OpenStreetMap

First area mapping from survey

Posted by Rostranimin on 16 April 2012 in English (English)

I'd be very happy to receive comments on some new mapping I've been doing. West Kilbride in Ayrshire has a piece of parkland/woodland which is of significance to the town. It's difficult or impossible to map much of this from Bing images because of tree cover so I carried out a survey using two GPS devices and repeated trips along the main walkway/paths. The GPS traces give me a good shape of the paths, but there are significant errors in exact path location (obvious from variation in path traces for the same path). Have tried to use the traces, Bing imagery, and the OS opendata info for the stream path. There are errors or misalignment between all the sources of data. The Bing images are out of date, the OS opendata is well out of date, the GPS traces aren't very accurate. Clearly this makes for a bit of a challenge.

My approach has been to do the best I can in order to get this area mapped. There are outstanding questions which need to be answered - does a path actually have this shape - has the stream moved since the OS data was created - how exactly should we deal with the area where a woodland merges into a garden - what's park and what's just grass - and so on. I also know that there are areas I need to revisit to check my information - but this won't be something I'll be doing soon.

So as I say - comments welcome.

Comment from EdLoach on 16 April 2012 at 14:56

If so, then it all looks OK from a quick glance, except perhaps the natural=grass area, which perhaps should be landuse=grass

Hide this comment

Comment from daveemtb on 16 April 2012 at 15:04

Your approach sounds good to me. The more GPS tracks you have the better the guess you can take about where the real path lies. I use the various other sources as an aid to interpreting GPS tracks.

Hide this comment

Comment from Rostranimin on 16 April 2012 at 16:04

EdLoach: Yes you have the right area. You'll see (on the updated Mapnik rendering for instance) the green area running north-south through the town. I've corrected the natural=grass to landuse=grass which had just been a slip.

Anyone think that this kind of area should be left until it can be surveyed properly? It feels inadequate to be using out of date information and weak GPS traces. Where will the original OS data have come from for this kind of area? Did people do really accurate surveys with proper (pre-gps) surveying equipment? How long ago is this likely to have happened - am I right to assume it may have been a very long time ago, later to be copied from map to map??

I'm intending to return to this area another time to try to see if my mapping makes sense as I walk the paths. As a map-user I know I look for features like how a path gets closer or further away from the edge of a woodland, or closer and further away from the main stream, or how the stream or path bends left and right. These are probably more significant for a map-user than the actual exact geographical location of these features.

Hide this comment

Comment from Pink Duck on 16 April 2012 at 16:11

As an aside, OSM’s OS OpenData StreetView tiles are currently years out of date. They could be freely updated if someone has access and knows how to do so.

Hide this comment

Comment from Hawkeye on 16 April 2012 at 18:34

Not sure how the StreetView data is collected but I think it is largely based on aerial imagery. I guess the answer is to wait for the new european Galileo satellites which will give sub-1m accuracy. But that could be 2019.

Or a bit of balloon aerial mapping?

I would say GPS would be okay for this type of surveying. Maybe adding a note= tag that it is approximate.

Hide this comment

Comment from z-dude on 17 April 2012 at 03:29

Mapping before the 70's was done with aerial imagery.

Airplanes also used radio beams from base stations to aid in keeping a straight path. http://en.wikipedia.org/wiki/Lorenz_beam http://en.wikipedia.org/wiki/Radio_navigation

Apparently, sextants were accurate to 200 meters. http://en.wikipedia.org/wiki/Sextant

Hide this comment

Comment from Central America on 17 April 2012 at 09:27

Looks really good, well done. Depending on how you want it to look you could also consider

barrier = hedge

and

leisure = track

this last one may require a multipolygon

There is a nice example of Queens park in Glasgow that was mapped by a mapping party a while ago.

http://www.openstreetmap.org/?lat=55.8312&lon=-4.26877&zoom=17&layers=M

Streams are always difficult because they change. In JOSM there is a validation tool which would point out that the stream should have layer=-1 and the bridges, layer=1

Do you use JOSM?

I think you did a really good job, I hope that you enjoyed doing it

Hide this comment

Comment from Rostranimin on 17 April 2012 at 10:03

Interesting stuff...

I was using the OS VectorMapDistrict vector information. Simple way to get data but I chose to use it to create the main stream way rather than using the riverbank data - this gives the illusion of accuracy even more.

I've assumed that this is the same data as on the Opendata tiles?

Surely streams like the one I was mapping couldn't easily have been done from aerial imagery. The trees - old ones on the whole I think - obscure the stream. I suppose with a low enough flight maybe?? It's the fact that the data is of the stream bank not just of the stream route which made me think that it was done accurately - and that small bridges which are completely invisible in Bing aerial images are included.

Central America - you are right, adding the hedges and walls will be good. I want to do a second check of these before I add them though as I'd like to include any gaps in them. Couldn't work out a way to tag a ruined wall with multiple gaps either. I'd used leisure=track previously for this same running track too - but the recent survey convinced me that what looks like a running track is in such bad repair as to make it unlikely that it's actually used properly any more (so I changed it to a simple track). The old tennis courts (middle rectangle) and old football pitch (south rectangle) likewise. I tagged these as leisure=pitch rather reluctanctly. Interestingly I've just Googled briefly for information and council minutes seem to confirm all old facilities except the grass football pitch (north rectangle) are unused. Is there some way to tag 'unused' or 'disrepair'?

Is it all of a stream which should be layer=-1 or just the bit under the bridge?

Hide this comment

Comment from LivingWithDragons on 17 April 2012 at 12:54

OpenStreetMap is never going to be accurate enough for legal use or blind-robot-navigation. Going for a walk with the map is a good test, are there curves on the map where you walk in a curve, are the path junctions roughly the right distance apart, etc.

Aerial imagery (Bing) will be offset by a certain amount, depending how far way in the world you are from their reference point. In the JOSM editor for OpenStreetMap, you can define an offset. So I normally go to a nearby place with no tree cover or tall buildings stand still for 3mins till my GPS is accurate to 12ft or so, and then I walk round some obvious land marks. Like road junctions. Then at your computer you can define the imagery offset so that the imagery roads match up with your GPS lines.

Sometimes you can tell from imagery where the stream is, because the tree cover looks slightly different (less dense, or it dips down). Other times you just have to remember roughly what line it took in relation to the path you walked on. Really the importance with streams is what side of the path they are on. They can change a bit over years anyway. They might never again be mapped by someone so focused on details as you.

Hide this comment

Comment from Central America on 17 April 2012 at 13:20

I tend to use the layer = -1 on all waterways, plus bridge layer=1. I originally got the idea from one of the german diary entries. It doesn't affect the rendering, or routing for waterways. If a highway crosses a stream or river and it they don't have different levels then it can affect the routing checking for highways though. I suppose it should be just layer = -1 for under the bridge, maybe I have just been lazy.

I see from the map features page there is a disused= yes tag

http://wiki.openstreetmap.org/wiki/Map_Features

I use barrier=wall for ruins, it renders really well

I did loch leven castle that way http://www.openstreetmap.org/?lat=56.200142&lon=-3.390579&zoom=18&layers=M

Hide this comment

Comment from Zartbitter on 17 April 2012 at 15:56

Streams should ghave no layer=-1 if they are not underground. But brides need to have layer=1 if they are crossing the stream.

Using layer=-1 on whole streams is imho a bad attitude of lazy mappers to avoid error reports of quality assurance tools (QA) if you forgot to mark a bridge on a street crossing this stream. Without layer=-1 QA can find errors where you just forgot a bridge. With layer=-1 these errors will probably be found only by chance.

Hide this comment

Comment from netman55 on 19 April 2012 at 15:34

With paths in woodland, surveying in winter is better than summer.

On my garmin, I switch gps to satellite view which gives an estimate accuracy figure. Start just outside woodland area. Stop walking, turn facing south and let signal stablise then waymark on best accuracy figure. Walk further 10m/strides whatever along path to be traced, then stop, turn south let signal stablise again then waymark, move another 10m/strides whatever, then stop etc, etc until path is completed. If at the same time you note where the signal is weakest you will know where your main problems are going to be.

For very bad signal spots if you can, walk 90 degrees to the path to a stronger signal point then waymark, turn around walk back across your path you are tracing, still at 90 degrees on the other side of the path at roughly the same distance then waymark again. Hopefully you path will be somewhere between the two waymarks.

If you think the above takes for ever, yes it does!

Another way is to stick a BT gps receiver on a very long pole :-)

Hide this comment

Leave a comment

Parsed with Markdown

  • Headings

    # Heading
    ## Subheading

  • Unordered list

    * First item
    * Second item

  • Ordered list

    1. First item
    2. Second item

  • Link

    [Text](URL)
  • Image

    ![Alt text](URL)

Login to leave a comment