In which I try and think more about bus-bike delays, by creating a toy model looking at interactions in bus lanes.
One thing that frustrates me, and others, is how current appraisal processes implicitly privilege a very problematic status quo. This comes up in safety audits, for example, where approaches long used in various Continental countries are rejected as risky, while the way we’ve always done things is seen as default – despite inherent risks, most notoriously the left hook risk built into our deadly combination of nearside cycle lanes, ASLs, and unprotected multi-lane junctions.
Anyway, I didn’t want to talk here about junction safety. The more general point is the costs of the current situation often appear hidden, even though they are in some other respects obvious. What’s the relation here to bikes and buses? Well, some of the debates has focused around trade-offs. If we improve things for bikes, buses will suffer. Cue arguments.
In one respect this isn’t wrong. The whole transport system is always going to be a mess of trade-offs, given the elasticity of demand and the limits of space. However, too often we don’t appreciate, study or model the problems built into our current system, in detail. And although we often have great high-level objectives (Pedestrians first! Then cyclists! Then public transport! Let’s be healthy! Sustainability! etc. etc.) they don’t seem to translate into scheme design, when we need dramatic change. The detailed assessments too often only see the apparent down sides of change, and not the down sides of leaving things as they are. It’s built into Cost-Benefit Analysis. Who’s worse off? Who’s better off? (While the current system will continue to ‘invisibly’ discriminate against many users, for example, those cyclists who cannot simply ‘dismount’).
So, back to buses and bikes. I have been suspecting for a while that there might be time disbenefits associated with current approaches to provision, and that these are not taken into account. So a while back, I asked around some lists to find out whether people had actually looked at the extent to which buses and bikes might delay each other, when mixed. To my surprise (and I’ll be happy if I’m now corrected!) I found very little.
The way that we’ve traditionally modelled traffic interactions – and the marginalisation of bicycle traffic – seems to have obscured the potential for understanding bus and bike interactions, from a time / delay perspective. (See this CIHT pdf download for a useful outline of different traffic modelling approaches). Traditionally, the modelling of traffic behaviour was done at an aggregate level, with vehicles represented as streams, using Passenger Car Units to represent the amount of road space taken up. From this perspective, mixing buses and bikes doesn’t appear so problematic. Bikes – particularly at the low levels prevalent in much of the UK – don’t take up much space, and when one factors in bus stopping, bikes and buses have relatively similar average speeds. This was the comment made to me by a colleague who specialises in buses: the speeds are similar, so mixing doesn’t cause delays.
This, it struck me, as someone who on a bike is sometimes stuck behind a bus and vice versa, doesn’t sound right. It may look ok in equations, but it doesn’t actually allow us to look at what happens when buses are stopped, or when they catch up with slower moving bicycles. A cynical thought also came into my head. Those ‘Cyclists Stay Back‘ stickers, still now festooning the TfL bus fleet. Could by any chance the hope be that riders will read ‘Cyclists Stay Back’ when approaching a stopped bus, and think “hmm, I’d better wait and let the bus move off”?
Anyhow, the good news is that microsimulation techniques allow us to look at bus-bike interactions in detail, and to explore the time impacts on both modes of, for example, (successfully!) persuading some cyclists to wait behind buses (or vice versa), and of mixing versus separation. With microsimulation, you can have heterogeneity; you can have cyclists behaving differently when there’s more of them (known as ‘platooning’), or you can have bus drivers acting differently in different contexts. This gives us the ability to explore the level of bus and bike delays in different situations, not just equate them out of existence.
The bad news is that there is still very little published work on this. Eventually I did find a paper with the initially unpromising title ‘Quantifying the total cost of infrastructure to enable environmentally preferable decisions: the case of urban roadway design‘. The authors focused on what happened in contexts where motor vehicles could or could not overtake bicycles; they were particularly interested in larger vehicles as these are the ones most often restricted from overtaking by lane width. The key point of this paper for my purposes here is:
‘We find that increasing bicycle mode share can have a significant impact on motor vehicle delay, and indeed greatly increase total costs where sufficient right of way is not provided. Conversely, with sufficient space to allow wide vehicles to pass bicycles, a reduction in total costs for all users is obtained through an increased bicycle mode share.’
I wouldn’t endorse the 1.2m bike lanes that they propose to allow passing (which look even in cross-section horrid) but what’s important here is that separation of flows led to time benefits for the larger vehicles (as ‘costs’ largely depend on travel time). They don’t seem to look at stopping vehicles though, which to me is an important part of the picture – it’s the flip side of buses trying to overtake bikes, and important for bicycle time and delays.
So, yesterday I finally decided to start having a look at it myself, using the free agent-based modelling software package Netlogo. I made a very, very simple ‘toy model’.
I created one link where I hypothesised that people on bicycles decided to ‘stay back’ behind stopped buses. On a second link, bicycles would instead overtake stopped buses if the adjacent lane was clear. And on both these two links, buses would overtake bicycles if the adjacent lane was clear. The assumption being that, as with most UK bus lanes, widths are not sufficient for bicycles and buses to overtake entirely within the lane, therefore, the presence of motor traffic in the adjacent lane can prevent overtaking.
(I could have included a scenario where buses had been instructed to slow down behind cyclists; we could imagine that instead of ‘Cyclists Stay Back’, the approach had been ‘Cyclist: Stay Back’).
On the third link, I created a separate bike track, with a bus stop bypass. I decided pedestrians would have priority here and so cyclists would need to wait for bus passengers to cross the cycle track. Of course, this is very abstract, and I’m only including links, not junctions, etc. etc. I kept numbers of buses, bicycles and cars the same, whereas varying them would be an obvious next step for taking this idea forward.
The way the model behaves is not earth-shattering, but it hopefully lets us think about quantifying things currently left un-measured. Firstly, we know people generally prefer to interact with other non-motorised traffic while cycling, rather than interacting with buses and other motorised traffic. For most people, the fewer interactions with motorised traffic, the better. Not surprisingly, the third link performs best on interactions with buses, and the second link, where you have bus-bike leapfrog, worst. (Even where – similar to the situation modelled in the paper linked to above – lanes are wide enough to freely permit overtaking within the lane at any time, bus-bike interactions by definition still exist, unless buses and bicycles never overtake each other: which would have obvious time penalties for both).
The second aspect is the one I’m more interested in here, though. The first link seemed to create quite a substantial time penalty for cyclists, as they waited behind stopped buses. The second link by contrast (probably in its abstract form the most similar to current London situations) seemed to generate some time penalties for buses, while reducing them for cyclists compared to the first link (the extent to which this happens will depend on levels of traffic in the adjacent lane). The third link enabled buses to flow freely without delays due to cyclists, although cyclists had some time penalties, due to waiting for pedestrians. (These looked generally lower than for the first link).
I’d love to get the chance to model this more properly, including using more realistic road networks and dedicated software. (Please do email me if you’re a Vissim/Aimsun etc. whizz and would be interested in collaboration.)
But I find it interesting that potentially separating buses and bikes at street level seemed to imply the possibility of time benefits for buses (and hence bus passengers) compared to the current situation of bus-bike leapfrog. It could even be seen as contributing to maintaining the flow of traffic (that duty under the Traffic Management Act 2004, which has so often stymied improvements for cycling and walking despite technically including pedestrians and cyclists as traffic).
Of course lots of other things contribute to timings and delays (e.g. junction design, layout of any bus stop bypasses, user behaviour, etc. etc.), but I think we need to better understand the potential impacts of mixing versus separating on delays to buses and bikes, and include this in the mix. It might be one way in which some apparent conflicts are not quite as they seem.