TT #187: More Frequency means More Go: Attitudes to PT frequency

If you're catching a plane to London or New York you're making a serious journey that is typically planned days, weeks or months in advance. Unless you're in the top 1% such a trip would be made rarely. On account of its significance, cost and rareness you're happy to rearrange your life schedule around it. Because of that it doesn't much matter much if flights are only a few times a day (or less). 

There may be changes to another flight if flying to smaller airports but the wise traveller plans long enough stopovers to lessen risks of missed connections. And you'd have to hope that there is good enough land travel at the other end to your destination. If there isn't it's accepted that you'll either hire a car, get a taxi or have the people you are visiting drive an hour or two to the airport then back to their place. Similar applies for transcontinental rail trips, especially the tourist types.  

Why frequency is key for metropolitan transit

Metropolitan transit is the opposite to all the above. For a start trips are shorter, cheaper and thus more frequently made. Travel time predictability is essential as travel may be to time-sensitive jobs, school lessons or appointments. But finish times can vary so the system must also be robustly amenable to changes of plan. 

Want all the above in on a public transport network? Good frequency on as many routes as possible is key. If you don't have it then the network isn't very useful as waits can easily exceed travel times. Yesterday I devised the 'five smileys' diagram below that I think summarises this neatly.  

The assigned frequencies and attitudes, I think, suit Melbourne and suburbs. 

Prevalence and user experiences of various service levels

Every 40 min

Melburnians loathe 40 minute frequencies yet they plague our bus network, including on some of our most productive routes in western and northern Melbourne. I have no doubt that 40 minute frequencies cause much of the deserved low regard that buses are held in. Key train lines to destinations like Wyndham Vale, Sunbury, Broadmeadows, Coburg, Epping and Greensborough are also only every 40 minutes, notably Sunday mornings but sometimes other times as well. The short-changing of the north is very obvious in the frequency map presented last week. 

40 minute headways lack the 'memory timetable' that services every 60, 30, 20, 15 and 10 minutes have. Perhaps their only merit is that they harmonise with trains every 20 minutes, as typical in Melbourne's west, north and outer south-east. Hence 40 minutes is a very common off-peak frequency for bus routes, even though the patronage on some in areas like Tarneit, Point Cook and Craigieburn would easily justify a 7 day 20 minute service.  

Every 30 min

30 minute frequencies are what make our SmartBuses not so smart on weekends and most of our rail network inconvenient for evening and occasionally also daytime travel. Even trams have half-hourly gaps on Sunday mornings and evenings. Like with every 40 minutes, a 30 minute wait due to a just missed connection, can double travel time for most trips. 

Still there's one thing worse than 30 minute waits and that's the abovementioned 40 minute gaps. Getting this to every 30 min is a small but worthwhile improvement. And it's super-cheap. This is because a 30 minute frequency is four trips every 2 hours versus three for a 40 minute frequency, you just need to add one return trip per two hours to boost the service. 

To take a Mernda line Sunday morning example, instead of 40 minute gaps before the 10:36am arrival at Flinders St (current timetable) the 40 minute service could finish with the 8:36am arrival with the improved 30 minute service until 10:36am. That's with one extra return trip per week scheduled. 

Even better would be one additional earlier return trip to cut the earlier 70 minute gap (before 7:56am) to 40 minutes. That's a lot of value for just two extra return trains per week. Similar Sunday morning gains are possible for at least parts of the Sunbury, Upfield, Craigieburn, Hurstbridge and Sandringham lines with a combined total of 12 return trains per week. That's about 24 trips each way, or about 100 per month. For context Metro runs about 64 000 train services per month (November 2019 figures), meaning an increase of just 0.16% in monthly services. 

That's tiny in the whole scheme of things given it would get almost our whole rail network from a maximum 40 to a maximum 30 minute wait for 18 hours a day / 7 days a week. It does not speak well for the planning, leadership and internal advocacy capabilities of DTP that Secretary Paul Younis has been unable to win government support for even these minor upgrades implemented across the rail network. 

Every 20 min
  

20 minutes is a sort of middling frequency. Its not horrid but not great either. It might be OK for non time critical trips involving no changes. But it can still greatly slow travel if making one or more connections that are not timed. Still, it is vastly better than the 30 and 40 minute headways so widespread on the network.

I have seen 20 minutes referred to as a 'check and go' frequency - in other words you are still looking at a timetable but it does not require the sort of detailed planning that you might do for a 30, 40 or 60 minute headway. 

The roll-out of much more 20 minute service was proposed for metropolitan rail lines and bus routes in 2012's Network Development Plan (Metropolitan Rail). This plan wasn't much favoured by the following government. Still, Werribee, Williamstown and Frankston train lines gained maximum 20 minute waits in the good but undersold 2021 rail timetable change. And before then quite a few bus routes gained 20 minute service. 

20 minute service is not something that people are begging politicians to implement. It's less saleable than every 10 minutes in that regard. But it's also vastly cheaper to do over multiple lines if your starting point is 30 or 40 minute service. The fine print of the diagram lets you compare the relative cost of upgrading to the next higher frequency in terms of extra trains, trams (or buses) per hour. 

As noted before, a boost from 40 to 30 minute is the same as going from 3 to 4 return train trips every 2 hours. That 0.5 extra trips per extra hour of improved service. 

The next step, of going from 30 to 20 minutes maximum waits, needs an additional return train trip per hour to be added. That's a bit dearer but is still a bargain when it is recalled that most of rail's costs are fixed. Also you've shifted the whole perception of the rail network from 'at worst bad' to 'at worst neutral' by halving maximum waits from 40 to 20 minutes.

These gains hugely aid reliability, build network robustness and cut travel time variability, especially when rail replacement buses operate. A similar 30 to 20 minute upgrade would also assist our popular SmartBus routes, which with few exceptions run their frequent service on weekdays only.

The better you make something the more people use it. Thus upgrades to this and even higher frequencies induce higher patronage, especially during off-peak times. With rail you could get a lot more 20 minute service with a small increase in monthly service.

To use round numbers, the Melbourne rail network averages about 2000 trips per day. A 1% increase is about 20 trips, or 10 return trips. Adding one return trip on 10 lines with 30 minute evening service would give an extra hour where services are every 20 rather than every 30 minutes. Go for a 3% increase and the maximum evening wait falls from 30 to 20 minutes until about 10pm network-wide. The point is that you can do a surprising amount with just a 1 to 3% increase in monthly train services, with about a 5% rise needed to finish the job.

Buses would need a bigger percentage increase as they're usually less frequent to begin with, especially on weekends. But trams would need a much smaller increase as their timetables are already every 20 minutes or better except for early and late on Sundays.  

Every 15 min

15 minute frequencies are largely confined to a few lucky train lines in the east and some weekday buses (notably premium SmartBuses). Many trams have 12 or 15 minute off-peak and weekend frequencies. 15 minutes is about what many people start to consider as turn-up and go for a single trip. It's only one extra trip per hour compared to 20 minutes and two extra trips per hour compared to 30 minutes so is fairly economical. 

While an improvement on all the previously discussed frequencies, 15 minutes is still not great for trips involving a change. This is because waiting could still amount to 30 minutes of total travel time in the worst case. An unreliable 15 minute service is also poor as the cancellation of just one train opens up a half hour gap, with similar issues possible for delayed main road buses stuck in traffic. This you might have 15 minute service as a top tier service in a small to medium sized city but a large city requires something better. 

Every 10 min

This is proper turn up and go service suitable for key suburban lines in a major city. It requires 2 extra trips per hour compared to 15 minutes and 4 extra trips per hour compared to a 30 min frequency. Thus you don't go running 10 minute service on every back street bus route. 

Still a 10 minute frequency makes the network vastly more transfer-friendly, encourages sympathetic transit-oriented land use and lifestyles and makes road time and space re-allocation (including traffic light priority and bus lanes) more justifiable. All this induces even higher patronage. And scope exists for bus network reform to deliver 10 minute frequencies on key corridors in parts of Melbourne that really need them (eg around Northland and Highpoint shopping centres).  

Widespread 10 minute frequencies were a major service aim in 2012's Network Development Plan (Metropolitan Rail). That plan also specified it for trams and top tier bus routes in its multimode coordination framework. Reference to this framework was regrettably omitted from the Victorian Bus Plan released 982 days ago today.

Nevertheless there has been some progress towards more weekday routes every 10 minutes, notably the Route 202 Melbourne University shuttle and routes 235 and 237 in Fishermans Bend (which provide a welcome precursor for a tram unlikely to be built soon). 

Other cities and expectations

The above is a Melbourne perspective. Those from large Asian, European and even some Canadian cities expect far higher frequencies than typically run in Melbourne. A Melburnian would very happily accept a 9 minute wait for a 10pm train while a Torontonian or Vancouverite would grumble. 

Even Sydney's expectations would be higher. In contrast, Brisbanites would set their sights a little lower than us, with most Canberrans, saddled with embarrassing 120 minute weekend bus headways, a lot lower.  

What do we want from our network?

Then there are more objective measures such as what you want the network to do. If it's to provide a versatile all day service for a multitude of trip types and destinations then there is going to be more interchanging (at all hours).  Passengers have a certain 'interchange time and inconvenience budget' for each trip they make. If the network imposes anything beyond that they people will either drive or not travel at all. 

Day and night high frequency is a necessary condition for the versatility the network needs to grow its role and patronage. Official understanding and responsiveness has varied over time. 

Understanding probably peaked during several decades between which the rail network was primarily regional and goods focused (late 1800s) and when it became a more extensive electrified but narrowly suburban commuter service (post 1950s). During this early to mid 20th century 'sweet spot' the trams, trains and even some buses were double their current frequency, especially at night, albeit on a smaller network than now. 

Suburbanisation, electrification extensions and core area off-peak service cuts followed. Interest in frequency narrowed to become solely as a means to maximise peak capacity rather than as a benefit in itself. The peak of this narrow commuter thinking was probably the 1970s when the City Loop was under construction (for the commuters) but evening timetables were savagely cut in 1978. Not unsurprisingly public transport reached historic patronage and mode share lows around 1981 (when Melbourne was losing prestige to Sydney and jobs were suburbanising). 

Subsequently patronage recovered (albeit in a larger city) and some service frequencies were restored. Indeed day and night services on the politically marginal Frankston line are now more frequent than they've ever been. Ideally we'd have continued with NDP Metropolitan Rail frequency upgrades in tandem with bus reform that was then accelerating. Unfortunately this basically stalled with everything going on infrastructure builds instead.

Of these the Metro Tunnel is a better project than the one-trick City Loop. Similar could be said about the Suburban Rail Loop, though one might still quibble about timing, design decisions and costs. 

The huge borrowing for these and other projects (including roads) may make it easier for government to say there's no money for service, given rising interest payments. Still, that's not what the Metro Tunnel banners say. 

Let's hope they're right and its opening heralds a renewed realisation in the power of frequency that we've waited too many decades for.  

Index to Timetable Tuesday items here

Big south-east train service upgrades turn 28

Today, 18 February, is a historic day in the history of Melbourne's train services. Old timetables were cast aside in favour of more frequent interpeak service across south-east Melbourne. 

More specifically, on this day in 1996:

* Frankston and Dandenong weekday trains improved from every 20 to 15 min (matching Ringwood and Glen Waverley), delivering what more would treat as a turn-up-and-go service. 

* Pakenham and Cranbourne line trains beyond Dandenong increased from every 40 to 30 min interpeak and from every 80 to every 40 min on Sundays. 

* Daytime Sunday services to Alamein, Belgrave, Lilydale, Dandenong and Frankston increased from every 40 to every 30 min.  

* Alamein line trains increased from every 60 to every 30 min Monday - Saturday night

This upgrade package constituted the most decisive increase in metropolitan train services for many years, if not decades. They certainly trump the early 1990s Sandringham line boost (also every 20 to every 15 min off-peak) and the peak gains the Frankston line got when its third track was built. 

It happened just before the March 1996 state election (at which the Kennett government was returned). A 15 minute frequency also raised more of the Melbourne's rail network to the service levels seen on Perth's newly electrified system which adopted 15 minutes as a base frequency (initially weekday, eventually 7 days). 

Rail frequencies set a sort of base service level for buses, especially in the middle suburbs. Before the rail frequencies were improved, the main bus routes ran every 20 minutes. The upgrade to trains every 15 minutes did not result in immediate gains for buses. However later some of the major routes in the south-east became SmartBuses (eg 703, 888/889, 700 and then later 900, 901, 902 and 903) which featured 15 minute weekday interpeak frequencies to retain a degree of headway harmonisation with trains. And even where buses were not upgraded, the boost in train services to every 15 minutes reduced average waits for those changing from buses.  

What about the West and North? They do not celebrate today. The Kennett government, especially in its first term, was south and east centric, reflecting its voter base. While the latter gained from the abovementioned service upgrades, they started a geographic inequality in train frequencies that has only widened since. 

Leaving aside the politics (with the north and west rusted on Labor) it may have been possible to justify the priority treatment for the east and south-east on patronage grounds. This is because Melbourne was more lopsided then with the eastern lines being busier. That's since changed with any previously valid usage excuses for the north and west not to get more train service evaporating.

There ought to have been a subsequent service catch-up to give the west equality with the east (like implementation of 2012's Network Development Plan - Metropolitan Rail would have done). However there hasn't been with the Frankston line 'haves' now enjoying half the waits of the Craigieburn and Mernda line 'have nots' at most times. Hence the distribution of rail frequency today increasingly reflects 'safe seat' politics more than objective measures like patronage or social needs. 

Returning to the 1990s, the other major Kennett era service boost happened before the 1999 state election in which metropolitan train and tram lines had Sunday services boosted between about 10am and 7pm to match their more frequent Saturday timetables. This was a major gain with benefits spread across Melbourne (unlike the 1996 upgrades).

Sunday evening services remained at every 40 minutes until a couple of years later when the Bayside portion got improved to every 30 min with the Hillside portion (under Connex) following suit a little later. Pre-10am Sunday morning services levels remain unchanged to this day except for the addition of earlier trips under Night Network and Werribee, Williamstown and Frankston upgrades in early 2021.  

These 1996 and 1999 train service increases are historically important as they represent a decisive turnaround from the cuts during the 1950s, 60s and 70s. Sure these might have cut costs but they also reduced patronage and contributed to increased car traffic (which carries its own cost in road construction and maintenance). Hence cutting service is counterproductive and doesn't necessarily save much money.

Better frequency on the other hand is beneficial as it reduces waits, making train travel more useful for diverse trips throughout the day, not just during the commuter peaks. And of the times where you could boost service, off-peak frequency is most cost-effective as the marginal cost of increasing it is low and it optimises fleet usage. Interestingly this was more realised and acted on in the late 1990s than in the 2020s (so far).  

The future

The current state government has found it easier to borrow tens of billions for new infrastructure than to find more modest (but recurring) amounts to make service more frequent and thus useful. Over the last decade or so this has been notable for all modes of transport except V/Line trains (which have enjoyed both large infrastructure and service gains).

Hence it's a common sight in the suburbs for there to be shiny new stations served by trains as far apart as 30 or even 40 minutes at important times people travel. Melbourne's 'infrastructure first' approach is quite different to 'service first' planning practice elsewhere, such as Sydney, that have done much to reduce maximum waits to much lower amounts (like 15 or even 10 minutes) over a much wider span of the day (while simultaneously pursuing a large infrastructure program). 

Will Melbourne's approach change any time soon? Will service finally get its day in the sun? Large Metro Tunnel banners raise hopes though without specifics it remains to be seen how widely spread the benefits will be. 

Further reading

https://melbourneontransit.blogspot.com/2011/05/melbourne-interpeak-train-service.html

https://melbourneontransit.blogspot.com/2011/05/melbourne-evening-train-service-levels.html

https://www.ptua.org.au/files/2008/2023/12/PTUA-News-Volume-23-No.-2-June-1999.pdf

UN 169: Do busy lines get the best service?

One of the reasons you hear people give for suburban trains being not as frequent as they could be are level crossings. That might apply during peak times on busy lines but not really outside them. Ditto for signalling capacity, rolling stock or the Metro Tunnel not being open yet. People who should know better sometimes peddle myths around train frequency and how difficult it is to have all day.

That needlessly lets service sceptical state governments off the hook and makes trains less useful for diverse trips than they should be. This has been a particular problem in Melbourne whose evening and Sunday morning trains are now only half as frequent as Sydney's (whose state government has been much stronger on service matters than ours has). 

Low usage is another reason sometimes given for not running much service. Although it's a chicken and egg situation as poor service can induce low usage as much as frequent service encourages patronage. If you've got reasonable population density around stations (or good buses running to them) then that declines as an excuse too.  

Fortunately station boarding numbers exist and we can settle the question. Are busier sections of the network better serviced than quieter sections? Or are service levels more due to history and politics than catchment density, patronage and community need? It would be particularly extraordinary, and a failure of responsive planning, if a popular line also had low service with timetables basically unchanged for decades. You'll find out whether that is the case in a moment.

Gauging service by maximum waits

My interactive 10pm Sunday frequent network map has a layer where you can see stations colour coded by maximum wait between approximately 7am and midnight (on any day of the week).

I changed the presentation of that to show maximum waits by line section based on major junction stations in the suburbs. Because the central area (inside North Melbourne, Jolimont, Burnley, Richmond and Footscray) has many lines I didn't colour code frequency there. I then used per station 2022-23 boardings data obtained from DTP to add up boardings per section and plot them for each section. 

The result (Metro services only) is below: 

In most cases the longest waits happen on Sunday mornings, often closely followed by evenings. Pakenham had 60 min gaps while all northern suburb lines (and Sandringham) had intervals up to 40 min between trains.

30 minute intervals were common in the inland east, including busy junctions like Camberwell, Ringwood and Dandenong. These were also encountered not only on Sunday morning and evenings (particularly weekends) but also weekday interpeak at stations east of Ringwood.

Lines with the shortest maximum waits (20 min) were those mostly hugging the bay including Frankston, Werribee and Williamstown. These had 30 or even 40 minute gaps but a timetable improvement in early 2021 reduced maximum waits to 20 min. Sandringham almost qualifies but didn't due to its infrequent Sunday morning timetable. 

The maximum wait approach I've taken is not perfect. For example it doesn't differentiate between lines that have frequent service most of the time and those that don't. For example it rates Werribee higher than Dandenong despite the latter's 10 minute  midday service all week. But when you look at after 7pm weekend evening and Sunday morning service lines like Werribee take the lead with 20 minute frequencies versus 30 minutes for Dandenong. 

I make no apology for this since having short to medium maximum waits over say a 6am to midnight span is a necessary condition for a railway to be useful for diverse trip types. And if a line already runs frequently or semi-frequently 7 days then the extra annual service kilometres needed to cut maximum waits from 30 to 15 - 20 minutes at the ends of the day is trivial. So it's purely due to inertia that we don't have it. 

A move from a 'peak heavy' to 'all day frequent service' pattern is more aligned to modern working patterns, especially post-pandemic. A higher patronage to service elasticity assists the cost recovery ratio. Frequent all day service is also what differentiates a good system from a poor system and can drive more efficient land use and housing policies as well as widening households' choice regarding car ownership, with significant cost of living benefits. 

Patronage versus maximum waits

What about the patronage? I worked from the outside in, although the stars are at junction stations. For instance the 4.3 million is Werribee - Seaholme, 2.4 million is Newport - Seddon and 4.2 million is  just Footscray and South Kensington (actually mostly Footscray). Not surprisingly the inner area had the highest at 53 million (mostly comprising suburban residents returning home). 

What sections have similar patronage levels but very different services? Most notable is the Craigieburn versus Frankston lines. Craigieburn has higher patronage than the Frankston line yet gets only half the service. This applies not only when waits are maximum (every 40 vs every 20 min on Sunday morning) but other times eg midday (every 20 vs every 10 min 7 days). Evenings also have a gap with Craigieburn's 30 versus Frankston's 20 minute frequency. Frankston is a famously politically marginal line while Craigieburn is populated by Labor loyalists who (in public transport services) have not had their political faith repaid by successive Labor governments. 

The above north-south disparity reflects 2024 service levels but there wasn't always a difference. Thirty years ago the Frankston line had a similar off-peak service level as lines in the north. However there has been a long-term trend to boost Frankston line service while timetables in Melbourne's north stagnated with the disparity growing over time.

Hence Frankston line passengers now enjoy a 7 day turn up and go service (with a maximum 20 minute wait at night and on Sunday mornings) versus 20 to 40 minute frequencies on northern lines. To be fair the latter got extensions into growth areas (Watergardens, Craigieburn, South Morang, Sunbury, Mernda) however basic off-peak frequencies were never improved despite strong patronage.  

Also in the north-west/north are the Sunbury and Mernda lines. These have similar patronage (around 6.5 million) and, like Craigieburn, are in the 40 minute maximum wait club on Sunday mornings. Sunbury should get improvements when Metro Tunnel starts though nothing conclusive has yet been published since the 2016 business case. Each line's annual usage closely matches Ringwood's usage (6.4 million) which gets a 30 minute maximum wait. The difference between these lines is highest on weekends thanks to Ringwood's 10 minute midday service versus 20 minutes on the other lines.   

Pakenham, with its 2.6 million annual boardings gets a particularly raw deal on the maximum wait scale due to it's terrible Sunday morning timetable with hourly gaps. It's a big contrast with nearby Cranbourne (which has barely half the usage) and an even bigger contrast with Williamstown (with 1/5 the usage). It does however enjoy a better weekday interpeak frequency (20 min) compared to the also short-changed Belgrave and Lilydale sections east of Ringwood (every 30 min). Stations in this ara like Mooroolbark and Bayswater (recently rebuilt with level crossings removed) are demographically similar to the likes of Carrum and Frankston yet get only one-third the off-peak frequency (30 vs 10 min). 

Conclusion

It's almost all politics! Not 'objective' criteria like population density, usage and social needs. 

The maximum waits you have for trains depends on who your neighbours vote for and especially how politically loyal they are. Service differences between some lines have gone from negligible 30 years ago to about 2:1 today, despite major sections of the less served line attracting higher usage (notwithstanding the inferior service). 

The record shows significant policy difference between the main parties when in government. Labor has had a stronger public transport infrastructure agenda than Coalition governments. But on service, especially metropolitan rail service, if anything it's been the Coalition that has been stronger.

For example past Coalition governments (even Jeff Kennett's) greatly boosted Metro train services (mostly in their east-of-Yarra safe or marginal seats) while Labor governments returned few similar favours for their loyalists in the north.

The result is thus a widening have/have-not gap in the geographical distribution of 7 day frequent train service so necessary for Melbourne to function as a prosperous, liveable and equitable city. 

Another look at at patronage data is provided by Daniel Bowen here. 

Index to Useful Network items here