Bringing pedestrian maps to Crystal Palace

I’m leading a Transition Town project to bring Legible London to Crystal Palace. You’ll have noticed these signs around central London, conspicuously absent across most of the rest of the capital:

Bringing pedestrian maps to Crystal Palace

At the local Transition Town AGM, somebody suggested we should try to bring these to Crystal Palace.

But why wait on TfL? Using OpenStreetMap and TileMill, we can try to produce our own similar map style and stick them up ourselves!

Here is my prototype for the wide area map:


You’ll notice dark blue lines along the edges of roads and though the park. Those are pavements and footpaths, and indicate where you can walk.

Here’s a prototype for the more detailed map of the local area, with local points of interest, pedestrian crossings and bus stops:

Bringing pedestrian maps to Crystal Palace

You’ll notice that some roads don’t show any pavements. Eagle-eyed locals may also spot missing cut-throughs and wonder about missing points of interest.

To that end, this Saturday I’m running a stall at the Crystal Palace Food Market on Haynes Lane from 10.30am-1.30pm where people can help us gather data for footpaths, pavements and other useful features for pedestrians.

I’ll also be getting feedback on the cartography. One idea I’m pondering is producing themed maps. For example, we could omit pavements where air pollution is over legal health limits, or we could draw on political boundaries and colour-shade the infamous five boroughs that meet in the area, or we could add blue plaques.

Any mapping and cartography enthusiasts are welcome! I will have printouts of the local area for people to scribble on and bring back to the stall, or to drop off in a local cafe where I can collect them later in the week.


Robert commented that he has seen Legible London outside of central London. It’s true, I said, they do go further afield to some major town centres. Well, courtesy of the latest evaluation report for the scheme here is a map of their extent:

Bringing pedestrian maps to Crystal Palace

Just imagine if that map could be absolutely covered with yellow dots, courtesy of a community-led, low-cost OpenStreetMap solution!


  1. I really like the idea of using a casing to show details of pavements. I’ve been experimenting myself with TileMill, but modifying the standard OSM cartography with a view to making it more relevant for walkers. My current idea is to only display missing pavements on residential roads in built-up areas. For country roads I would like to display whether there is a pavement, a walkable verge or neither.

    15th July 2014
  2. Tom Chance said:

    It’s not actually casing, it’s a separate set of rules to draw offset lines, so it will distinguish between sidewalk=both/left/right.

    I’ve been in two minds about what to do with very minor roads with very little traffic – at the moment I use sidewalk=shared and draw it as a fat footpath. This works in urban areas, where these are usually small service or residential roads without pavements which people walk down as a matter of course, without needing to concern themselves about cars. There are some other very minor, low traffic service roads, e.g. in the main park, which I’ve left as sidewalk=no because there nobody really uses them in that way. Knowing the area well, I can make the judgement based on observed behaviour.

    I’ll post the stylesheet when I find the time to clean up my nasty hacks.

    I started out with the OSMBright package from Mapbox.

    15th July 2014
  3. Amaroussi said:

    sidewalk=shared for when there is no footpath and pedestrians aren’t banned.

    sidewalk=no/none for when there is no footpath and pedestrians aren’t banned, but pedestrians don’t use it anyway because, for example, it’s in the middle of an inner city gyratory.

    I do not use foot=no, unless it is specifically signed, or the road obviously looks like a motorway (e.g. A12 in Leyton).

    15th July 2014
    • Amaroussi said:

      Sorry, but I forgot to add:

      sidewalk=seperate, for when there is no footpath and pedestrians aren’t banned, but there is an adjacent footpath in lieu of it (e.g. staggered crossings or separate walkways on the A4 Great West Road).

      15th July 2014
  4. Robert said:

    Just to annoy you, I’ve seen the legible london pillars all the way out at Uxbridge.

    15th July 2014
    • Tom Chance said:

      Yes, they do crop up in major town centres and various other places too, but not in Crystal Palace… yet! 🙂

      15th July 2014
  5. You might care to put cycle parking on the maps; it’s not unknown for cyclists to stop and use this sort of thing. In OpenCyleMap style only big parking spaces (20+ places) on the big map, all of them on the small.
    Of course, this latter would require a survey of the area to be covered to update the cycle parking nodes, but that is not beyond the wit of man.

    15th July 2014
    • Tom Chance said:

      Tom, that’s a good idea, it might actually be interesting to have pedestrian and cycling maps side by side. We could show the LCN+ routes and bike shops as well.

      16th July 2014
  6. Another thought: I would avoid having the bus stop symbols cross the street and consider having the routes served by the stop annotated. Yes, this can change, but so can anything. Might also be worth having the railway station name next to the symbol as well.

    16th July 2014
    • Tom Chance said:

      That’s a limitation of how far I’ve got with the technology (TileMill, using CartoCSS and PostgreSQL). I tried to have it automatically move the bus stop symbols so they touch, but don’t overlap, with streets but haven’t managed to make that work yet. I just do that adjustment manually when I produce printed maps.

      The routes – it’s an idea I’ve thought about. Again it would be great to do it automatically with the technology. I might try adding it in and seeing if it looks cluttered.

      16th July 2014
      • Ed Merritt said:

        This is great, Tom. Glad to see that you’re taking on the challenge of bringing Legible London to Crystal Palace. I’ve just stumbled across your blog. I’m a cartographer and, coincidentally, I used to live in Crystal Palace. I will keep an eye out for your progress with this. I now live too far away to be able to get to your stall at the weekend, unfortunately. However, I might be able to help with one of your TileMill queries.

        Regarding the road names which clash with the bus stop symbols, in situations like this, I tend to use two different layers for the roads. One layer for the roads themselves and a duplicate layer for the road labels. I then move the layer with just the labels above, in this instance, the layer containing the bus stop symbols. TileMill will then, hopefully, reorganise the labels so that they don’t overlap with the bus stops (assuming that you don’t set them to overlap).

        I’m also wondering, if you’re using image based markers for the bus stops, whether you could add an empty gap at the bottom of the image file. This might have the effect of offsetting the marker further from the centre of the road when it gets applied in TileMill?

        Anyway, keep up the good work.

        16th July 2014
      • Tom Chance said:

        Thanks Ed. I have now managed to fix the issue of bus stops overlapping with roads, I just needed to dig out some trigonometry from the depths of my school days (I had forgotten to convert degrees to radians).

        17th July 2014
  7. Tod Robbins said:

    Reblogged this on New Provo and commented:
    Civic-powered wayfinding via OpenStreetMap. A very cool project.

    23rd July 2014
  8. Harry said:

    Have you considered that parks generally have more than one entrance, and therefore need a differently oriented print according to the direction you are facing when you look at the map ?

    Sadly, although we were all taught at school that *paper* maps should be drawn facing North, our teachers often never bothered to tell us that this applies only to paper maps which can be turned as you progress along a route. But when you’re printing a map to be displayed on a signboard, you need to know which way the board will be facing and print the map accordingly.

    That’s one of the strengths of the street maps in Central London — they are always correctly oriented and if you go round the back there is another map for when you are facing the opposite direction. Its not quite as simple as rotating the map to the correct angle — you also have to optimise the lettering for optimum legibility in that orientation.

    Sadly, most councils delegate signage to somebody that doesn’t understand orientation, consequently most parks have extremely unhelpful maps that face North even though the signboard is not facing in the corresponding direction.

    28th July 2014
    • Tom Chance said:

      Hi Harry,

      Rotating maps is something I’ve thought about, prompted by the Legible London maps that – as you say – are oriented to the direction of the viewer.

      It’s an interesting challenge because the tools I am using just turn out north-oriented maps.

      In Crystal Palace Park the main maps are actually oriented in a completely different way. The park is roughly shield shaped, so the maps show the “top” of the shield at the top of the sign. That’s neither north, nor the orientation of the viewer. The topography is also quite seriously distorted. You can see an example here:

      Even if I can’t get the orientation to match the viewer’s line of sight, at least getting a more topographically-correct map oriented to the north and placed around the area will be an improvement on what we have at the moment! I’m a firm believer in getting something done that is good enough, and gradually working in iterations towards the perfect solution.


      28th July 2014
      • Harry said:

        I’m not sure if it will help, but because the raw data on OSM is just a collection of nodes and attributes, it is possible to extract the data for an area collectively and rotate it to a new orientation. This does not without affecting its ability to be rendered by a mapping program, which will use its current rules to optimise text direction in the new orientation.

        As an example, I’ve extracted the current data for the park area and rotated it through an arbitrary 120 degress. I don’t think I can show an image as part of the reply, but the rotated data can be found at I’ve also put a png run through maperitive at

        4th August 2014

Comments are closed.