The Spatial Miscellany


A weblog. A website. A geospatial miscellany…

Is the writing on the wall for the OS Grid Reference?

I recently shared a few beers with an old university lecturer. We noted the ever improving access to the excellent Ordnance Survey mapping data but tongue in cheek he pondered…

“Is the writing on the wall for the OS grid reference?”

Here is a guest post from University of Edinburgh GIS lecturer, Bruce Gittings…

I see multimap have become the first to offer OS data on a slippy map
interface to a web service.

The writing is certainly on the wall for anyone contemplating a system which isn’t slippy – - that’s now the benchmark, anything else is dead technology. We’ve known space is continuous for a few years now and the average GIS toolkit gave up on ‘coverages’ and ’tiles’ a while back.

Multimap OS data

At the same time multimap also seem to have moved to using latitude and longitude, instead of eastings / northings on the national grid, probably as a function of the software they are using. The map is actually skewed to the window – most unusual. So, as more and more applications want WGS84 / GPS compatibility will the OS grid reference disappear? For decades this has been the de facto (indeed virtually the only) referencing system use for locating places on maps in the UK (excepting one-off grid systems used on town plans for example), despite a lat-long grid appearing subtly on the edge of OS sheets. Yet, the dash to lat-long driven my pan-national web services seems to be quickly moving the amateur place-referencer quickly towards lat-long too, if the dreaded wikipedia is anything to go by (wash my mouth out with soap for mentioning that word!). I wonder if they realise the implications?

The Wikipedia Flaw?

There’s a lot more misunderstanding about lat-long than the good old simplicity-itself national grid. The oft-used precision (e.g.
55.12345678901234556 degrees north) – just because the computer can calculate to that number of decimal places – takes positioning to a ridiculous sub-atomic level. Even multimap – who should know better – have fallen into that trap. Also, most people don’t understand it is not possible to convert between lat-long and eastings/northings terribly accurately, and that, unlike eastings and northings, two numerically identical lat-long positions don’t necessarily refer to the same point. But let’s not pretend I’m a mathematical whizz who actually understands coordinate systems and transformations, but its all here if you want the details: A guide to coordinate systems in Great Britain.

But all of that only slightly undermines multimap’s achievement. I wonder if they had to get special permission from the OS to use their data in this way? It makes OS copyright a little easier to infringe, but also again (like google’s first) shows the remarkable usability of the slippy-interface and its underlying AJAX technology. This ‘blogger’ particularly likes multimap’s feature which identifies the ‘locality’ which continually updates as you move cross the map. They must have built a load of theissen polygons from the OS gazetteer to underpin this. The result works well in some places, but there’s no sign of the name on the map in others. Less good is the use of getmapping aerial imagery, which seems to have a lot of gaps, many more than it should have – people’s map wins there. And performance seems to be a struggle, you need big banks of powerful servers before you can offer a professional slippy / AJAX service – the demands of the clients are probably two orders of magnitude more than the previous technology.

Good try though – 6 out of 10.

Above Cartoon by Peter Steiner. The New Yorker, July 5, 1993 issue (Vol.69, no. 20) page 61

9 Comments, Comment or Ping

  1. Multimap moved to latitude and longitude a little while ago, this was done so that all their data was in one map projection. They had two separate versions before converting OS grid data to lat/lng.
    It’s a good move and with the Open API you can see the benifits
    As to demise of the OS Grid. No chance OS will only map the UK.

    Here is a good link using OS 50K data using the Open API
    (Full Screen OS 50k – in lat/lng but Barry has add OS Grid tile conversion script)

    Open API


  2. Rob

    I had a look at the OpenAPI at the weekend…awesome.

  3. Glad you’re enjoying the slippy OS maps on and through our Open API.

    You can use OS Grid References (and eastings and northings) as search terms on via the address search box, and through the ‘qs’ (quick search) parameter of our API. For example here’s a search for OS Grid Ref NN166712 (the &dp=841 part gets you OS map data):|NN166712&dp=841

    Eastings and northings can be specified using the form osgb:easting,northing:|osng:254343,123433&dp=841

    We also support mercator co-ordinates in the form mercator:x,y and latitude longitude as lat,lon in the same search field.

    We’re currently working on a much improved linking (query string parameter) specification for the site (and accompanying documentation), which we hope will expose this functionality far more explicitly.

    In addition, if you scroll down below the map on UK maps at appropriate zoom factors, you’ll see the OS Grid Reference of the centre of the map is displayed.

    Hope this is of interest.

    Richard Keen

  4. >unlike eastings and northings, two numerically identical lat-long positions don’t necessarily refer to the same point.

    Are you referring to differing datums? Because really, any northing/easting system can suffer datum misidentification just the same as lat/lon. It happens that OSGB does define a standard datum, but so does most every national mapping standard, whether it be grid based or lat/lon. Outside of a datum issue, I am not aware of any situation where “numerically identical lat-long positions” refer to a different point. Could you give an example?

    BTW, I still remember your Domesday book of DEM data sources from the Dark Ages. :)

  5. Jonathan

    Can you explain “the implications” of moving to lat-long? All you seem to be saying when you admonish its use is that a) it’s too precise, b) it doesn’t convert from eastings/northings very well and c) two identical grid refs can refer t the same points (really? I’ll take your word for it).

    But what relevance does any of that have to the users of Multimap? And why the Wikipedia swipe?

    PS: To answer your question about whether MM had “special permission from the OS to use their data” – I’m willing to bet that this took the form of a shedload of cash, no doubt somewhat spiced by the knowledge that if OS didn’t allow it, it’s competitors would (the notion of “competitors” here being a little perverse seeing as OS benefit from a large slice of tax-payers money, but I’ll let that pass…)

  6. Jonathan

    Nice CAPTCHA by the way – makes a pleasant change from the usual nagging.

  7. bruce

    OK, In turn:

    MAPPERZ, take care. Lat-Long is a grid-system, NOT a projection. Lat-Long can have multiple projections, and each projection multiple datums. Beyond that I’m no expert.

    That also explains why to lat-long locations can have different points on the map.

    JONATHAN: I’m not necessarily admonishing its use (in fact I think we have little choice because it is (a) the de facto system of google and etc. (b) international, but it does present problems for those working happily within the OSgrid world which we have been used to in the UK. I think the (a), (b) and (c) you pull out of my post are pretty good reasons to be worried. Very few people properly understand geodesy, grid systems, projections and etc. well enough to use them properly and I fear many systems are being designed incorporating genuine mistakes. I did a poor job myself earlier in the summer when I implemented the basic maths for the conversion as a PL/SQL function, but forgot the datum shift. Hence errors which should have been 1-5m became 80-150m.

    Re WIKIPEDIA: that would take a long time and another post to explain in full, but in short I am not a fan of a so-called factual encyclopedia which is written by amateurs, has not proper editorial control (consensus can often be wrong), changes day-by-day (or minute by minute), has untrained contributor who regularly plagiarise. It is certainly NOT a way to arrive at a definitive place name gazetteer.

    For more details, see:

    Agree over OS.

  8. Rob

    Thanks for the comments, nothing like some bold statements and speculation to generate a bit of interest.

    As for the question, “How did OS mapping data end up in Multimap, and their API?” Shed loads of cash? Perhaps the answer lies in ‘The Power of Information’ report by Steinberg and Mayo, which recommended to government the OS release a web mapping API [OpenSpace], note paragraph 67…

    “Interviews with OS suggested that the major barrier to launching the application [OpenSpace – an OS mapping API] came from its relationships with smaller suppliers, who resented the possibility of non-commercial re-users obtaining free access to information that they had had to pay for as part of their commercial arrangements”

  9. Jonathan

    “…resented the possibility of non-commercial re-users obtaining free access”?

    And what did the report say to that? The only reasonable response I can think of is “tough s**t!”

    In any case, the flip side of the objection to other people getting stuff for free is that you can too. Unvarnished jealousy seems to be a rather odd reason not to fulfil a public service.


Free GIS Software...

Download ArcGIS Explorer, a free geobrowser from ESRI!


Use ArcGIS Explorer to visualise geographic datasets. The latest build provides full access to Virtual Earth imagery and comprehensive support for several data formats including GeoRSS, KML and ArcGIS Layerfiles.

Before you go

Going so soon? Test your geography with the...


Do you support the campaign? Should government-funded and approved agencies such as the Ordnance Survey collect data with significant indirect contributions from the UK tax-payer, but then charge users and companies for access to it?


Download Flash plugin