[Talk-GB] Composite mapping (OSM and OS, PRoWs etc)

Wed Sep 7 14:29:30 UTC 2016

First, let me thank you for all your comments. They’ve been very helpful
indeed.

Second, let me apologise for the length of this reply.

Dan S:
> Looking good! Is there a key? Would help to make sense of the various
> footpath markings etc

Automating the legend was one of those jobs I kept putting off. I’ve just
pushed an update that provides a basic legend; there’s still more to add.
You can find the legend at http://geo.gy/legend/.

Robert Whittaker:
> That does look really nice. I particularly like the long-distance
> cycling and walking routes being shown. Maybe decrease the dot spacing
> a bit at higher zooms though.
> …
> Are Public Rights of Way actually shown with specific markings,
> separate from other paths/tracks with access tags set? I couldn't seem
> to see anything obvious to distinguish them in the areas I just looked
> at.
> …

Good points. A few people have mentioned the access markings, and it’s
clear we need to rethink these. I will review these, and add new access
categories that distinguish between a legal public right of way, and a
maintained footway. I’ll also separate bridleways and cycle paths, add a
category for Scottish core paths, and add permissive bridleways.

The current site is just intended as a way of accessing the maps for
preview purposes. I’ll make the search accept place names shortly, and
allow links to a specific grid reference. Using the History API I should be
able to update the URL as you pan around the map, but I can also add a grid
reference readout.

Dave F:
> Could you add a permalink, primarily so any specific queries/problems
> can be clearly identified?

Sorry Dave, not sure what you mean? Do you mean the ability to link to a
specific location on the map (lat/lon or grid ref)? If so sure, I’ll make
that happen shortly.

> It appears to render OS filed boundaries as grey short dashed lines. Is
> this intentional?

Yeah I appreciate the dashed lines are a bit odd. These aren’t actually
taken from OS maps in the majority of cases. We processed the 2m LiDAR data
from the EA and NRW, and used these to infer hedgerows and walls.
Unfortunately as you noticed it’s not perfect, and worked better in some
regions than others. I want to revisit this and try improve it by using
their raw point cloud data.

In areas where OSM has field boundaries, the system throws away our LiDAR
walls and uses OSM’s instead. The intention of the grey dashes was to
suggest there’s a surface feature rather than definitely a wall, because
from the LiDAR data we can’t be certain. I’ll come up with another way of
showing them that’s less confusing.

> As a by-product of the PROW comparison, would it be possible to create a
> list (Geojson?) of paths not in OSM but in LA's definitive maps? It's
> something I always thought would be helpful to complete PROWS.

I’m happy to do whatever I can to help out in that respect, but Robert
Whittaker’s system seems better suited. I don’t know if it’s helpful or
not, but you’ll find the code I use for pulling in PRoW data in all sorts
of different formats at [1], with some extra files for the awkward
authorities that used numeric codes in their data etc. It pulls in the
files then attempts to classify them automatically using common codes and
abbreviations. Might help Robert keep his tool up to date?

SK53:
> With PostGIS I have run into various problems including non-noded
> intersection errors in the latter steps.

PostGIS and I now have a difficult relationship at best. Processing the
whole country (with lots of ST_MakeValid calls and the like) to match all
the PRoWs does take several hours. One thing we also do is try to deal with
partial matches of the paths, so a new segment will be added only when OSM
doesn’t already have a path stored, otherwise the legal status is updated
instead.

> Churches
Good point. I haven’t yet extended the building schema we use to
accommodate ‘special’ buildings, but it’s on the plan. That would allow us
to show a thicker border or different colour for significant buildings in a
community, including churches. From a cartographic standpoint, there’s a
system in place to prioritise some features, and aggregate features if
they’re nearby into a single plural label. Churches would definitely be
more helpful in navigation than memorials.

> Access land
I’ll look into this. The data from Natural England is a bit dodgy in
places, with very very small gaps that can’t be shown on a 1:25,000 scale
map. We remove some of the detail when we import it, but maybe we removed
too much if these narrow patches are missing. One of the things I want to
look at is moving features that are near to roads slightly.

> Grid lines
Good idea. I’ll make the 10km lines thicker.

> PRoWs through gardens and so on
Good points. My intention is to use OSM as the base for highways always, so
if we match a local authority’s PRoW data against a path in OSM, the
geometry will be taken from OSM and maybe we’ll indicate the discrepancy
between the real path and the legal status in some other way.

Richard Fairhurst:
> The major thing you need to look at, I would suggest, is access tagging.

Absolutely. One thing I am concerned about is making it too complex (i.e.
having to consult the legend all the time), but it’s clear we need to
improve on the access tagging we have at the minute. Watch this space.

Paul Berry:
> It's a great effort; please keep at it!

Thanks – hopefully we’ll keep updating this once every few weeks.
Unfortunately we can’t update as often as OSM does, because of all the data
matching.

Ed Loach:
> both the correct route for the Essex Way (as in OSM, here-ish
http://osm.org/go/0EHx9iB1-- )
> where it has been diverted north and south of where it passes under the
railway, and also show
> the pre-diverted footpaths (not in OSM) which used to cross the fields
diagonally

I’ll look into this. The official data I have downloaded might be old. Over
time I want to move to also comparing timestamps on OSM updates, to deal
with these changes more sensibly. The same issues likely apply to
buildings, because OS OpenMap hasn’t been updated since 2015 so could have
reinstated some demolished buildings.

> PS: I just tried searching for the grid reference without spaces, and
then it seemed to work.

Oops. Fixed now.

Thanks once again for all your comments.

Cheers

Luke

[1]
https://github.com/lukessmith/grough-map/blob/master/bin/linux/gm-import-prow.sh

On Wed, Sep 7, 2016 at 11:02 AM, Ed Loach <edloach at gmail.com> wrote:

> I’m not sure which of your data sources give which bit of your rendering,
> but at about TM 151 312 (and I can’t get search by grid reference to find
> this location – slightly NE of Bradfield, Essex) you have both the correct
> route for the Essex Way (as in OSM, here-ish http://osm.org/go/0EHx9iB1--
> ) where it has been diverted north and south of where it passes under the
> railway, and also show the pre-diverted footpaths (not in OSM) which used
> to cross the fields diagonally.
>
>
>
> But generally I think it looks great.
>
>
>
> Ed
>
>
>
> PS: I just tried searching for the grid reference without spaces, and then
> it seemed to work.
>
>
>
> *From:* Luke Smith [mailto:luke.smith at grough.co.uk]
> *Sent:* 06 September 2016 16:03
> *To:* talk-gb at openstreetmap.org
> *Subject:* [Talk-GB] Composite mapping (OSM and OS, PRoWs etc)
>
>
>
> I mentioned a while back that grough was developing a composite map,
> blending OSM data with OS OpenData to fill in the gaps, and using public
> rights of way data directly from the local authorities which have released
> it. Over time, hopefully we will rely progressively less on other data
> sources.
>
>
>
> I'm happy to say there's now a beta available, at http://geo.gy/ with
> more details about the project at http://map.grough.co.uk/.
>
>
>
> There'll also soon be a 3D version available, building on the prototype at
> http://3d.geo.gy/ to cover all of Great Britain and improve the controls.
>
>
>
> The source code behind generating the maps is open source, although not
> suitable for on-the-fly tile generation because of the preprocessing. The
> idea was to create a map which could be printed and used at a fixed scale
> (1:25,000 scale), with labels moved around to avoid obscuring detail etc.
>
>
>
> If anyone has comments or advice for us, it would be gratefully received.
> We're aware of some issues already, so this is only a beta release.
> Similarly if anyone would like to use the maps, we'd be more than happy to
> help if you run into problems.
>
>
>
> Luke
>
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openstreetmap.org/pipermail/talk-gb/attachments/20160907/1952237c/attachment.html>