Just over a year ago I started running daily street name completeness checks for Switzerland based on a list of street names by municipality generated out of the federal “Gebäude und Wohnungsregister” (GWR), see my original article for more information and http://qa.poole.ch/ch-roads/ for the daily updates.

For technical reasons I decided to move the contents to a new server late December and during the process I’ve made a couple of updates and changes that need some explaining. On the one hand I’ve updated the GWR list to the December 2013 one on the other hand I’ve somewhat changed the logic of the road (contrary to other object types with names) statistic generation. One of the more annoying trends in the GWR list is that more and more municipalities are no longer correctly filing the object type in their submissions and are either using “unknown” or leaving the field empty (which I map to “none” in my statistics”), despite the large amount of building going on, the number has actually gone down in absolute terms by roughly 500 over the year 2013. This naturally makes the data substantially less useful for us, and I would go as far as saying it makes the data less useful for its primary purpose too.

However it is clearly not our job to discipline such behaviour, we just want as good as possible estimates of how many named streets there actually are. To achieve that I’ve now added some heuristics to take this undesirable behaviour in to account:

• assume that if a GWR object has either no geometry type or “unknown” and the corresponding OSM object is a road, it should have been a road in the GWR too.
• if there are no roads at all for the municipality in question in the GWR and OSM has roads as described above, add all the relevant GWR objects to the GWR road count.

The GWR numbers reported are corrected correspondingly. As of today this reduced the object count for “unknown” by 8’374 and for no object type by 4’438, adding a total of 12’812 to the GWR road count. For continuity and documentation purposes the old statistics are still calculated daily and are still available.

More interesting than the above changes is how much progress we have made over the past year. In the following table non-road objects are summarised in one number.

As can be seen from the above we managed to add 20’024 names in just over a year, confirming that we should have as good as possible coverage of street names latest in two years.

To conclude, and most impressive, a map from early 2013, and current versions using the old and new logic

January 2013

December 2013 Old

December 2013 New

Since the beginning of November 2012 I have been running statistics on how OSM data compares to “official” sources. The results can be seen qa.poole.ch/ch-roads/  a wiki page with some more information is available here.

The official numbers as of December 2012 contain 108’271 road names, 2’684 named nodes, 27’339 areas and a total of 52’211 of objects that were not classified by the muncipality collecting the data. Overall this amounts to 190’000 named features.

The Swiss OSM extract as of January the 10th 2013 contained 115’940 named streets, 5’053 areas and  9’674 nodes tagged with “place” and a name. The comparision just on a number base looks quite good, however actual matches are only 84’162  in total as of today, which is less than 50%. Just considering the streets we find 67’125 matches which is  61%.

The good news is that in the roughly 10 weeks the statistics have been running, we increased  the street matches by 3’556 and the total by 5’783. This indicates that we should have all roads covered in latest two years. I actually suspect that the coverage increase will be far larger in the summer months and that we will be able to acheive very good coverage far earlier.

Achieving rapid progress on non-road objects is however going to be difficult. Most of these are not sign posted and publicly available information for verification or a possible import seems to be limited to a proprietary dataset owned by the Swiss Post and the GWR address data. Both can be purchased at considerable expense, the later however has very restrictive usage terms. Essentially Swiss Topo has been granted a monopoly on all non-internal use. It is strange that in the age of the OpenData movement that this would still be possible (the legislation is quite recent), but I suspect that it will be very hard to change.

A good example of the difficulties is the small mountain village Ftan where we customarily spend our annual skiing vacation.

Essentially none of the roads have road signs, nor do they actually have names. The full addresses use a “quarter” name in place of a street name. As can be seen from the list produced by my statistics (original file: http://qa.poole.ch/ch-roads/ftan-20130119.html current version: http://qa.poole.ch/ch-roads/GR/3761.html switch to the Mapquest Open map on the OSM site to see what was actually added) nearly all the missing features are not roads, that dosen’t stop Google showing street names (google maps) where there aren’t any.

Surveying such data is going to take time, it implies using local knowledge and other clues available on the ground. It will be interesting to see how much can be gathered in the week I will be on site and how we progress with the multitude of similar settlements over the next months.

The last numbers on development of the OSM coverage in Switzerland were from early 2012. Since then we have had on the one hand the licence change process with a small corresponding loss of data, and on the other hand quite a lot of interest in the project as a whole. So I beleive an update is in order. While I didn’t use quite the same methodology as the previous numbers presented in the wiki here, they seem close enough to be comparable.

The statistics show an increase of a total of at least 10’000 km from 142’000 km at the beginning of the year to 152’000 km now. This without taking a further 8’000 km of “service” roads in to account that were not listed separately in previous statistics. As would be expected the length of all roads with higher classification does not show much movement, a clear indication that the major road network is very complete.

The low number for the length of combined footway/cycleway would seem to indicate that we are mistagging the most frequent occurring type of cycleway.in Switzerland. What is further slightly surprising is that we have mapped a combined length of 1500 km of driveways and parking aisles, this would seem to be an awful lot compared to roughly 1300 km of cycleways.

Detailed Numbers