Thanks for the interest and comments on my post about estimating coverage. In answer to some of the questions:
The way I have handled dual carriageways (and motorways) is to assume that both carriageways are plotted separately on OSM. So the figure that results should be twice the length estimated by DfT. However, for primary roads, DfT themselves show the total length of primary road, and the length of this which is dual carriageway. So by adding their two numbers, I effectively double up their figure for dual carriageways, to reach the same (in principle) as the OSM total. Hence I can ignore "dual carriageway" tags on the OSM stuff. I think. Non-adopted roads, service roads, etc. Are certainly a potential source of error. Of course I am ignoring stuff tagged footpath, bridleway, cyclepath and the like. However I am not sure how DfT count slip roads, roundabouts, and the like, which would all add to the OSM total, but not necessarily theirs. At the moment I am ignoring stuff that is tagged "service road". Of course there may be all sorts of roads that are not counted by DfT that I cannot exclude. The question, I suppose is whether this makes a significant difference, or not. However, there are about 180 different categories of highway in the UK map. This initial attempt takes a rough cut, until I work out what is important. My initial guess is that outside the main categories few if any will make a significant difference to the totals. On reflection I probably shouldn't have included "unsurfaced". I am trimming ways at the admin boundary. I wasn't originally, and it makes quite a big difference, particularly on the figures for primary roads. There are also quite a lot of local government boundaries running along major roads, and my guess is that in some places the precise positioning of the boundary makes quite a big difference. I'm not sure that boundary plotting is accurate enough yet to classify these marginal areas properly. I have uploaded a summary of my view of the current status of relations for admin boundaries here - http://www.reedhome.org.uk/Documents/LGboundaries.csv As you will see this table makes a distinction between the status in my extract database, and the current state of the OSM database. I extracted my data about a week ago. My main emphasis is on England, but I will try to add Scotland, Wales and N. Ireland in due course. The main database has moved on since I did the last extract so I've also run a quick check on that to show how it currently seems to stand. The process I used is slightly different, so there may be some inconsistencies. This list also shows English authorities (down to district) that I cannot find, and whether or not relations I can find are complete. Where I know they are the ceremonial, not the admin county I have marked them as such. I've also started to fix some of these myself in the main database by adding a new "county council" boundary - leaving the ceremonial one in place. I've also uploaded my summary of road length estimates here - http://www.reedhome.org.uk/Documents/Roadlengths.csv The "Other" column gives an indication of how difficult I found it to classify the different types of road. Most of this is just tagged "road" but quite a lot is "unsurfaced", and should probably be excluded. The differences between different types of road are more significant than the totals. They may highlight some inconsistencies in tagging, and the may be of interest to people looking at an individual authority. At this stage, though, I would like to concentrate first on getting rid of anomalies in the overall calculation. Generally the figure for motorway is pretty close. I suspect that significant differences in the motorway figure are down to errors in the boundary position. It should be possible to compare adjacent authorities to see where these have resulted in a motorway appearing in the wrong authority - but this is on my ToDo list for a later stage. Note also the variations in the area shown on OSM and by government land use statistics. I think some of these will be due to areas of sea / estuary, etc included within the OSM boundary, but not land-use stats. Others are likely to highlight errors in the OSM boundary - particularly where I've picked up a Ceremonial county boundary without realising it. For what it's worth, my thoughts on tagging boundaries are: * Regions are a higher level of admin boundary in England, falling between "4" for England and "6" for "County". So they should have their own level of "5" * A Ceremonial boundary is not conceptually the same as an Admin Boundary, and should have its own relation, even if this is following the same line segments, in the same way as a district and county often share a common boundary * To group similar relations they should be given a common tag - e.g. <Local_Authority_type=London Borough>, not their own common relation On moving the coverage estimates forward, I'm in for a couple of busy weeks at work, so I'm not likely to revisit the actual estimates for a few days. But once things go quiet again I'll take a new extract to pick up any improvements to the boundaries, and try to refine the calculations on the basis of all the help I'm getting. Thanks again for all the interest. _______________________________________________ Talk-GB mailing list Talk-GB@openstreetmap.org http://lists.openstreetmap.org/listinfo/talk-gb
