UN 138: Why reforming buses is like playing with string

Reforming bus networks is like unravelling an (often tangled) ball of string. You can cut it in any number of places but cannot vary its total length. One of the ground rules is that it must still go to a lot of places without too large gaps in between. Another rule is that it can't be spread too thinly. 

You could get crafty and split the string into fine strands to get more length but the result will be fragile and may break. Or you might double or triple up parts of the string for a stronger result. That's also got problems as you're then compromising on the distance it can extend to and/or the size of the gaps. 

Bus network planning is essentially a string running exercise. The service kilometres you can run each year is your ball of string's length; typically a non-negotiable budget. The frequency is the thickness of the string. You are not usually allowed to go to worse than one bus per hour so we'll use this as a base standard and not permit splitting. Paralleling strings for higher frequency remains allowed but you are going to have to shorten or more widely space strings to avoid running out.

There's another nasty as well. These are special strings that need renewal throughout the day. So to keep them up for 24 hours costs more than 12 hours. But there's an extra fixed cost involved in erection and dismantling so even if a string is only up for 3 hours a day it can work out quite expensive. So you don't want too much climbing up and down ladders twice a day. Hence there's much to be said for leaving string up all day unless there is truly no one around to appreciate it. In case you haven't twigged, the bus parallel with having the string up for only a few hours a day is running a frequent service during am and pm peaks with buses away in depots the rest of the time.   

The Department of Transport's job as network planner is to run the string available in the 'best' way possible. There is no one single definition of this. Every service planning decision favours some areas, some people and some types of trips over others. There is also a general wish to avoid too much disruption for existing passengers or to require people to walk too far. Again there are varying definitions of 'too far'; some people will walk 1000 metres while others will struggle with 200. It can depend on the service quality; if it is fast and frequent then willingness to walk more may increase. 

DoT wants to know the public's view on this before it starts to reform bus networks in areas like northern Melbourne, north-east Melbourne and Mildura. Hence its surveys are heavy with questions asking about people's willingness to walk 400, 800 or even 1200 metres for a service and under what conditions.

Assuming the same amount of string, you can have (roughly) double the frequency if you have double the route spacing and thus doubled maximum walks. Having a bus every 20 instead of every 40 minutes permits better travel time flexibility and shorter average waits for connections. The cost? Some people may have to walk 5 minutes more (which may or may not be acceptable). 10 minute frequency is even better with many people considering it as 'turn up and go' service. 

I will present 6 hypothetical bus networks. All use the same amount of string (ie bus service kilometres per year resources) so can be directly compared. 

Network A is an extremely fine meshed network with a grid of 16 bus routes 400 metres apart. No matter the location you are within 200 metres of two routes. Thus this network is very good for coverage. Unfortunately each route runs only hourly. So interchanging to get anywhere not on your two routes will be difficult unless you are very lucky. And even if a destination was a single seat ride away the hourly timetable is unlikely to be convenient.

Despite its good coverage this is a 'charity' type network whose use will be largely confined to those without cars because its frequency is so poor. Buses will run mostly empty and fares revenue will only contribute to a small percentage of operating cost. Adding extra service will be expensive and not necessarily attract that many more passengers. 

2006-era bus planning practice had a tough coverage requirement (eg 90% of people within 400 metres of a service) but a weak minimum frequency requirement (every hour or better seven days). This, combined with bus-hostile street layouts in many 1960s-1990s suburbs, delivered a network comprising many infrequent and indirect local routes but few if any frequent main road routes (SmartBuses are scarce, largely confined to the east and have had no additions since 2010). 

This outcome is like Network A, though with less direct routes and a tendency of them to converge on key destinations rather than stick to a grid. Typical examples are in Melton (where nothing runs better than every 30 min, even in peak) and large parts of the outer north, east and south-east (eg Clyde, where there's also many overlaps between routes every 40-60 min and weak termini). 

Network B below uses exactly the same bus resources as Network A (count the lines if in doubt - each line represents one bus per hour). The difference is that the 16 hourly routes are cut to just four wider spaced routes running every 15 minutes. That higher frequency gives more flexibility of travel times. And if your trip requires a change the most you will wait is 15 minutes. 

The trade-off is the increased walking distance; instead of 200 metres it becomes 800 metres.  Still, this simple frequent network is good for some trips rather than Network A which is poor for all trips. So it will probably get higher usage than Network A, including from those with an option to drive. Concentrating routes on a few corridors may permit enhancements such as bus priority and good shelters at all stops that should further boost usage. Either increased fare revenue or crowding might  strengthen the case for improved frequency, eg to every 10 minutes, and create a virtuous circle of usage and improvement. A network based on such a frequent grid has been proposed for Melbourne's west and was reviewed here. Unfortunately its wide route spacing gives rise to a risky dependence on flexible routes that are inherently inefficient and don't scale up well for moving more than small numbers of people.  

A network like B won't be universally acclaimed. The extra walking distances won't be welcomed by existing passengers, especially if walking conditions are poor due to neighbourhood severance caused by large roads and roundabouts. The latter can also cause problems at intersections where the network forces passengers to change to complete their trip. And some passengers are physically unable to walk long distances. Hence Network B is the coarse and frequent opposite to the fine but infrequent Network A.   

The nearest Melbourne examples of a network like this are in areas like Mt Waverley where buses don't always penetrate areas between widely spaced main roads. The routes involved are amongst the most productive on the network. However frequencies are likely to be only every 30 minutes rather than the desirable every 10 - 15 minutes so the frequency gains for the extra walk aren't always there.  

Network B was predicated on 800 metre walks, though this assumes a road network with good permeability into local streets and ample crossing opportunities without backtracking. Remove any of that and walks might be nearer 1200 metres.

Interestingly DoT asked about willingness to walk 1200 metres in its bus reform survey. My guess is that it was put in as an extreme to test the limits and make the 800 metre option (which they'd want most people to tick) not look outlandish. It's a bit like you're in an appliance shop and there is a choice of three different models. While the third most expensive unit might not sell nearly as much as the middle-priced model it is still produced as it makes the middle priced item look more reasonable and neither 'cheap' nor 'expensive'. 

Is there a middle ground between the extremes of Networks A and B? There are several possibilities. Take a look at Network C below. The 800 metres between routes means that no one is more than 400 metres from a route, similar to accepted coverage standards now. This is done by moving one trip per hour from the main corridors to a roughly parallel street. This creates a two-tier network where people have a choice of walking up to 800 metres to a more frequent corridor or 400 metres to a basic hourly service.

This network is far less austere than Network B so is a much easier sell to a community used to good coverage. While there is a frequency trade-off on the main routes (from 15 to 20 minutes) the latter meshes better with trains in many areas (especially if more lines move from a 15/30 to a 10/20 min frequency pattern). 

The two-tier Network C has similarities with the 2014 and 2015 revamped networks in Brimbank and Wyndham. The main exceptions is that the 'coverage' routes are typically every 40 rather than every 60 min (at least on weekdays) and service on some main road 'connector' routes often fall short at every 40 rather than 20 minutes off-peak. These are good patronage performing routes that justify an upgrade and it was only due to a lack of resources that has kept their frequencies (and operating hours) less than desirable.  

If you wanted to go even further along the coverage road, like how we went from Network B to Network C, then you might end up with Network D (below). Like Network A it has maximum walks of 200 metres. However the gaps are rectangular rather than square. This allows four routes to be doubled-up with a 30 minute frequency. However this is unattractive, especially in areas where trains run every 20 minutes. Hence I do not recommend this network style unless short walking distances are of utmost importance. Even then this network will have most of the problems of Network A with mostly unattractive service. 

If we wanted a 20 minute grid like Network C but still valued some infill coverage, an alternative two-tier option is Network E. It has only five routes with four being direct routes along main roads. This fifth route is a local service. It's indirect so it is about twice as long as the other routes. However as it replaces four coverage routes it can run twice as frequently with the two lines denoting a 30 minute service. As well as providing coverage this route might serve a major station or shopping centre, providing a local one-seat ride to areas that would otherwise not have it. 

Street layouts in many 1960s-90s subdivisions may force a bus network that looks more like E than C. This is because coarse main road grids lack finer grids within (though some newer subdivisions are better here).

An example is Hoppers Crossing (below) where routes like 161 and 181 provide within block coverage while also connecting people to major local attractions like Werribee Plaza that direct routes like 160 and 180 do not. No one would take the 161 end-to-end but it does perform a useful coverage role and is a decent patronage performer. It also supplements direct routes like 160, 170 and 180 whose patronage productivity is extremely high.

The Wyndham experience demonstrates that an efficient network need not all be straight routes; you may be better off to have several simple straight routes and one or two very indirect 'mop up' routes rather than have them all somewhat indirect in an effort to retain coverage. Having a mop-up coverage route also allows you to respond to customer feedback for deviations to retirement homes etc without compromising the basic simplicity of the direct route network. 

So far we've only discussed a two tier network. In DoT parlance these are 'Connector' and 'Local' routes. The bus reviews also have 'Rapid' routes for the top level. Eg the SmartBus type services that we have none of in the outer north, west and south-east. Rapid routes would basically be like Connector routes but with turn-up-and-go frequency and innovations such as bus priority, 'rapid running', off-vehicle validation and, in some cases, exclusive lanes.

How can you accommodate a frequent rapid route without needing more string? Of the above network options, Network E is most amenable to change. Supposing the windy route (which is twice as long as the straight routes) was cut from every 30 to every 60 min? That would allow one of the connector routes to go from three to five buses per hour (ie a 12 minute frequency). Now suppose that you added some priority and/or 'rapid running' to remove mid-route dwell time with a headway-based timetable. The extra speed gained might get you up to 6 buses per hour, ie a 10 minute service with the same number of buses. This is why the three tier Network F below has the equivalent of 17 rather than 16 lines.  

What style of network will reform produce?

There is broad agreement that Melbourne's more frequent and direct network is underdeveloped everywhere except perhaps the City of Manningham (which has 7 SmartBus routes versus zero in municipalities like Wyndham, Melton, Casey and Cardinia). Having a majority of survey respondents tick either 800 or 1200 metres would embolden the DoT to pursue cost-effective network rationalisation that could add more frequent routes. It is clear that DoT is encouraging people to give answers along these lines which would be in accord with successful bus network reform in places like Sydney and Auckland. 

Networks A and B should be ruled out as impractical extremes. They are either infrequent or have unacceptable walking distances. Adelaide tried the latter in a 'big bang' approach to bus reform and failed. Network D is insufficient radical with 30 minute frequencies being neither attractive nor harmonised with trains in most areas. That leaves either Networks C, E or F as being acceptable options. 

C most suits areas where there is a strong fine-grained road grid such as Melbourne's inner north, east and south-east and even some newer subdivisions such as Tarneit. E, the approach implemented in parts of Wyndham, could be another acceptable reconciliation of frequency and directness with coverage. Both are also consistent with the Useful Network mini-reviews covered here with a 20 minute frequency criterion adopted for 'useful' (or 'connector') routes matching SNAMUTS standards for network analysis. Because bus routes in Wyndham are very productive there's a good justification for more resources, including higher 7 day frequencies including potentially some running every 10 minutes. This could result in a network more like F, though if the local routes are decently productive (which they are in Wyndham) you'd retain or improve their service rather than reduce it like I did in the exercise here. 

I've focused on coverage and walking distances but bus reviews need to examine other criteria such as access to useful local destinations. There is some interaction because unless you have limitless money you cannot have a frequent bus from everywhere to everywhere. You are going to have to accept some level of interchange. This is made easier if routes are more frequent and this is made easier if the number route kilometres (as opposed to service kilometres) is kept to near the minimum needed to assure reasonable coverage. 

DoT's surveys has also tested peoples willingness to change services to complete a trip. The current network often offers many destination choices but buses may only be every 30 to 60 minutes, especially on weekends. People saying that they are unwilling to change may limit the radicalness of reform proposals, just as would be the case if they all ticked 400 metres in the walking distance question. This is because DoT must work with the ball of string it's been given, with extra balls a rarity (unlike the more generously funded infrastructure projects). 

Like with walking distances, a greater acceptance of changing is essentially a vote for network reform. It could lead to fewer but more direct and more frequent routes (since string is freed up to boost frequencies). 

Current interchange facilities, particularly for bus-bus changes at busy suburban road intersections, are both rudimentary and passenger-hostile, with stops too far back. Features needed to improve interchange including information, service frequency, shelter and accessibility are asked about in the survey. This is important because in some cases a reformed bus network will also need road works, including roundabout removals and stops being brought nearer intersections, to make interchanging practical.

See other Building Melbourne's Useful Network items here