[Talk-GB] The curious case of USRN 20602512

Mark Goodge mark at good-stuff.co.uk
Fri Jul 10 10:27:24 UTC 2020


Apologies for the long read, but this may be interesting to some folk. 
This follows on from my earlier response to Kai Michael Poppe about 
"Fairfield Road" in Ealing.

On 04/07/2020 12:02, I wrote:
> 
> To find the USRN of the path, you need to use the lookup tables supplied 
> by OS. Doing that, we find that the associated USRN is 20602512.
> 
> Now, there's no open data source which will directly tell you the name 
> of a USRN (at least, not until we start putting them into OSM). The long 
> way of doing so is to find the matching LineString in OS OpenMap Local, 
> and see what name it has there.
> 
> However, it can be done directly via a non-open source. If you go to 
> https://www.findmystreet.co.uk/map and zoom in on the location, then 
> click the street to bring up the USRN details, it will give the name 
> (and also confirm that the USRN from the OS lookup table is correct). Or 
> use the search box and search for USRN 20602512.
> 
>  From an OSM point of view, that would normally be a dead end. Even if 
> you can view the information on a non-open source, you can't incorporate 
> it into OSM. However, in this case, we already have an abbreviated name 
> from an open source. So all we are learning from the closed source is 
> the full text of the abbreviation. Whether that makes it acceptable to 
> include the full name into OSM is a matter of debate. I'll leave that 
> decision up to others, but, for reference, the name of the street is 
> Fairfield Road.

I've been doing a bit more research in this, as it piqued my interest. 
And the results are a little surprising.

For a start, USRN 20602512 doesn't match Fairfield Road in OS LocalMap 
Open. In fact, there's no Fairfield Road anywhere near there in OSLMO. 
Matching the coordinates indicates that, as far as OS is concerned, it's 
a part of Southdown Avenue. That's not particularly unusual, access 
roads off named streets often don't have a name of their own, they're 
either completely unnamed or share the name of their parent street.

However, I did wonder whether this might just be a limitation on OS Open 
Data, and whether MasterMap might actually include the name. That's not 
reusable in OSM, of course, but it might help point to an open source 
that does contain it.

But it seems that even MasterMap doesn't have that name. You can check 
that by looking at Ealing's online GIS website:

http://maps.ealing.gov.uk/Webreports/Planning/Planning.html

This is a planning application map, but it's just a window into their 
GIS system and you can turn off the planning layers. Anyway, zoom all 
the way in to the street in question - I can't give you a persistent 
link, but it's just above the LA boundary in the bottom middle of the 
map - and... it still has no name. At the highest zoom level, this is 
MasterMap, and every named object has its name displayed. But there's no 
name here.

Google, also, knows nothing of a Fairfield Road here. Using the Maps API 
to query the coordinates of USRN 20602512, we either get Southdown 
Avenue, again, or Boston Gardens, which is the postal address of 
buildings facing Boston Road. You can see that name on the road sign via 
Google Streetview:

https://goo.gl/maps/KGLbRC75mQw43PCV6

So, it seems that Fairfield Gardens isn't known to either OS or Google. 
It is shown (in abbreviated form) on streetmap.co.uk, but at that zoom 
level, in London, that's based on the Bartholomew A-Z maps rather than OS.

Given that, we can't include the name "Fairfield Road" in OSM as it's 
only available from non-open sources. But even those non-open sources 
don't agree on the name. That seems to me to lead to two possibilities:

1. It doesn't exist at all. It's just a map trap designed to catch out 
unwary copyright infringers. That's certainly a possibility, and A-Z 
maps are known to use those. But that doesn't explain its presence in 
the USRN database.

2. The USRN name is wrong, but that error has propagated to the A-Z maps.

Personally, I think that the second option is the most likely. And, if 
so, it wouldn't be the only error in USRN. One of the things I had to 
deal with a few years ago, in my capacity as a district councillor, was 
a country lane in my ward that had the wrong name assigned to it in 
USRN. After a bit of investigation, we concluded that it had simply been 
a transcription error back in the late 90s when the local gazetteer was 
first digitised, but it had gone unnoticed for a couple of decades 
simply because the wrong name never appeared anywhere in public until it 
eventually cropped up on a planning application. Getting the name 
corrected wasn't an easy task, because of the length of time it had been 
wrongly recorded, but we did eventually manage to get it sorted out and 
the correct, historic name of the lane assigned to the USRN.

But that's not the only one. The USRNs for where I grew up, out in the 
middle of the countryside in West Suffolk, are listed as either "Poultry 
Road" or "Sedge Fen" in the USRN database. You can see that by looking 
at https://www.findmystreet.co.uk/map and searching for 14610160. But, 
in reality, the road name is "Sedgefen Road", and this is correctly 
shown by MasterMap, Google and OS LocalMap Open.

This is, of course, one of the problems with proprietary data. It can be 
difficult to spot errors, because the people who are most likely to spot 
errors - members of the general public with local knowledge - tend not 
to have easy access to the data.

So this is a bit of a warning, really, for the open mapping community. 
Although the open data release of USRN ids and coordinates is welcome, 
don't be tempted to look up street names on the street list published, 
with a restrictive licence, on https://www.findmystreet.co.uk and then 
copy them to our own data. Because it simply isn't reliable enough as a 
guide to actual usage, even if it is what the "official" name of the 
road may be. Stick to OS Open Data and local knowledge.

Mark




More information about the Talk-GB mailing list