Friday, August 13, 2021

UN 101: Suburban connection points - the secret to useful & versatile trains?

How do you analyse a transport network? The first thing most people look at is coverage. If you don't have coverage you don't have a network. Most transport authorities produce network maps that show where routes go. Like PTV/Department of Transport

Second you can look at frequency. Unless you have carefully timed connections (and even these break down if you're travelling in the counter-peak direction) you need frequent service to assure easy connectivity unless you're making one of the rare trips that does not involve a change.

Some cities produce frequency-based network maps that highlight the parts of the network that support easy interchange. We don't officially do that in Melbourne (though I do via interactive frequent network maps here). Hence on PTV/DoT maps a bus that's a few times a day gets the same (or occasionally more) prominence as our top tier routes every 10 or 15 minutes. 

Most peoples' mental image of the public transport network, which the Department of Transport does nothing to dispel, are to the effect that trams are almost always frequent, trains sometimes are while buses never are. The latter are also full of traps for the unwary, like missing trips on public holidays or short operating hours even on direct and popular routes.

Because, with few exceptions, trains and trams are radial, while buses are not, there are only a few non-CBD points where people can trust frequent services to intersect and form a versatile network in multiple directions. 

Which brings us to network topology. That's another measure to which I have hitherto given too little attention. You could have good coverage and good frequency but if routes all run roughly parallel, only meeting near the CBD (like on the map above) then you don't have a true connective network. What's there requires significant backtracking.  Which is both inefficient and time-wasting, with the problem worsening with distance from the CBD. 

In a city where most jobs and destinations are not in the CBD a network that's grid or spider web shaped is more useful than one where everything runs to the CBD. Background on this here and here. The late Paul Mees referred to such as connected network as 'Squaresville'. 

In short you need routes that: 
(i) meet each other in a rough grid (right angle is geometrically better than dramatically acute or obtuse angles), 
(ii) are frequent enough to cater for efficient interchange in all directions, and 
(iii) are at least moderately fast on their own rights of way or at least with reasonable priority. 

You can get away with just (i) and (ii) but end-to-end travel will still be slower than driving, with the difference increasing for trips longer than about 5 or 10 km or with a significant non-rail component. Thus if we're seriously interested in efficient metropolitan-wide transit you need all three. 

Very simply that narrows down the network to include just trains, trams on their own right of way and fast busway buses. This ends up being a much coarser network than one that includes all the infrequent buses and routes that run in-traffic. In fact in a city like Melbourne you might be left with the rail network, some sections of trams and very few buses. Even much of the rail network disappears if you apply even a very ordinary frequency criteria (eg every 15 minutes off-peak) that limits waiting and connection times. 


To keep things simple, we'll just look at heavy rail networks, bearing in mind that some cities will have frequent light rail and buses on their own right of way that should also be included as portions of the 'congestion free' transit network. I'll exclude CBD and near CBD connection points and only discuss non-CBD junctions that could be useful for non-radial trips. 

Melbourne (now) 

The current network has just three suburban connection points since I've considered the likes of North Melbourne, Richmond and South Yarra as part of the CBD network core. All are in the inner east and south-eastern suburbs since lines in the north and west are too infrequent to qualify (mostly every 20 minutes). The same applies for junctions like Ringwood (feeder lines every 30 minutes) and Dandenong (feeder lines every 20 minutes). The three shown are Caulfield, Burnley and Camberwell.

All have limitations set by, amongst other things, network geometry. For example while both the Frankston and Dandenong lines have good frequencies (every 10 min interpeak) the angle between them is rather acute at something like 45 degrees. This means that the ratio of the distance that needs to be travelled and the distance of the trip for a journey like Clayton to Cheltenham is high. This is unlike where if the lines were angled at 90 degrees and Cheltenham was the same distance south as Clayton was east then the ratio would be only 1.4 with this more likely to be cancelled out by the train's faster speed. 

The Glen Waverley and Ringwood line have an even more acute angle at Burnley. Contrary to the map they run close to parallel with trams and buses usually being the better way of getting between them. The main penalty is the waiting, especially for trips involving two changes as buses, with widespread half-hourly frequencies, don't assure reliable connections.  

The geometry at Camberwell between the eastbound Ringwood line and the roughly southbound Alamein line is better. However the Alamein line doesn't go anywhere significant. Even the tram that intersects it doesn't have a particularly good physical connection due to the location of stations. Hence this connection is largely used by locals in the Burwood/Ashburton area getting home rather than also by cross-metropolitan travellers going to potential Alamein line destinations (if extended) like Caulfield, Chadstone and Oakleigh. Hence not only should there be connection points but the line you're connecting to should go somewhere useful, preferably also including another connecting point. 


Sydney's network is vastly different. Firstly its basic off-peak train frequency is 15 minutes which means less waiting than for a large part of Melbourne's network. Secondly it has many more suburban rail junctions between frequent lines. Hence it has about three times the number of suburban rail connection points than Melbourne with better frequency and geometry. An example of the latter is in cases where lines intersect at about right angles and go on to feed other lines. Every time Sydney extends its rail network it appears to add more middle suburban connection points, further increasing the network's versatility for diverse trips (contributing to the 'network effect'). 

Other Australian capitals

The smaller capitals have the Melbourne (and Chicago) problem of a very radial network. Brisbane and Adelaide also have low frequencies (every 30 minutes) on most of their lines. Brisbane has suburban branches but they are like Ringwood with 30 minute frequencies making connections chancy (I haven't checked whether timetables are timed). 

Perth also has a radial network but with 15 minute basic frequencies on all lines. It has one suburban junction at Cannington where passengers can change to the Thornlie line (also every 15 minutes) to go one station. All other interchange points are in the CBD. However projects currently underway will see this increase to three suburban junctions (like Melbourne has now). One will be the Mandurah line's Cockburn Central (at which the extended Thornlie line will terminate) while the other will be  the Midland line's Bayswater, to be served by the under-construction Airport/Forrestfield line and after that a line to Ellenbrook. Even better is that the radial Thornlie line will become part circumferential, permitting an entirely rail trip between the radial Mandurah and Armadale lines. Suburban rail - rail trips like this are possible in Sydney but on no other Australian suburban system without a CBD change.  

More connection points for Melbourne

It should be quite clear that in terms of the versatility of its rail network, Melbourne is severely handicapped relative to Sydney. This is due to two factors: frequency and network geometry. The first is relatively easy to fix while the other requires the construction of cross-radial lines. 

Increasing to a 10 minute frequency on main lines (as proposed in the 2012  Network Development Plan - Metropolitan Rail ) would create new connection points at Clifton Hill and Footscray. This would require an upgrade from 20 to 10 minutes interpeak for Mernda and at least the inner portions of the Hurstbridge and Sunbury lines. Sunshine could be added to that list if short trips were added so that trains ran every 10 minutes to Wyndham Vale. The gain would be even greater if CBD connection possibilities were counted with frequency improvements for other lines. Craigieburn is the one not mentioned so far with the likely highest priority. A frequency upgrade would have the added benefit of better airport connections via the 901 bus at Broadmeadows. Getting Ringwood to every 10 min interpeak would be another top priority due to its very low cost, the number of stations that would benefit and the marginal seat electoral benefits for whichever party does it. 

Then there's the construction of cross-radial lines, most famously the proposed Suburban Rail Loop. That could give us more suburban connection points than Sydney has now. However this requires not only for the SRL to be built but also frequency upgrades on many radial lines. The record of recent governments is that building infrastructure, for all its complexity, can come quicker than adding service frequency, even where we already have sufficient train sets.


The lack of rail-rail connection points, especially non-CBD, is where Australian urban rail networks fall down compared to many systems overseas including in cities that Melburnians travelled to (pre-COVID). This is largely because of how the networks evolved. This website has subway maps from around the world. Here's some examples. 

Toronto's subway network has fewer lines and less track kilometres than the rail networks in Australian cities. But it punches above its weight with regards to patronage. Like Melbourne it has three main connection points (actually four but they involve the same lines) that are likely to be busier than our three given their vastly higher frequency.


Busy and useful rail networks that serve diverse trips do so by permitting connections at multiple points. Despite the relatively large size of our suburban network it has few suburban locations where people can easily change. This is due to (a) low service frequencies that make changing chancy without checking timetables and (b) a network geometry based on radial lines. Where there are junctions they are typically line branches that form acute angles pointing inward. And there are no circumferential lines. This works for CBD travel but is less effective for cross-suburban trips. 

Consequently our network scores quite poorly when analysed with tools like SNAMUTS that value connectivity. The abstract diagram below shows how the more junctions you add greater the number of trips that become convenient. And it's an exponential rather than linear relationship. Because of this, well-targeted spending on rail connectivity can be highly cost-effective. 

A network development program to make our train network useful for more trips (especially post-COVID where peak CBD travel may assume lesser importance) needs the following components: 

1. Increase frequencies on existing lines to permit connections between them without a timetable. Smaller cities like Perth accept this as being 15 minutes but a city like Melbourne should probably aim for 10 minutes, given that many of our lines have outer branches or single track sections that suit a 20 minute headway (30 min not really being good enough). This approach matches that of the Network Development Plan - Metropolitan Rail written nearly 10 years ago and the coordination framework mentioned here. Top priority lines from a connectivity viewpoint include Sunbury (to Watergardens), Craigieburn, Werribee,  Ringwood, Mernda, Hurstbridge (to Macleod), Glen Waverley, Upfield and Sandringham, probably in something like that order. 

2. Optimise connections between trains and key feeder tram and bus routes. This is a lot of small projects and service upgrades including bus network reform, adding service kilometres, harmonising service frequencies, longer operating hours and improving physical interchange between modes with emphasis on access and shelter. Recent discussion on this here and here. This is unspectacular painstaking reform at a lot of sites but it can really make the network connect on a fine grained level that often gets ignored. 

3. Faster buses and trams. The first two are super-important but by themselves won't deliver really fast journey times. SmartBuses look good on paper but in practice are often too slow and not frequent enough. Until the SRL gets built Melbourne has to do with trams and (especially) buses what other cities have the luxury of doing with rail (because their built their system as a connected Metro rather than radial country lines that got electric suburban service added later like we did). This means really good full-time tram priority (not just when behind schedule) and 'bus wormholes' in the suburbs to connect rail to key destinations via fast frequent buses uninterrupted by traffic. You might also consider tram extensions where lines stop just short of stations or there are major 'missing links' (eg CBD fringe or to connect key hubs such as Caulfield to other lines). 

4. Big infrastructure like the Suburban Rail Loop and similar schemes. This is the final thing. It should only really be built after we've done all we can with the first three (which are highly cost-effective and often low-cost). The SRL is in contrast a dearer longer term project, with a price tag of billions of dollars per connection point added. But given we've committed to it the design needs to allow excellent train - train connections from the radial lines to the SRL stations. This means very short walks, a minimum of climbing, complete shelter and freedom from having to cross roads. I discuss world class interchanges here. But if done properly concepts like the SRL could be transformative in making our network much more versatile.  


Brett said...

Sydney also has a train line that is completely suburban (the T5 Cumberland Line) which no other Australian city has yet and won't for many years p

Peter Parker said...

You don't count our Stony Point line? Not all suburban but not CBD either.

Unknown said...

How wonderful it would have been if they kept the Outer and inner circle lines as well as the Roseland (I think) lines.