[Talk-GB] OS and OSM

SK53 on OSM SK53_osm at yahoo.co.uk
Mon Mar 21 14:30:54 GMT 2011


On 11/03/2011 10:19, Steve Chilton wrote:
> I have been asked by editor of the Cartographic Journal to write a short piece on the effect of the release of OS OpenData on the OpenStreetMap project, and I am just trying to gather my thoughts, and make sure I cover all bases.
> I was present at Blackadder's Society of Cartographers talk on "Why OSM won't be bulk importing OS OpenData" and am aware of the work Chris Hill has done on admin boundaries etc.
> Obviously also aware of the ITO work with OS Locator and what people have done with that.
> There was work on importing detailed water features, was that Chris as well (goes off to read back through his blog).
> Can anyone point me to others who have explored the possibilities that OS OpenData provided - PARTICULARLY if they can evidence WHY it is NOT of value to OSM?
>
> Cheers
> STEVE
>
> Steve Chilton, Learning Support Fellow
> Educational Development Manager
> Centre for Learning and Teaching Enhancement
> Middlesex University
> phone: 020 8411 5355
> email:steve8 at mdx.ac.uk
> http://www.middlesex.wikispaces.net/user/view/steve8
>
> Chair of the Society of Cartographers:http://www.soc.org.uk/
>
> 'Inspire Me!' lunch time showcase on Assessment and Feedback, organised by the Centre for Learning and Teaching Enhancementhttp://inspireme.middlesex.wikispaces.net/
>
>
>
> _______________________________________________
> Talk-GB mailing list
> Talk-GB at openstreetmap.org
> http://lists.openstreetmap.org/listinfo/talk-gb

Sorry this has been sitting in my outbox for a few days. Ignore if not 
of interest now (post new release of VDM & duplication of others comments).

A quick summary of other imports known to me:

    * Boundaries: Notts/Derbys (me, will_p), Norfolk (PinkDuck),
      Tendring (EdLoach)
    * Power-lines: a few (me), lots in Scotland (tms13), chillly
    * These are relatively straightforward imports as the information
      can be left as is. In practice powerlines are much more easily
      added from Bing imagery than from OSGB VDM.
    * Woodland, numerous places by various people.
    * Waterbodies/rivers etc: Tendring (EdLoach), lots in Scotland
      (tms13), River Trent & lower reaches of Derwent (me)
    * VDM Building outlines: done at CenterParcs, Sherwood Forest
    * Buildings derived from OS StreetView (mapseg by TSC): various
      locations, including Preston, Surrey, West Bridgford area.

Known areas substantially traced from StreetView:

    * Bolton (steely)
    * Oldham (steely)
    * Middlesbrough (CGT)
    * Darlington
    * Grimsby & Cleethorpes (warofdreams)
    * substantial areas of S. Yorkshire (warofdreams)

Areas with names predominantly sourced from OS data, /inter alia/:

    * Oldham (done by me)
    * Darlington
    * most of NW London (Hayes, Hillingdon, Northolt ...) (done by me)
    * Liverpool suburbs (done by me)
    * Corby (and I presume many other areas in Northants)

Where I added names it was only to ways already traced beforehand (both 
from aerial imagery & from OSSV), and I usually did the minimum of 
alteration to what was already there.

Looking at each available class of data (sorry rather long):

*Boundaries*. Some don't like them for mainly reasons of 
policy/philosophy. Others (like me) see them as an integral part of a 
map. Main problem in importing them was merging with existing data (to 
retain things like relation history). Quite a lot of work and somewhat 
error prone, but substantially improved data as well. Experience shows 
that boundaries are *very* prone to be being broken in OSM. On the other 
hand I think that the process of using boundaries direct from shapefiles 
is beyond what most people feel willing or capable to do. So, OS data 
was relatively straightforward to use, somewhat tedious to integrate, 
convenient for *some *uses in OSM. I'd say pros & cons evenly balanced.

*Powerlines.* See chillly's blog on this. Major powerlines are 
conveniently provided as many short lines with a gap where the pylon is 
located. Really tedious to manage directly. Much, much easier to do from 
Bing data. Largely useless.

*Water. *Riverbanks and lakes as large scale features were generally 
better than existing OSM features, often because those had been traced 
from landsat/NPE maps etc. Requires significant post-processing to 
remove bridge outlines etc. (artefacts of the fact that VDM is a render 
dataset). See Mike Collinson's recent email for issues concerned with 
this. I have generally only used this on waterways I have recently 
encountered (mainly the Trent S &W of the end of Yahoo Aerial imagery). 
Some canals have been imported, as have the banks of smaller waterways. 
I'm not at all sure about this: primarily I would like to have base 
hydrography as a connected network with flows. Again render artefacts 
are a pain. Absence of flowlines in the data for many streams is a real 
pain. Generally the most useful of the VDM layers by a long chalk.

*Woodland*. In principle as for water, *but *suffers from a number of 
issues. The main one is that the woodland is subdivided into an absurd 
number of small parcels (some of this is a production artefact). 
Secondly is that perhaps 20% is not woodland, but scrub, small 
plantations (e.g., new planting on golf courses), foliage of individual 
trees in parkland, avenues, narrow wood belts, or erroneously 
interpreted (or coded) reedbeds, sedge-beds and other 'green' areas. 
Thirdly, woodland paths often separate woodland parcels even when the 
path is narrow and in summer canopy is closed over it. So the woodland 
data has to be treated with a great deal of caution. In general most 
useful as a hint as to where to look at Bing to trace woodland.

*Residential buildings.* Not useful at all, even though it has been 
imported in one or two places (see above).

*Other VDM layers.* I'm not aware that these have been used in earnest 
by anyone. In principle they could be used to identify major features 
which are still pretty inaccurate in OSM (e.g., a lot of motorways, 
which are often 10s of metres off GPS tracks, presumably because they 
were mapped early on). Fun to play with, but not hugely useful yet for OSM.

*OS StreetView.* I've used this to add names to pre-existing un-named 
streets on the basis that at least OSM data would then be more useful. I 
have only added streets by tracing OSSV to flesh out a set of roads I'd 
surveyed and only had stubs at their starting point (Buckingham). I do 
use it now to add streets when reported in a MapDust bug report. 
Generally useful for improving alignment of other features. Tracing from 
OSSV has more in common with tracing from aerial images than other 
sources (old maps, landsat etc): a lot of ground can be covered quickly. 
A lot of freshly traced OSSV stuff has never been followed up on the 
ground. My personal worst case was the 'Carlton incident' when a user 
realigned a lot of ground surveyed roads to OSSV: many have not been 
restored, and I was by no means the only mapper affected.

*Buildings from OSSV.* mapseg and similar. I don't regard these imports 
as successful. I did one and am now working through all the buildings 
redrawing them properly. I originally did it because I was finding 
adding addresses in the particular area to be a nightmare, and the Yahoo 
imagery was not really good enough for useful building outlines. The 
data look OK at zoom 15 & 16, its only when one gets to 17 & 18 that its 
clear that it's not good quality. In probability all such data will 
probably need to be re-done one way or another. mapseg also used the 
full new conversion rather than OSGB36 so is probably out of alignment 
with other data which used EPSG:27700.

*OS Locator*. Has been useful, both for adding names (for MapDust, short 
stub roads) and for quality assurance (ITO, OSM Musical Chairs).

*CodePoint. *Has become useful now that buildings are being added from 
Bing, particularly with Chris Hill's postcode tiles.

*Other issues:* VDM areas are obviously assembled from 1km by 1km tiles 
as the polygons all have boundaries on these lines. Projection and 
conversion issues across the OSM toolset, particularly from shapefiles 
to osm XML.

*Summary:* most useful things from OS have been names, and probably 
postcodes. Almost everything else has substantial drawbacks: as has 
become really obvious with the availability of Bing data. Apart from the 
deterrence (TIGER ?) effect  of apparently well-mapped areas, many 
mappers have invested signficant time into playing with OS OpenData 
which may otherwise have been directly used to map or add to OSM tools. 
Adding more data from OS OpenData has merely confirmed what we already 
knew: "Build and they will come" is not true.

Sorry for being so long-winded.

Jerry

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openstreetmap.org/pipermail/talk-gb/attachments/20110321/65300246/attachment.html>


More information about the Talk-GB mailing list