Limitations of the data on this site

The basic UPRN and USRN data on this site comes from two sources, the Ordnance Survey Open UPRN and Open USRN databases and the Office for National Statistics UPRN Directory. These are augmented by other OS Open Data products, including OpenNames and the Open Linked Identifiers which link UPRNs to USRNs.

However, there are some severe limitations on the data that is supplied. For UPRNs, all we get are numbers - the UPRN itself (which is a number from 1 to 12 digits long) and a pair of geographic coordinates in OSGB format. Something like this:

1,358263,172798

The ONS data contains a list of government statistical service (GSS) codes for each UPRN, allowing us to link it to various administrative geographies. A line from the ONS data looks something like this:

0009707683,E00062491,E99999999,E99999999,...
(not all codes shown, for the sake of space)

These codes can be linked to the names of various administrative geographies such as district and county councils, local authority wards and NHS regions. However, none of this tells us any more than could be derived independently from the geographic coordinates.

In particular, none of this tells us anything about the specific circumstances of an individual UPRN. It doesn't tell us which property it is associated with, or the address of that property.

In some cases, the identity of a property can be derived from a map, if there is only one UPRN assiged to a distinct property. For example, on this street, it's fairly clear that each UPRN belongs to a single property on the street. But in others, that's not the case. For example, if we look at UPRNs near this postcode, you can see that the Lidl supermarket has two different UPRNs on top of it. And this map also illustrates another issue with the open data UPRNs. Immediately to the east (right) of Lidl on that map, there are five UPRNs over what is open ground - look at the satellite image to see that. Those UPRNs are legacy identifiers for the buildings that used to be there before, but are now inactive. But they're still in the open data release.

Other places have different issues. Here, for example, there appears to be only one UPRN on the Royal Mail depot. But, in fact, there are 71 all at precisely the same coordinates - they are all the different "large user" PO Box addresses that are serviced from that depot. And the same problem occurs with blocks of flats - each flat has its own UPRN, so where a multi-tenanted residential building has multiple floors then it will have mutiple UPRNs stacked on top of each other. There is no way, from the published open data, of distinguishing between them.

For USRNs (streets), we get a little more. But not a lot more. As a street is a line, not a point, we get a set of coordinates that make up a line (or a set of lines), together with a "street type" field. Here's a line from the USRN database:

9,[GEOMETRY - 266B],27001252,Designated Street Name

What we don't get here is an actual name. Which is important, because most of us refer to streets by name rather than number.

The OS Linked Identifiers Database (LIDS) also enables us to link UPRNs with an associated USRN. Again, this is just a pair of numbers:

100030148369,14000313

This allows us to see which USRN a UPRN is linked to, and also find all the UPRNs that are linked to any given USRN. That's useful. But it's also unreliable. In some cases, such as this one, all the UPRNs associated with a street do, in fact, seem to be on that street. But, in other cases, such as this one, as well as UPRNs that are actually on the street in question there are also others that are on other streets close to it. And at this rural location, the URPNs associated with the USRN are all over the place, and some have no connection to the street at all.

Even as it stands, therefore, the open data releases of UPRNs and USRNs are not 100% reliable. There are clear and obvious errors in the data, and we can only hope that these are corrected with later releases. In particular, legacy UPRNs need to be removed from the data, and the UPRN/USRN linking needs to be more accurate. But we also need more information about UPRNs themselves, for example a "z-index" value for when they are stacked on top of each other, such as a block of flats, is essential.

The lack of street names in the USRN database is also unhelpful, especially given that there are also known errors in the non-open National Street Gazetteer of which USRNs are a part. The only way to get street names from an open data source is to try to match up the USRN coordinates to roads and streets in OS LocalMap Open and Open Streetmap. This works some of the time, but not all of it. It would be more useful to have the names directly in the USRN open data release. Or, better still, make the entire NSG open data.

Rather than go into that in detail here, I've written a separate blog article about that on my general website. If you agree with me, please lobby your local councillors and your MP to get their support for further opening up of the NSG as open data.

uprn.uk is a Good Stuff website

3.238.184.78 Mon, 23 Nov 2020 19:11:25 +0000