line towards London Waterloo, the driver, having been informed by his colleagues that there was a fallen tree in the vicinity, was looking for and noticed the remains of the fallen tree struck by train 1L13 on the adjacent Down Main line.

During the journey to London, the bad weather subsided, and the sun set at 16:42 hrs.

The driver reported no conditions of low adhesion or any other difficulties with the train’s braking or acceleration during this journey.

On arrival at London Waterloo, the driver changed trains, leaving train 1L52 and taking train 1L53 which departed from London Waterloo one minute late at 17:21 hrs.

Soon after departure, the driver performed a routine running brake test in the Vauxhall area, from a speed of 53 mph (85 km/h).

The driver reported no issues with the braking capability of the train during the test.

No further running brake tests were carried out during the journey and none were required by the SWR professional driving policy, unless the driver judged there was a problem with the train’s braking.

The train called at Clapham Junction and Woking, where it departed six minutes late due to being held at a signal.

The train then stopped at Basingstoke and Andover, departing Andover six minutes late at 18:30 hrs.

It was dark during the journey from London Waterloo and weather conditions were dry.

The driver reported that he did not experience any WSP activity nor any instances of low adhesion that affected the train’s ability to slow, stop or accelerate at station stops during the journey.

Between 18:35 and 18:40 hrs a band of drizzle moved south-eastwards over the Up and Down Main lines in the Middle Wallop area, near Salisbury.

The drizzle ceased in this area before train 1L53 approached it.

At 18:41:09 hrs, approximately 1 minute 47 seconds before the accident, the OTDR from train 1L53 recorded that the traction power had been shut off.

The train then coasted down the prevailing 1 in 169 gradient towards signal SY29R.

Ten seconds later, OTDR data showed that the driver acknowledged the AWS warning horn for signal SY29R.

Signalling data and forward-facing closed-circuit television (FFCCTV) showed that this signal was displaying a double yellow aspect.

This was because signal SY29 was displaying a single yellow aspect, and signal SY31 was displaying a red aspect.

The driver approached Broken Cross bridge at about 18:41:36 hrs with the train still coasting.

The driver stated that Broken Cross bridge was the point where he would normally start braking in anticipation of stopping at signal SY31.

However, because he felt that there was a potential for low adhesion conditions at the location of the fallen tree, he decided to delay braking and use the fallen tree and associated debris as the marker point to start braking.

The driver believed that this would still leave sufficient time and distance to bring the train to a stand before reaching signal SY31.

Tree that was struck by train 1L13 (branches had been cut back by the train driver and guard) and later observed by the driver of train 1L53 (when driving train 1L52 on the Up line to London Waterloo)

Enhanced image taken from train 1L53 FFCCTV system

Events during the accident

Around 18:41:42 hrs, train 1L53 coasted past the location of the fallen tree and debris at a speed of 89 mph (143 km/h).

Eight seconds later, at18:41:50 hrs, the driver acknowledged the AWS warning horn associated with the warning sign for the approaching 50 mph (80 km/h) speed restriction.

At 18:42:02 hrs, with the train travelling at 86 mph (138 km/h) and around1,560 metres beyond signal SY29R, the driver, having not noticed the fallen tree, applied the brakes.

At this point, the train was on a descending gradient of 1 in 169, approximately 780 metres on approach to signal SY29, and approximately 1,560 metres on approach to signal SY31.

Evidence from the train’s OTDR data shows that the driver initially selected brake step 1, followed almost immediately by brake step 2.

When the driver applied the brakes, he stated that he immediately felt and observed that something was wrong.

The driver stated that the feeling ‘under his seat’ indicated that the train's wheels were sliding.

The driver realised that the train’s speed was not reducing as he expected.

Six seconds after his initial brake application, at 18:42:08 hrs, the driver made a step 3 (full service) brake application.

This was followed a further six seconds later by the driver making an emergency brake application.

Equipment fitted to the train recorded that the WSP system was active from the start of braking and throughout the various brake applications, and this would have demanded the sanding system to discharge sand onto the railhead.

However, this had a limited effect on reducing the speed of the train.

At 18:42:24 hrs, train 1L53 passed signal SY29 (displaying a single yellow aspect) at approximately 77 mph (124 km/h).66 At 18:42:42 hrs, the train’s TPWS detected that the train was travelling above the set speed of 34.5 mph (55.5 km/h) for the OSS of the approaching signal SY31, which was showing a red aspect.

The TPWS therefore commanded an emergency brake application.

However, this had no additional effect as the driver had already made an emergency brake application.

At 18:42:49 hrs, train 1L53 passed signal SY31 and approached Salisbury Tunnel Junction.

About 40 seconds earlier, GWR train 1F27 had passed over the junction on the Up Dean line on its journey from Salisbury to Southampton.

Meanwhile, train 1F30 was approaching the junction on the Down Dean line heading towards Salisbury.

The driver of 1L53 saw train 1F30 appear from the left and move into the path of his train.

Believing there was little he could now do to prevent a collision, he got out of his seat with the intention of exiting the cab into the saloon.

However, in doing so, he tripped over his bag, and fell onto the floor just before the collision occurred.

As train 1F30 was entering Fisherton Tunnel, train 1L53 collided with its right‑hand side, near the front of the fourth carriage.

Evidence from the OTDR on train 1L53 suggests that the collision occurred at 18:42:57 hrs, with train 1L53 travelling at a speed of approximately 52 mph (84 km/h).

The OTDR on train 1F30 recorded its speed as being around 20 mph (32 km/h) at the time of the collision.

The 32 mph (51 km/h) speed differential between the two trains meant that train 1L53 travelled forwards in heavy contact along the right-hand side of the fourth carriage of train 1F30.

It then continued along the right-hand side of the third carriage, with the leading left-hand edge of the cab of train 1L53 embedding itself in the front end of this carriage.

Diagram showing train 1L53 (carriages marked as B1 to B3) and train 1F30 (carriages marked as A1 to A4) immediately before the collision

Diagram showing train 1L53 colliding with train 1F30

Diagram showing trains 1L53 and 1F30 derailing into Fisherton Tunnel

The fourth carriage of train 1F30 was displaced to the left by the impact, resulting in its leading left corner coming into heavy contact with the facing portal wall of Fisherton Tunnel.

This caused a failure of the coupling between the second and third carriages of train 1F30, opening a gap of 23.8 metres between the two halves of the train.

At this time, passengers intending to leave train 1F30at Salisbury station were still seated, and no passengers were standing in the vestibule area where the train parted.

Impact mark from carriage A4 on facing portal wall of Fisherton Tunnel. Vestibule of A3 is visible and carriage A4 has been removed

Both trains travelled into Fisherton Tunnel, with the tunnel walls acting to keep the carriages upright.

The rear two carriages of train 1F30 and all carriages of train 1L53 derailed, with the cab of train 1L53 being severely damaged.

The front carriage of train 1L53 came to rest adjacent to the leading end cab of the third carriage of train 1F30.

Diagram showing carriages A2 and A3 separated within Fisherton Tunnel

View from London portal of Fisherton Tunnel showing the rear carriages of train 1F30 (A3and A4) and front carriages of train 1L53 (B1 and B2)

Events following the accident

The driver of train 1L53 was injured in the collision and was trapped in the heavily damaged cab.

At around 18:46 hrs, the guard on train 1L53 attempted to contact the driver but was unsuccessful in doing so.

The guard then tried to make a railway emergency call to the signaller using the GSM-R8 train radio in the rear cab, but this again was unsuccessful.

They then used their mobile phone to contact the emergency services and provided information from 18:45 until 19:15 hrs when they were met by responding fire and rescue personnel.

View from within Fisherton Tunnel showing the front cab of train 1L53 (B1) and carriage A3, having separated from carriage A2, of train 1F30. Parts of the driving cab of train 1L53 have been cut away by the emergency services to release the trapped driver

The driver of train 1F30 felt the impact of the collision and immediately made a GSM-R railway emergency call, reporting to the signaller at Salisbury that they believed their train had derailed.

In response, the signaller placed all signals in the area to red.

On completion of the emergency call, the driver of train 1F30 left their cab to investigate what had happened.

They walked back through the train and found the rear two carriages had become detached.

Realising then that another train may have collided with train 1F30, they returned to the cab and made another railway emergency call to the signaller to update them with this information.

From 18:46 hrs onwards, a number of calls were made to the WICC at Basingstoke.

Due to the volume of calls and the chaotic nature of the scene in the tunnel, around 17 minutes elapsed following the collision before staff at the WICC made contact with the guard of train 1L53 and established that it had also been involved in the accident.

Both GWR and SWR mobilised ‘on call’ teams to the accident site.

As members of train crew at Salisbury station and depot became aware of the accident, they also went on foot to the scene to offer assistance to passengers.

On arrival, they found passengers were already trying to evacuate from train 1F30 via a window that had been broken.

Network Rail and SWR staff from Salisbury station arrived at the scene and, at 19:06 hrs, Network Rail declared a major incident.

Wiltshire Police and Dorset and Wiltshire Fire and Rescue Service arrived at the scene at 19:10 hrs and confirmed the declaration of a major incident with the other agencies attending (such as the South Western Ambulance Service and British Transport Police).

By 19:56 hrs the fire and rescue service had reported that, with the exception of the train driver of train 1L53, all passengers and staff had been evacuated from the trains and had exited from the tunnel.

At 20:45 hrs, the fire and rescue service reported that the evacuation was complete.

The damage to the rail carriages prevented the movement of passengers between some carriages.

Consequently, it was necessary to evacuate passengers simultaneously from both the front and rear of the trains involved.

Some passengers, predominantly those in the leading three carriages of train 1F30, were evacuated to Salisbury station.

Passengers in the remaining carriages were evacuated from the rear of the trains and walked trackside to a railway access gate.

RAIB was notified of the accident at 18:53 hrs and immediately deployed a team of inspectors, who arrived on site at 21:20 hrs.

RAIB handed the accident site back to the railway in stages, with the final areas being released on 7 November 2021.

Repairs and reinstatement of the track and signalling were completed on15 November 2021, with services through Salisbury Tunnel Junction resuming on16 November 2021.

Analysis

Background information

Wheel/rail adhesion
Because trains rely on friction between wheel and rail to stop, the level of wheel / rail adhesion available is critical to the rate at which a train can decelerate.

Research indicates that the level of adhesion available is mostly affected by railhead contamination and moisture on the railhead.

Railhead contamination can be caused by a number of things, but the most common cause is leaves falling onto the railhead and then being compressed by the wheels of trains rolling over the material, building different layers and an increased thickness of contamination.

The layers of leaf fall are created and bonded as a result of chemical reactions occurring during the rolling and sliding of wheels over the contamination.

The level of adhesion between a train’s wheel and the railhead is normally expressed as a coefficient of friction.

The lower the value of the coefficient of friction, the lower the adhesion between wheel and rail.

Adhesion levels can vary considerably over relatively short distances and timescales.

Range of railhead adhesion on the rail network as defined by the Adhesion Working Group

Research undertaken on behalf of the industry’s Adhesion Working Group (AWG) led to it publishing guidance for the industry in 2006, most recently updated in 2018.

This provides the following comment on low adhesion conditions:

‘The rail surface and the wheel treads can become coated with a range of contaminants. The worst of these are crushed leaves, which, when combined with moisture particularly in the form of dew or condensation, reduces the adhesion level. For a train on dry rails adhesion is typically around 0.25, on wet rails it is around 0.15, but on damp leaf (layers) it can be as low as 0.015. Rails with damp leaves (layers) significantly constrain the rate of braking.’

Guidance on the thickness and levels of contamination and action to be taken are included within Network Rail standard NR/L2/OPS/045-4.07,‘National Operating Procedure, Section 07 Railhead contamination levels’ (issue 2, December 2019).

Leaf layer thicknesses will typically measure up to 0.2 mm but can be thicker when the base layers become bonded to the rail.

Train wheels under acceleration and braking can result in the contamination ‘creeping’ (a series of very small horizontal movements), which can further increase the thickness of any railhead contamination present at certain points.

Sliding train wheels, in contrast, may have the opposite effect and result in layers being removed.

Post-accident examination of the Up and Down Main lines between signal SY29R and the accident site showed that, although the topography and tree species were very similar in number and density either side of the railway, the levels of railhead contamination varied, with contamination on the Down line being medium to heavy at locations where it was only minimal on the adjacent Up line.

This demonstrates how localised changes in wheel/rail adhesion conditions can be.

The speed at which layers of contamination build up is dependent on a number of factors, including the number of trains that have passed over the location since the last treatment, train braking and the leaf fall within a given period.

The rate of leaf fall is in turn affected by the environment, meteorological conditions, topography (such as cuttings and embankments) and tree species.

However, the relationship between the rate of build-up of contamination and these factors remains largely undefined.

For the purposes of managing wheel/rail adhesion, Network Rail considers the autumn leaf fall season to run from 1 October to 13 December each year.

While the rate of leaf fall can vary depending on weather conditions throughout the year, the rate of fall normally begins to accelerate in the middle of October with the ‘peak leaf fall’ period occurring from around 22 October through to around 31 October.

A further acceleration of leaf fall can also occur from around 31 October through to around 24 November, when around 40 to 60% of all remaining leaves normally fall from trees, depending on the species involved.

The cross-industry Adhesion Research Group (ARG) promotes understanding of wheel / rail adhesion, train detection and related topics, and ensures a common and system-wide approach to these issues.

It acts as a sponsor for research projects and co-operates with the Seasonal Challenge Steering Group and Seasonal Challenge Communications Group to implement research on the running railway.

RSSB in collaboration with the Adhesion Research Group developed an adhesion research programme called ADHERE.

The programme has been running since 2018 and continues to provide research to improve the rail industry's knowledge, and management, of low adhesion conditions.

The programme includes workstreams investigating: the modelling of low adhesion and braking; the effectiveness of rail cleaning activities and treatments; driver behaviours; current train-borne technologies, including double variable rate sanding and magnetic track brakes; and adhesion forecasting and observational capabilities to improve real-time decision-making.

Topography data showing the embankment/cutting profiles, including the previously classified high risk of low adhesion site (pink shaded area - see paragraph 143), the location of Broken Cross bridge, tree/associated debris (green dots), the braking points of train 1L53 (red dots) and signals

Post-accident sample locations

Network Rail classification of contamination levels (NR/L3/OPS/045/4.07)

Location (left) and railhead condition (right) at 80 miles 40 chains

Location (left) and railhead condition (right) at 82 miles 27 chains beyond signal SY31 andapproaching Salisbury Tunnel Junction

The main lines between Overton and Whitchurch looking towards Salisbury showing the condition of the vegetation before (taken in summer) and after (taken in winter) vegetation work was completed in 2019/20

Network Rail (Wessex) data in diagrammatic format showing mileage, topography, tree species and leaf fall risk assessment data scores (June 2021)

RAIB diagram showing the action zones from Network Rail standard NR/L2/OTK/5201, train profiles similar to class 159 trains, and tree growth as found on certain sections of the main lines between signals SY29R and SY31 (red dotted line)