Saturday, 28 September 2013

Magical Modelling, Digital Deceit and Swept-path Subterfuge

This week, a Twitter debate about traffic modelling has inspired me to lift the lid on a couple of the shortcomings and sharp practices used in traffic engineering where the use of computer modelling is concerned.

Before I start, there will be opinion in this post and I encourage you to do your own research, ask difficult questions and challenge assumptions; because as any computer expert will tell you "garbage in, garbage out". I am not an expert in the intricacies of traffic modelling, but I have rumbled cheats over the years (more on that later). 

By modelling, I don't of mean mock-ups in Lego (although Copenhagen is nicer than London for cycling), I am referring to computer modelling of traffic which is often used to test different road layouts. In my opinion, it is often an utter waste of time when a traffic authority has a policy of prioritising motorised traffic (yes, Transport for London, I can see you there hiding at the back).

There are lots of different pieces of software out there that help us design our schemes which is really useful and is not in itself a bad thing. The problem is what people use the data for, or how they manipulate it, or what they are trying to show and for which audience. 

Let's start with very simple example which is based on a real investigation I dealt with a few years back (it still haunts me). This story can be applied to countless locations in the UK.

We have a signalised crossroads type junction where an 'A' road runs east-west with two local distributor roads (which are probably just as busy at peak) coming in from the north and south.

Each road is single carriageway and where the road approaches the junction, it flares out into two narrow lanes. In reality, each arm has an advanced cycle stop line (no feeder lanes), but that is not important for this example.

In all cases, the left lane is left / ahead and the right lane is right only. For pedestrians, there are no green men at all, but a narrow refuge in each arm. The traffic signal staging is a doddle. the east-west arms run together, then the north-south arms run together and that's it. Pedestrians try and leg it on the all red signals between opposing stages while drivers in the middle finish their right turns!

The investigation I was involved with was simple - what was the feasibility of providing green men (pelicans) at the crossing. So, we needed to model the junction and that involved spending a ton of money getting traffic and pedestrian counts for the junction at various times of the day, sticking the results in the computer and then modelling different layouts. I knew that in traffic capacity terms, we couldn't make it work as the area is always stuffed at peak times. I knew councillors wouldn't go for it as it would stuff the junction for drivers even more. But, we pressed on, I thought it was important.

Staggered pelican crossings are often used to maintain vehicle
capacity over pedestrian comfort and convenience.
We modelled three things. Sticking in a new stage for all-round green men (all traffic held on red); widening the junction to provide staggered green men crossings to allow pedestrians to move when traffic is not in conflict with them; ripping the lot out and sticking in a roundabout with zebra crossings at each arm. 

We did the modelling for the two signal options with LinSig, which I won't even attempt to explain in detail (as the intricacies are mainly beyond me), but it allowed the data to be input to create a model which can then be used to test different arrangements. The base method of control was tested to see what the capacity was at the time and it showed that things were running pretty much at capacity (for traffic), with significant traffic queues at peak times - which told us what we knew from observation and so the model was working.

Junction capacity is an interesting concept to get one's head around. When looking at a signalised junction in isolation, for every cycle of the signals, there is a theoretical amount of traffic (including pedestrians and cyclists) that can be stuffed through. The trouble is, that when a junction runs beyond 90% of theoretical capacity, the flow starts to break down and capacity drops off. For those in the know, this is like a water pipe running just under full. The inside of the pipe exerts friction on the water and it can be shown (by that there mathematics) that the point where the pipe is just under full is actually more efficient than full, because of the little space at the top where friction doesn't act on the water.

Traffic is the same concept. The "friction" in this case will be things like people hesitating at a gap rather than turning, traffic blocking the exit from the junction, a large vehicle turning (if the turn is tight - classic issue in constrained city sites) and even things like pedestrians being invited to cross by a driver. So at 100%, the demand on the junction and its capacity or equal and so it is full up (i.e. saturated). The relationship between demand and capacity can be expressed as a percentage know as Degree of Saturation (DoS) and hence the magical 90% figure. Beyond this, the flow gets broken down as explained above. Strangely, it is also possible to exceed 100% of the theoretical capacity. 

The 100% will have things like safety built in where it is assumed that people obey the signals and take sensible gaps to turn. Locations exceeding 100% are where people are regularly (every cycle) going through amber and red signals, taking very small gaps to turn or driving to closely to the vehicle in front. This will very much depend on the junction and the types of drivers, but is something which can be observed on busy commuter routes where the assumption will be that generally the same people use it every day and pedestrians are often excluded from having green men unless they can be mixed in with the signal phases.

This road is stuffed. The only way to improve capacity is to add traffic
lanes. Bye bye trees, front gardens and possibly buildings. Hello
pollution, noise, crashes and eventually, more lanes of
traffic. Would protected cycle tracks reduce demand and
so increase capacity?
To increase capacity, you cannot just add green time as the overall cycle time increases leading to congestion. So, like a pipe of a certain size, a junction will have an optimum capacity and to increase the flow, you need to increase the size of the junction to cope (like a bigger pipe for more water flow). Of course, capacity can be increased if demand is reduced and the now spare capacity can be given to other modes - traffic falling off within inner-London?

Long-winded, but I think useful to know. For the roundabout option, ARCADY was used, which again uses flow data, plus the physical layout of a roundabout such as number of lanes into and out of the junction, lane width and so on. For roundabouts, we use Ratio of Flow to Capacity (RFC) to see if the junction will work using the magic number of 0.85 as the threshold. Essentially 85% of the theoretical capacity is the maximum flow rate. The zebra crossing component is a nightmare to model as we don't have pedestrians crossing in nice little groups as we would for green men and so there is a fair bit of assumption. Of course, people crossing in large numbers on one arm  for more than a few seconds would lock up the roundabout.

So, the results showed that all three options would increase queue length to the junction at peak times with the roundabout / zebra combination being particularly bad. So, what did I present to councillors? Well, I suggested that the the roundabout was out and the staggered green men would be a large land take and therefore cost (moving buried utilities and so on) The staggered option made the junction larger and off-set some of the congestion created by the green men. To have completely mitigated the issue, the junction would have been huge and there was not enough land available! The introduction of green men on all arms was the simplest, but would have created chaos for drivers. 

The summing up was that on the one hand the junction could be left alone, as when looking at the casualty-rate, green men would be unlikely to affect the number off pedestrians being hit each year; yes, sadly, we do deal in cold harsh numbers and pedestrians will statistically get hit at signalised junctions as there will be drivers jumping red lights against green men and pedestrians will cross on a red man against a green traffic signal. At this junction, the actual rate was about the same as would be predicted for signals. Therefore, the do nothing option would mean drivers are not impacted more then was current. On the other hand, the junction is horrible for people to cross and green men would allow people to cross easily, but at the expense of traffic congestion. This is pretty much Hobson's choice as the councillors didn't fancy causing any more traffic congestion there remains no green men at the junction today.

I stated earlier that this junction haunts me. Not because I get complaints from pedestrians (who put up with most things most of the time), no because in hindsight, I should have been more pushy. Of course, my job is to be objective and impartial in giving advice. I know that I would have been on a loser and probably criticised, but perhaps I should have ramped up the issues facing pedestrians?

Traffic capacity reduced to create space for protected cycling. I bet
this wasn't modelled, it just took political will to shift the status quo.
This is a classic example of modelling being a waste of time because I knew the green men on all arms would be the cheapest and simplest option to deliver and I knew traffic queues would be worse. For the councillors who read and understood the report, they knew the same as me. So, why didn't I just save a load of money and propose green men and to hell with traffic flow? I suppose it is the game that we play. Do loads of work, write a technical report and go with the charade of explaining the implications to committee because if we just put the scheme in and it annoyed drivers, we would be asked why didn't we foresee the problems. Why didn't we model it? 

In this case, the only winner was the consultant that did the modelling (and I am not criticising them for that, they looked at everything and earned their fee). It is a charade we go through and I think that in many cases those thinking that a situation needs modelling, are half expecting there to be capacity issues. Wouldn't it be simpler and cheaper for politicians to start from the point of view that they wish to prioritise walking (and indeed cycling) and where there is no space to build, admit that this will be at the expense of drivers. No modelling, no arguments, just get on and build it. OK, the flip-side is to sod non-drivers which this country has a great tradition of doing - but at least the public will know where they stand and the politicians will not be able to hide behind a technical report.

Of course, I do modelling some disservice. It is possible to create very complex layouts and test them and some of the software is very clever indeed. VISSIM, for example, can create 3D virtual layouts which show little cars and buses driving around the model, cyclists pootling along and pedestrians crossing the road. I have seen VISSIM running on junctions I know well and it is quite spooky. This is the world of micro-simulation and could be a whole post in itself (by someone else who understands it!). 

Modelling allows one to test ideas and optimise existing layouts, but it is not a magic spell that will give you extra highway space and people need to bear this in mind. You might be able to tweak the junction today to improve capacity based on today's flows, but if those flows change or traffic grows, your advantage has gone. I have worked on another junction in a busy town centre (which has green men) where the conclusion is that unless demand decreases, we are stuck with the congestion. We looked at widening the road which reduced congestion, but it meant footways were lost to tarmac and to be fair to the councillors I presented the findings too, they wanted to prioritise pedestrian space and so we left things alone.

So that is the nice end of the stick, the plucky local authority engineer doing his best knowing that any argument to provide for pedestrians at the expense of motorists can be like farting into the wind or the micro-simulation nerds (in the nicest possible way) who can do amazing things with computers. What is the nasty end of the stick? The foul underbelly of highways if you will? 

Well, that would be developers. Now, developers are not necessarily born evil, some are very nice indeed. But some have hearts as black as their stinking souls and even worse are their unholy consultants. Before the saintly consultants protest (they may even read this blog!), I am not having a pop at you, but I am not naming names - they all know who they are! So, what do the naughty ones get up to?

Cribbs Causeway, Bristol. A veritable shopping city. I bet the TA was
enormous, the roads were nicely modelled and it has had no negative
impact on the city at all.
Image Google Maps.
From the developer's point of view, the time between buying land and selling homes or letting units is a temporal money pit. Time is literally money. Before building can start, the developer needs to get planning permission and as part of the planning application, a transport statement (TS) or transport assessment (TA) must be prepared. At the cynical end, the TS / TA must show that the development won't stuff up the roads as a result (for traffic as usual). At the enlightened end, it can positively enhance the design and value of the development. 

A TS will tend to be very light, support modest schemes, give basic and sensible facts. It is unlikely that a scheme needing a TS will create any major issues. A TA is a different animal. They can be the thickest document of a planning application. As well as summaries and useful information, they will have appendices of bus and train timetables, walking routes that Sherpas would be proud of and they always contain reams of computer prints showing traffic flows through junctions. Do you know what, I have never read a TA which admits a scheme will cause a problem to the road network and I have read an awful lot.

If I am a consultant working on a planning application, my job is to get my client its planning permission, or at least show there are no highways and transport impacts. Say you are considering a new development without parking as it is in a city centre with parking controls. At worst, the development generates refuse collections and the odd delivery. The impact will be more demand on non-car modes and so cumulative developments or a single large one can mean a higher frequency bus route or even a new railway station. Where car parking is provided (depending on local planning policies), it will generate vehicle movements and this will impact on junctions or create conflicts where the site is accessed.

Consultants will always want to show there is spare capacity (remember DoS and RFC for junctions) and their client's scheme will not make a junction exceed capacity. So, what are the tricks used in the traffic modelling? Well, they can undertake traffic surveys when traffic flows are lower such as over the summer (they either didn't leave enough time to undertake the surveys or they are trying it on). They can look at the peak flows and top or tail the data so some of the peak is conveniently missed (e.g, they report 8am to 9am, when the real peak is 7:30am to 8:30am). They tweak the physical attributes of the models making lanes a little wider than they are or radii a little larger to make flows a bit smoother.

They also like to bury DoS or RFC information in the data appendix and that is one of the first things I read! You can quite often find the magic figures of 90% and 0.85 being exceeded, but they get averaged out in summaries and things seem fine at a glance. I have prepared evidence for planning appeals which includes dissecting appendix data and it does make one's brain hurt. The good consultants play things straight and advise their client early on and good developers make changes to their plans or admit there are going to be issues, but propose physical mitigation up front and good on them. In some cases a planning committee may decide to accept congestion being created if all of the other issues are positive or it suits the ruling administration's policies or plans.

OK, enough of the theoretical capacity arguments, what about kerbs and tarmac I hear you scream? Before we hit the streets we need to draw up plans. My office uses AutoCAD to draw up scheme plans (other design packages are available) and it is a fantastic tool. We also use Autotrack which is swept path analysis software (I will come back to it in a bit). AutoCAD is not just about pretty pictures, it allows schemes to be constructed with a real world layout using real-world coordinates. We design things tied into the Ordnance Survey eastings and northings system which your smartphone GPS can give you. If we have an OS map in AutoCAD, we can overlay our designs in a model of the real world.

Anything slightly complicated will need a survey and so you get a team of surveyors in, they pick up all the details on site with clever and expensive GPS-enabled kit. They can link it to ground penetrating radar (GPR) to pick up the line and depth of buried services (which can also include tracing pipes with a radiosonde which is pushed through pipes emitting a signal). It allows us to know if there are clashes with shallow services if we are moving kerbs or installing foundations - cool stuff indeed.

Of course, this is all open to abuse. One of the wheezes is to change the scale of the real world model so that there appears to be more space than there is and so shows that the layout fits. It is not until things get to site does the problem get discovered as the models stay in the designer's computer and get printed out for the groundworks contractor with a "do not scale" note (which is universally ignored!). The other bit of naughtiness is dimensioning. AutoCAD can dimension things for you and will tell not lies, but what if your parking bay is not quite wide enough? That's OK, we can just over type the dimension with the "right" measurement. Your 1.6 metre wide on-street parking bay, now measures a wonderful 2 metres! If things are space critical or there is a whiff of suspicion, we ask for drawings in digital AutoCAD format. We can check for anomalies and it is amazing how some consultants suddenly get very protective of their intellectual property and only want to issue a PDF or hard copy!

The excellent Autotrack in action.
Image from Savoy Computing.
Back to Autotrack. When you have designed your new kerb line and you have missed the buried utilities, it is quite nice to check that vehicles can use it. Autotrack allows you to drive little cars and buses around your layout (or follow paths drawn by you) where you can check there is space. The model shows the position of wheels and vehicle bodies as it moves and so is known as swept path analysis. How can anyone cheat which such benign technology?

Well, when you set up your test vehicle, there are a whole load of parameters to play with. The two best ones to tinker with are forward speed and lock to lock time. Imagine you have a signalised junction with a left turn that is a bit tight and it is on a bus route. A tight turn means that the bus drivers need to slow down to make the turn, but this reduces the amount of traffic that can get through (as they are stuck behind a bus) and so the capacity is compromised. 

If the swept path model is run at "normal" traffic speed, the bus will not be able to fit round the turn without bouncing over the footway, or a traffic island or whatever. But, if you reduce the forward speed to a crawl and make the lock to lock time very quick, that bus will now appear to glide through (just don't tell the guy doing the capacity modelling). 

Lock to lock time is interesting. Most people have never driven a large vehicle (I have!) in a constrained urban area. There is a lot to think about when you drive a large vehicle, but one of the key issues is it takes a while to turn the steering wheel from fully in one direction to another. You also have body overhanging (worse for buses) and so the body sweeps a larger area than the wheels so you might need to slow down. 

A car with power steering is easy to turn corners with, just whip the steering wheel round. With a large vehicle the power steering is pretty much a necessity, but it takes a bit longer to move the steering wheel than a car. When going round a corner, it can take a bit of time and effort to point the wheels the right way (you will now watch the driver the next time you get the bus). As the vehicle is large, there is a lot of it to get round the corner and the slower you go, the easier it is. So, a very slow forward speed and unnaturally quick lock to lock time will show that the kerb line is fine. In reality, the real world driver will struggle!

All of this is linked. Engineers are there to solve problems and with highways, it tends to be trying to squeeze as much as possible from the asset. Cheating and naughtiness aside, I think the use of computers is fantastic and allows us to do things unheard of, even a few years ago. But, all too often, we take the blame for not sorting out congestion (well, you just rephase the lights don't you) or making life better for pedestrians and cyclists (and the criticism is partly fair). But, we cannot work miracles. Our designs cannot bend space and time and they cannot create new space where there is none. At this point, the politicians need to take responsibility and not hide behind technical reports or their staff.

With the announcement over the summer to throw billions at creating motorised traffic capacity and the news this week that Local Transport Bodies are ignoring sustainable travel, the outcome may well be making inter-urban car use easier, but this will be at the continued expense of our town and city centres in terms of congestion, deaths, injuries, pollution, transport poverty and all of financial impacts resulting. The last time I looked, you cannot stick that in your model!


  1. This whole article is bang on the money. Well done. As a traffic engineer and former consultant I've witnessed similar; the AutoTrack bit in developments is particularly valid. It's also worth noting that people sometimes use a "medium car" too, which, if you look at the dimensions is smaller than most cars on the road today as it was defined decades ago.

    I consider myself one of the good ones and I can honestly say I've not done these things. However, in more than one case, I've been taken off a project explicitly because I wouldn't, which I think in part, underlines the problem the engineer faces.

    1. Well, this is an issue of professional ethics. If anyone involved in cheating is a member of a professional institution and they get found out, they could be out the door.

      Of course, principles cost money in terms of commissions for consultants and individuals who sometimes get the elbow.

      Yes, medium car is another trick - we always insist on seeing the data. On more than one occasion, I have stopped reading a TA based on partial swept path data and invited the developer to check the report before resubmitting it.

      The ones who often suffer are the contractors on site who end up having to bodge stuff in because it literally doesn't fit - had one only last week, but we made it work after bollocking the developer and threatening not to adopt his road.

  2. Pretty spot on in the most part. On Transport Assessments, it's worthwhile bearing in mind that they are simply a report of a substantive process undertaken in order to get a desirable outcome. When modelling a junction, the TA often reads like "the junction doesn't work, but what we're promoting makes it work" when in fact there is a whole iterative process behind getting to that point - testing one layout, retiming a signal etc., to come up with an optimal solution. In the interests of keeping the TA readable, this whole process is not mentioned. If it was, there would then be understanding of the process at least.

    Another issue that I find is relevant peak periods, and the almost religious use of the confined traditional highway peak. Any engineer worth their salt - or who has at least used TRICS - knows that different uses have different peak periods, and need to be modelled accordingly. I recently had this argument over a new sports facility. The TA said that everything worked fine in the highway peak, yet was silent between 6pm and 7pm when the peak use actually took place. The response? "Well, you prove it doesn't work."

    Fundamentally, a model is simply a tool to aid decision-making. Whilst validation ensures a strong degree of robustness, just because the model says no doesn't mean something shouldn't be done should there be the will to do it. The biggest problem the profession faces in my view is how the model seems to determine all, and too many use that power to tweak the tool to achieve their aims.

    But at least there are the likes of us who call at BS for what it is!

    1. The trouble is that a set-up or layout that "works" for the TA, may not be the appropriate answer. I have seen proposals such as a new connection to an existing road which will only "work" with signals as it is the only way that traffic will be able to get out of the new development. The fact that it makes life worse for everyone else (but still within capacity terms) is the elephant in the room as essentially it means that the development should be rejected. But as you rightly state - prove it doesn't work!

      Peak is interesting. Take an industrial estate and compare with an office development. The industrial estate will often earlier peaks than the office site because of the different hours they keep.

      I often find the AM peak to be relatively short, but heavy with the PM peak a little lighter, but going on longer and often merging with school travel times.

      Weekends are also becoming as busy, if not busier than weekdays. Our local shopping centre opens at 11am on a Sunday and so if I need to take the car out, I make sure I am off the roads by about 10:30 as all of a sudden, the place will be stuffed!

  3. Superb.

    Comments from a non-highway engineer (an end-user if you like).

    I appreciate you are dscussing a very limited piece of the "plumbing" but IMO that is part of the problem: farfield effects really ought to be incorporated into the modelling. For example, it doesn't seem to account for drivers choosing an alternative route or even (gasp) a different mode of transport if a junction started getting clogged. I did. Sometimes I drive, sometimes I cycle to work . A few years back part of my regular 17 mile driving route started getting choked in the morning and it now takes about half an hour longer than when I started*. I put up with for a while to see if it was temporary. When it turned out to be as good as permanent , I found another route - 2 miles more, but 20 minutes quicker. The choking is actually mostly due to an increase in the school run cars: in holidays, or coming back in the evening on the old route, traffic doesn't seem much different to 10-15 years ago. I know this is "natural' change and not "artificial" as you are required to do, but the basic processes and outcomes seem to me to be similar.

    I suppose letting junction choking happen and then sort itself out would be part of the political will aspect.

    A couple of technical questions:

    1. Sounds like you model road traffic as single phase flow. Just wondering whether or not there would be a difference if cyclists filtered/didn'tfilter, took the lane or stayed at the side etc and a multiphase model would be more appropriate: i.e. cyclists share the pipe but don't have the same effect as other vehicles.

    2. You don't say whether you change the signal phasings (to help pedestrains) between offpeak and peak times. A lot of people are out shopping or walikng kids/pets during the day round me and motor vehicles are far and few between, so it would seem appropriate to benefit the pedestrian when they are in the majority (this applies to cycle paths and toucans etc too - I had to wait about 3 minutes for the toucan to change in one location with hardly any cars around [I could have walked across on red but I wanted to time it]). Sort of like James Gleave's sports facility point.

    Probably inanely naive comments from me but cheers for the excellent insights.

    * the worst part is Weybridge - especially here, and even worse here To me it looks like you'd have to knock down virtually the entire town to ease the congestion - the devlopers no doubt would love that.

    1. LinSig certainly allows many junctions to be linked together, but for really clever stuff, micro-simulation is the key. VISSIM for example can be used to model entire city centres, but to work properly, it needs lots of data which can be very expensive to collect. The model then needs constant updating as time goes on and like anything, with a lack of maintenance it will go wrong.

      The very clever part is that events can be modelled and so this could include even how pedestrians behave when boarding a bus. But of course, the more detail you want, the higher the cost and the more to maintain.

      There is also SCOOT to play with for traffic signal applications ( which does indeed allow different pre-sets to be used. So, it traffic is light, pedestrians can be given more priority. SCOOT also uses loop and microwave detection to count traffic movements on-the-fly and so can adapt. It is very powerful when a city area has all of its signals connected up.

      The trouble I find with lots of this clever stuff, it is often used to show that improvements for walking, cycling and public transport do not affect general traffic capacity and in many situations, politicians will not be explicit in who they are prioritising (hiding behind the reports).

  4. Hi, I found your article fascinating. I know nothing about traffic engineering but from other areas of technology, I know the fun and games you can have with computer modelling based upon the input data and assumptions made...

    Again, from an "end-user" perspective, near where I live in London there is a proposed development for a new stadium for Brentford FC and housing. In general, I'm in favour of the new stadium but cynically suspect that despite fine sounding words from the developers about cycle facilities and sustainability, we'll end up with the usual rubbish.

    I've skimmed the Transport Assessment and most of the stuff you mention is there... ARCADY, LINSIG, AUTOTRACK...

    Unsurprisingly, the conclusion is "all links operate below a Degree of Saturation of 90%" As you mentioned, I dived into the Appendix and was somewhat surprised to see the max. RFC quoted as 0.24 given the junction (Chiswick roundabout) is somewhat notorious for congestion. If that figure is accurate, I wouldn't want to see it when it is much higher!

    I couldn't find a reference to the size of vehicle used in AUTOTRACK. Is this generally made available or does it need to be requested?

    Their evening analysis stops at 18:15 while from my perspective as a resident, it always seems the traffic later in the evening around 19:00 is busier.

    1. Don't forget, everything dealing with capacity and queues in an average. So, where a junction is theoretically fine, there will be times when it is busier than average and so there will be congestion.

      To be clear, the capacity may be fine at the peak of say 8am to 9am, but this is an average. Taken over a 5 minute period or one cycle (with signals) one may well see congestion.

      I don't know the Lionel Road site, but have been through the Chiswick roundabout a couple of times and it is classic long-distance roads meeting more local roads.

      It is pretty large and being signalled, it will run better (on average) than not being signalled - advantage will have been given to the smaller roads and pedestrians.

      The development may well be fine - what is it replacing - what traffic does the existing layout generate. If it is industrial now, then it will attract trips at peak times. When it is housing, it will be people leaving the site going elsewhere and this may be more likely on public transport, unless jumping on the M4 to leave London.

      The stadium element will have its own peaks and of course, on match days, the local area will be swamped with people walking to bus stops and the local stations which will of course have more demands on footway width and pedestrian crossing width/ timings (which is often forgotten).

    2. WRT AutoTrack and the vehicles used. The program holds a profile of each vehicle (including a nice side view of it, and all the appropriate data like lock-to-lock time), and good practice should be to place this profile somewhere on the swept-path drawing. If the drawing isn't showing it, I'd be wary.


    3. Now that is a great tip - we sometimes ask for the DWG which puts the wind up the cheats!

  5. Thanks for the reply. The existing Lionel Rd site is semi-derelict land so there is no doubt that a 20,000 stadium and about 900 residential units will generate more traffic than that. While we are talking Brentford here and they don't have big crowds, you clearly can't rule out that they may reach the Championship sometime and there is talk about a ground share with a rugby club as well.

    As I said, I'm generally in favour of the development as it will regenerate an area with very little in it at the moment. While it has yet to get through planning, the local council appear to be in supportive. I can't see how they would increase the capacity of the surrounding roads so it may be a trigger for the local council to come clean and explicitly favour other modes of transport over private cars with the design of the scheme.

    At least now I'm a bit better informed when it comes to discussing the impact on the roads!

    1. One option is not to increase road capacity, but to improve walking, cycling and public transport. This would be a political decision, but as places such as The Netherlands and Denmark (and even New York!) have shown, you need to start somewhere, but if you are determined, you can make a real difference (IMHO).

  6. As Jitensha Oni mentioned, clearly the elephant in the room is that motor traffic is not constant.

    Until we can model individual transport choices - ie, individuals deciding what mode to use, and what route to use, depending on congestion, price, weather, kind of trip etc - instead of modelling motor traffic as a constant - the modelling will act as a way of re-imposing the status quo.

    Over time, users adapt to what's there on the ground (road space, safe cycle infrastructure, bus or rail services) - and make fairly efficient usage choices. Where it's difficult to make different choices in the short term (say Hammersmith flyover, where many users are tied into one form of commuting), treating motor traffic as a constant is a fairly good guide to at least medium-term consequences of changes. In a central London location, on the other hand, where users have many different choices (and taxis are a big part of the traffic), treating motor traffic as constant in a busy location with restricted space just leads us to re-build exactly the same mix of compromises, and prevents us from changing what's on the ground to allow all users the chance to make different (and better) travel choices.

    1. Very interesting point. Other modes have to be attractive. Take my commute, It is around 3.5 miles each way and so I can walk in about an hour, cycle in around 20 minutes, drive in about 12 minutes, take the bus in about 20 minutes (I exclude trains as I don't live near a station).

      My long term decision has been the bike as it is the same journey every time regardless of traffic congestion and it is literally door to door. If the weather is bad, then it will be the bus, I have a stop nearby which runs a service every 20 minutes and so with a 3 minute walk I am there.

      Driving is door my end, car park and a 7 minute walk at the other end, if the roads are reasonably quiet (so leave by 7.30am latest). The trouble with the bus is that I still need to leave early as it sits in the same traffic as the car, but it stops opposite my office.

      So, all modes are roughly comparable time wise, but cycling is less affected by variables (snow is the enemy) and so that is my choice based on time (before the other benefits).

      Of course, other people might make more choices in a given week and so how on Earth can they be modelled? I think there is a lot to be said for political leadership in promoting cycling for the 3 to 5 miles trips and this could be coupled with closing junctions on main routes (which would make building protected cycle tracks easier and stop local rat running).

      Not sure we have any vision in the UK, though.

    2. Check out discrete choice modelling, Ranty!

    3. Now there lies the path to some insane mathematics - I am glad I only have to throw tarmac around!

    4. To be fair, for larger developments (and for modelling larger networks)we have SATURN, which does account for different route choice based on the economics of delay (to motor vehicles).

      As for accounting for cyclists in the traffic flow, the basic unit of ARCADY, PICADY, LINSIG and TRANSYT is the PCU (Passenger Car Unit), wherein a car is worth 1.0 PCU, a bus 2.0,an HGV 2.3, a motorbike 0.4 and a bike 0.2.


    5. I guess it is the right tool for the right job, but with a transparent objective!

  7. TFL are real bastards for this kind of trick. "You can't have any cycle lanes cos the computer says 'no'". As if the opinion of some silicon chips is indisputable and should be the final word on the matter.

    1. To be fair, many of the staff are stuck with the political dogma of "do what you like, but don't reduce traffic capacity".

      Until the politics *really* changes we are stuck with it. A typical argument is that pedestrian flows over the TfL side road are low, so a green man is not justified - no allowance for latent demand or the sort of flows we would like to see in the future.

    2. Well then surely the honest and correct reply to "why can't we have proper cycle infrastructure here?" is not "computer says 'no'".

      It should be "Person A has dictated that motor traffic flow must not be compromised at any cost, even the cost of the lives of cyclists. We have plugged in some sub standard measurements into our half arsed models cos it's too expensive to do it properly. As a result we predict that proper cycle infrastructure will have a detrimental impact on motor vehicle flow. Person A has already stated that this must not happen so there will be no proper cycle infrastructure here. If you wish to take the matter further please contact Person A"

      This means that the people making the decisions to kill cyclists don't get to hide behind a wall of corporate bullshit.

    3. Engineers have the responsibility to explain the issues in the round, but I fear that when faced with political and management dogma, walking and cycling is not on their radar; but, if those making the decisions are doing so properly, they should throw reports back to be looked at again. Back to leadership, it always is.

  8. Do you have a view or any pointers to data on the impact of 20mph on traffic flow? 20splenty is primarily a safety campaign; but as a cyclist in urban/suburban SW London it is very noticeable how drivers accelerate from 0-35mph when leaving a junction only to brake sharply down to 0mph to queue at the next junction typically 0.5-1mile away; whilst the typical cyclist does a steady 12-15mph and catches the motorist at each junction. (aside - buses do this too and it really annoys me as a passenger to be thrown around - god help the elderly and infirm bus passengers). In areas such as this, wouldnt universal 20mph smooth traffic flow and significantly reduce fuel consumption/exhaust output?


  9. This is a good place to start:

  10. I would like to query something in regards to the manual DMRB below about Ghost Islands.
    It states in the manual that its for trunk Roads.
    Would this manual also apply to C Roads in the event that there was no other Geometric Design standards to fall back on.
    I have been looking at MfS (Manual for Streets) & other road manuals including Councils road polices on this matter
    & have found nothing that states what the minimum Turning length for a Ghost Island should be on a C Road.

    It states in the Design Manual For Roads And Bridges at 7.32 (The turning length should be 10 m, measured from the centre line of the “minor road”
    irrespective of the type of junction, design speed or gradient).

    But it states in the Manual for Streets 2 at 1.3.3 (Where designers do refer to DMRB for detailed technical guidance on specific aspects,
    for example on strategic inter-urban non-trunk roads, it is recommended that they bear in mind the key principles of MfS, and apply DMRB
    in a way that respects local context. It is further recommended that DMRB or other standards and guidance is only used where the guidance
    contained in MfS is not sufficient or where particular evidence leads a designer to conclude that MfS is not applicable)

    As there is nothing in the MfS about Ghost Islands should it default to DMRB or is there another standard that should be applied.

    I would be very grateful if you could clarify the position on this matter

  11. Forgot to add this for the aboth.

    PART 6
    TD 42/95
    Chapter 7 Geometric Design Features

  12. Well written! I am a teacher and found this a very good source. I'm also using some visuals from, i think the combination makes sense to my students.

  13. Well written! I am a teacher and found this a very good source. I'm also using some visuals from, i think the combination makes sense to my students.