Train punctuality – WWJD?

Last week the government announced timetable changes, including adjustments to the Cranbourne/Pakenham line. (I’ll just call it the Dandenong line for short. Almost everybody else does.)

Of the roughly 265 services per weekday, 93 will be adjusted. A few will be quicker by a minute or two, but most of the 93 will be slower – most of them by 1-2 minutes, but a few by up to 5 minutes.

During peak hour, I suspect most people won’t notice, because between the City and Dandenong it’s usually not worth looking at a timetable – there’s generally a train every 3-6 minutes.

Beyond Dandenong it’s less frequent, of course, and off-peak 20 minute frequencies kick-in as early as 8am on weekdays at Pakenham.

Striving for the impossible?

But let’s cut to the chase. What’s the point of an impossible-to-achieve timetable?

Over the past 12 months, this line has had the worst punctuality on the network (October 2018 to September 2019: Cranbourne 84.8%, Pakenham 85.9% within 5 minutes)

If some running times can’t be achieved, it absolutely makes sense to make changes.

But they also need to look at the causes of delays, and identify fixes.

  • Are there infrastructure issues? (Yes, there have been major disruptions due to new infrastructure, but is it also causing every day delays?)
  • Is the Cranbourne single track contributing to delays? (It’s hard to believe it’s not. Even the government’s own web site says duplication would remove the bottlenecks that cause delays and allow the number of train services to be doubled during peak times)
  • Is the flat junction at Dandenong causing issues?
  • What about the sharing of the City Loop tunnel with Frankston trains? Nobody could pretend that doesn’t add to track congestion.
  • Peak crowding? I’m sure that’s contributing. So how about boosting the service? Again, that’s possible if the Frankston trains come out of the Loop.
  • Not all of the changes are in peak hour. For instance the 1:40pm to Pakenham currently arrives and terminates at 2:58pm. Under the new timetable it takes 2 minutes longer. If services like this are taking too long now due to crowding, is it a sign the base off-peak service needs boosting too?
  • Will the new HCMTs help cut station dwell times? (And are other measures such as platform staff and announcements and the “burn lines” helping?)

The preference should be to fix issues like these, and then adjust the times…

We know they’re working on timetable changes.

Realistically, right here and now, what do they do?

And why has this been ignored for so long that some trains need to be given five minutes longer?

Crowded train, South Yarra

What would happen in Japan?

One reaction I saw from multiple people was along the lines of:

Slowing down trains? Outrage! This would never happen in Japan!

Um well… scrupulously punctual trains don’t just happen by magic. They have to work at it.

So what do they do in Japan? A little research, and here are some answers, courtesy of this paper.

It’s worth noting that they’re often looking at some seriously intense railway operations. The paper looks at the Tozai line, which carries 1.6 million passenger trips per day:

In many railway lines in big cities, 25 – 30 trains are running per hour per direction on a double track line during morning rush hours. This means trains are running every two to three minutes in one direction. Still, trains are very congested and it is not unusual that more than 2,000 passengers are aboard a train which consists of ten cars and is 200m long.

That’s far more and longer trains than on the Dandenong line, though that’s where things are likely to be headed in coming years.

How do the Japanese railway companies handle it? The paper says they look at their infrastructure and try to make the most of it:

Timetable planners are very cautious to avoid conflicts imposed by capacity constraints.

And station dwell times are certainly an issue:

Because trains are operated very densely, once a delay (primary delay) happens, the delay is propagated to other trains and a lot of trains are also delayed (secondary delay). Primary delays are quite often caused by passengers. If more passengers than expected get on/off a train, the dwell time becomes longer and the train is delayed.

The paper goes on to say that if infrastructure can’t fix it, or it isn’t practical to upgrade it, then yes, absolutely they do adjust train timetables to improve punctuality:

Various factors are relevant to the robustness of timetables. In particular, improvement or increase of facilities such as construction of new tracks is quite effective to increase robustness of timetables, but usually prohibitively expensive and sometimes impossible due to limitation of spaces. So, railway companies are more interested in improving robustness of timetables by slightly modifying them.

And it flags some of the options for modifying the timetables:

…potential adjustments are:

Dwell time of Train Z at Station A should be increased by five seconds.

Running time of Train W from Station B to Station C should be increased by 10 seconds.

Departing order of Train S and Train T should be changed so that Train T departs first.

Train U should stop at Station D.

The whole paper is worth a read if you’re interested in such things.

So then… do they slow down train timetables in Japan to maintain punctuality? Yes they certainly do.

I’d love to visit Japan one day. But whenever and wherever I go on holiday, I’m wary of observing things just as a tourist, and I try not to make assumptions about how or bad and efficient things are, based on limited experience.

Dandenong line, 6pm

So is our State government off the hook? Certainly not. The question is: are the root causes of delays being looked at? And we know Dandenong to Cranbourne is to be duplicated – by 2023 apparently, almost 30 years after suburban trains started on that line. Why has it taken so long?

And how about the other single tracks around the Metro network: Upfield, Belgrave, Lilydale, Hurstbridge, Altona Loop? They all cause problems – delays, and short shunting to prevent flow-on delays.

This isn’t Puffing Billy, a tourist attraction with a train only every couple of hours. Melbourne’s rail network underpins the entire economy.

If they are serious about the public transport network, then sufficient infrastructure is a must, and the other barriers to punctuality all need to be looked at and resolved.

By Daniel Bowen

Transport blogger / campaigner and spokesperson for the Public Transport Users Association / professional geek.
Bunurong land, Melbourne, Australia.
Opinions on this blog are all mine.

20 replies on “Train punctuality – WWJD?”

I’m hoping the HCMTs coming online will help with dwell times, particularly when compared to the Siemens trains on the Pakenham/Cranbourne line. With only two doors per carriage, they take noticeably longer to unload then load than the three door Comeng trains. An extra carriage, sufficient doors and a ‘metro’ style layout should help.

All these timetable movements have had a knock-on effect to the Gippsland line, with a number of trains being forced to leave Bairnsdale or Traralgon earlier to compensate for an even longer suburban crawl.

I think the thing that rubs me, and probably a lot of other people the wrong way about this is that it is all ‘take’. There are a lot of disruptions and changes to accommodate operational improvements but with what feels like very little pay-off.

I completely understand that long dwell times mean that timetables have to be adjusted and slowed down for punctuality purposes, but Melbourne train patronage growth has been exceptionally poor, undershooting all targets and expectations, and is turning into a real disaster (it is far less than population growth, which means PT utilisation is going backwards). As outlined in this post, there is plenty that can be done using just timetabling and routing changes to take advantage of additional capacity in the network – or allow for more off-peak trips etc. (except that frequencies are abysmal, as bad or worse than dinky networks like Auckland).

Similar with things like the new train display ‘burn lines’. The state government won’t bother to improve frequencies using existing network capacity, so will make the network as user-unfriendly as possible to accommodate.

This is something I have also noticed with many of the disruptions for level crossing removals and so on. Apart from some small benefits in terms of new stations and so on, the bulk of the benefits are for drivers. There have been no real (stemming from the removals) service improvements following significant bus replacements and disruptions, so for train passengers, there is feels like there is little benefit overall.

Of course, nothing is ever that simple, and there are dependencies and complexities throughout the system. But when everything starts to look like an excuse, or more and more new reasons are fished up why service just cannot be improved, it shows something is really getting off track. If they spent half as much effort on delivering simple service improvements – many of which are low-hanging fruit – as they do spinning new reasons why nothing can ever be improved despite huge capital outlays, then we might have a passable network. In the mean time, the statistics show that the rate of public transport usage continues to fall every year.

“Is the Cranbourne single track contributing to delays?”

Definitely. Trains from the City to Cranbourne in evening peak are occasionally terminated at Dandenong if it is running late. Sometimes the fact that it is running late means a train coming the other way (Cranbourne to City) is using the single track already, so instead of stalling the outbound train at Dandenong while it waits, they just terminate it there instead and tell everybody to wait for the next outbound train to Cranbourne.

In fact, just yesterday this happened – the 4.09 PM Flinders Street to Cranbourne was terminated at Dandenong at 4.59 PM. This meant the 5.55 PM Cranbourne to City originated at Dandenong at 6.08 instead, which resulted in Cranbourne going for 62 minutes without a citybound train (one train at 5.21 PM, then the next train at 6.23 PM).

Those ‘burn lines’ don’t seem to start when the train opens its doors, either. They start when the system has decided the train is there, which often happens before the train even stops. So people approaching as the train slows to a halt, and concentrating on where they walk, can then arrive at the doors when they’re about to open and find that they can’t read a display telling them where the train is going.

One of the things I’ve noticed on a visit to Japan about the metro system is that they’ve really put a lot of dwell times at really busy stations. Trains waiting around an entire minute or longer at a platform is quite normal, to allow passengers to get off and get on, despite there being 4 doors a side per carriage, for a train with as many as 10 carriages. One of the terminating trains I’ve seen in Kyoto even had a platform pull into a dual-side platform to allow faster offloading of passengers. From what I was told, these dwell times are all put into the timetable, together with the stations and timing points that the drivers follow.

I can’t assume whether this is done here in Melbourne, but I have noticed trains that dwell for as little as 15 to 20 seconds when there are only a handful of passengers getting on and off, but staying longer at a different time when there are more passengers at the same station. The passenger timetable doesn’t show much of a different in timing. There is a bit more of a contrast on the V/Line timetables, where arrival and departure times are shown for major stations, with anywhere between 1 to 5 minutes between them.

I also feel that some lines here would benefit from a timetable fix, even if it means longer timetabled services. The media may portray that the trains seems “slower”, but as it is now at peak time, an 8.12 train at Laverton to Flinders Street for example often arrives between 8.14 and 8.16 in actual time, and this has a flow on effect of running other trains later, not only on the Werribee line but also on the Altona Loop and Williamstown services, and eventually other services pulling into Flinders Street Station. Perhaps the trains now can make up for lost time in the expresses between Laverton and Newport, and Newport and Footscray, but this advantage is often given up when congestion at Footscray causes the trains to wait before arriving at Footscray anyway. IMO, this can be fixed by starting the trains at Werribee station minutes earlier, and giving longer dwell times at stations between Werribee and Laverton. The recent implementation of countdown bars on platform display screens has helped a bit.

On a separate topic of “perceived slowness”, I am sure most passengers would hate the idea of the Limited Express services on the Werribee line being forced to stop at more stations and would make it appear slower, but I think making all services between Newport and Footscray stopping all stations would allow even more timetabled services on the Altona Loop, Werribee and Williamstown lines It will at most add 2 to 3 minutes IMO. As it is now, there is a need to buffer the timetable to allow the Werribee services.

Another observation I could make of Japan is the use of useful signages on the platform. The platform signs will say which positions on the platforms the train will stop at, and mark special carriages if required, e.g. 1-10 (10 cars) , 1-12 (12 cars), 4-6 (3 cars), Green Car 6-7 (premium seats, middle of the train).

In the case for Osaka, there are special symbols representing different types of trains marked on the platform, e.g. ● circle, ■ square or ▲triangle, which is useful for the passengers to know where to stand on the platform for quick boarding. Like in Melbourne, the Greater Osaka network (and probably a number of other cities) face the issue of catering to different rollingstock with 2, 3, 4 or other combinations of doors per carriage. Passengers knowing where to stand would allow for quicker alighting and boarding. I’ve spoken to people who are sceptical that this might work here, but I remember that platform markings in Singapore wasn’t introduced until years after trains have started running, and has become significantly less chaotic on the platforms.

I’m disappointed that the HCMT’s did not introduce four doors per carriage. This is a lost opportunity to further decrease running time. Perhaps on whatever replaces the rest of the Comeng?

I wonder if part of the problem on the Dandenong line re secondary delays was the decision to use Home signals all the way along the sky rail sections instead of Automatic signals, and the rule changes that mean drivers can no longer pass red lights at reduced speed after waiting for 30 seconds.

Instead they have to talk to the signaller, which theoretically improves safety but also extends the delay and propogates it to following trains.

What I feel no one is pointing out is that the timetable is going to be closer to the actual running times. The trains aren’t going to be slowing down, the timetable will just match the realistic time taken for a train on the Dandenong line during peak hour. The line is a lot less padded then those on the northern loop and the difference in time for peak and off peak journeys between Dandenong and the City is as little as 3 minutes. I do not remember a day my 3:43 Pakenham from southern cross arrived into Dandenong less than 5 minutes late in the 5 years I caught it.

What should have happened was the Dandenong line be quadrupled during the level crossing removal project to allow Vline services to run separately to metro services. The Cranbourne line should have also been duplicated at the same time too.

Totally agree Phil
This was called for by many but escaped scrutiny because of the anti sky rail mob
And guess what, they loved sky rail but didn’t know it,, another missed opportunity unfortunately it is a Melbourne thing

@Chris, even some insiders were disappointed the Siemens trains had only 2 doors. The total door width can be a factor (and the Siemens doors are wider than the Comeng) but the number of doors influences flow inside the carriage, and affects dwell times as people make their way to/from the doors.

@Cam, yeah hopefully the bigger timetable changes next year will finally add much-needed services to the Dandenong line to actually use the capacity gained by the skyrail.

@Philip, the worst Burn Line example I’ve heard about is at Upper Ferntree Gully, where the outbound train’s details disappear off the screen even though it’s still waiting for an inbound train to clear the single track. That’s just silly.

@Arfman, over time I hope we’ll see more separation of lines, conflicts become less important and specific timing isn’t required so much. During peak at least, put the trains on the line and try and maintain a high frequency, rather than specific times.

The issue on the Newport to City section is the remaining level crossings. Places like Yarraville would suffer long delays (with people unable to even reach the station on foot) if there’s a constant stream of trains.

@Lachlan, I think the powers that be have decided the long trips on Melbourne’s network means a certain number of seats are required. (The 7-car HCMT has about the same number of seats as a 6-car Siemens or X’Trap, but a lot more standing room)

@Canio, good input, thanks.

@Phil/michael, they were never going to build 4 tracks. The amount of property acquisition required made it exponentially politically more difficult than the skyrail was. But with the signalling and HCMT upgrades, it is a significant boost in line capacity.

They would certainly never build extra tracks just for the sake of a handful of V/Line trains. It’s not like City to Sunshine with 3 different V/Line lines feeding into it.

They probably could start looking at an additional 2 tracks outbound from Oakleigh, where there’s space.

(By the way, I find it funny that some Dandenong line peeps talk about having no express trains, when almost every service runs nonstop between Caulfield and South Yarra.)


I agree that it will create issues with gates down most of the time. A bridge or underpass will need to be provided at Yarraville station. Also, why don’t more stations have ways to get across to the other platform other than through road crossings?

I think the Anderson Street crossing should just be shut, with Somerville Road and Francis Street already being sufficient nearby crossings.

Hudsons Road crossing Spotswood station will be more of a problem, requiring a costly grade separation.

In the longer term, I hope the government will eventually go ahead with Melbourne Metro 2, fixing the Newport and Footscray corridor.

No one seems to understand that there’s recovery time in the schedules as trains approach a terminus. For instane, Frankston line trains take two minutes longer on the down (from Kananook to Frankston) than they do on the up (F to K).

Many Pakenham line trains are allowed eight (some six) minutes from Cardinia Road to Pakenham, but on the up from P to CR it’s usually five minutes start to start.

This is the sort of ‘gift’ to Metro Trains Melbourne that the current government refuses to remove, so depending on how late a train was earlier in its journey, it may arrive at the terminus within the allowable margin to be considered ‘on time.’ Hence, Metro will pay a smaller penalty (or conversely, receive a higher bonus).

A significant proportion of the lengthened running time services on the Pakenham/Cranbourne line are in the afternoon peak shoulder, where neither the infrastructure or the use of the loop by stopping all stations Frankston services constrains the ability to run extra services because they are running at off-peak frequencies. At these time, frequencies should have been increased (2 extra trains per hour, 1 each to Pakenham and Cranbourne, with the reorganised for even gaps between train with the new frequencies), rather than lengthening train running times.

@Edmund – without having checked the WTT, I’d guess that the down arrivals at Pakenham that are given extra time are more likely to arrive into No.1 road than No.2, accounting for the 40 km/h speed restriction across 7 points, plus the signal layout being designed as an intermediate crossing station rather than a proper terminus. Upgrading the turnout, modifying signals 4 and 6 and adding a down signal PKM10 between points 11U and PKM14 (either side of the level crossing, as space permits) should help significantly. There may also be an argument for a signal PKM28U on the down line between 11D and PKM16, but that is less critical.

@Arfman : Regarding timetables including stopping times, Swiss train timetables do that too. It’s particularly noticeable on the network graphic published by ZVV[1].

A lot of European lines make use of additional tracks and platforms at stations to schedule passing trains. Japan does this too, to a much greater extent, and most lines feature three to five service patterns, all running regularly with timetabled overtaking moves. Melbourne, however, has hobbled this by removing or booking-out the refuge platforms that could enable express trains without complex and expensive quadruplication projects—examples include Oakleigh and Essendon. Furthermore, we don’t use high-speed turnouts and have antiquated signalling, so trains have to slow down much earlier than is necessary, and can’t leave refuges and get up to speed quickly once back on the main. At the service frequencies proposed for the outer extremities of the Upfield, Cranbourne, Lilydale and Belgrave lines, the Swiss probably wouldn’t duplicate the track. They’d focus instead on well-placed passing loops with line-speed turnouts, additional platforms at stations, and proper timetabling to minimise delays.


I can’t imagine there’s much incentive for Metro to actually run the trains slower, but certainly running a timetable that actually reflects running times makes perfect sense.

I’m not sure what the solution will be for the Footscray to Newport corridor, other than it purely being for Williamstown (and maybe Altona Loop) with Werribee via Metro 2. Any grade separation at Anderson St will be very problematic, and closing crossings doesn’t happen – look at New St Brighton.

Just on the Altona Loop, is there any capacity to actually run it as a loop – i.e. one-way running to Laverton and running directly back on the express line?

In regard to passengers congegating at the spots where the doors will be – the best example of this I see is Tarneit – but even with that there are probably still relatively large dwell times, because you can only embark and disembark the V/Locity trains at each end of the carriages in single file. Definitely not designed for suburban commuting.

Comments are closed.