[Talk-GB] UPRN to postcode lookup

SK53 sk53.osm at gmail.com
Thu Jul 1 21:54:04 UTC 2021


Yes, I've played with it and it helps with edge cases where the postcode is
not obvious. I'm very chary about the implicit intellectual property
rights: there was a link recently tweeted by Peter Wells regarding IPR
associated with UPRNs, but this may be outdated as it was back in 2016.

Note that the UPRN->Postcode applies to non-addressable objects too which
is a little odd.

The main thing which the lookup assists with is correctly associating a
property both to a postcode & a street. This helps when UPRNs within a
street are a straight sequence of numbers ordered from low housenumber to
high housenumber (which is true of most properties which existed before
UPRNs were created & partially true for those afterward: Conwy, Gedling,
Nottingham, Wirral are areas I've tested this). Here is an example
<https://wiki.openstreetmap.org/wiki/File:Stoneby_proxy_address.png> for a
single street in New Brighton, Wirral where I've taken all UPRNs matching
the two postcodes, ordered them & assigned proxy house numbers solely on
the basis of that sequence. You can compare this with actual even house
numbers which I had previously surveyed. I have a number of other examples
where I've used the UPRN sequence to infer housenumbers. All done by hand
and in most cases I know either one or more house numbers already or am
familiar with typical numbering patterns in the area. I'm currently playing
with Wirral data as a testbed (& have the UPRN->PC lookups too).: I have
done some desultory surveying around Wallasey, two Birkenhead stations,
Hoylake & West Kirby. In addition the old Naptan data often has
housenumbers in the indicator field (true for much of the Merseyside
transport area & also for Greater Manchester). These existing addresses
provide a useful cross-check against those inferred from UPRNs.

My rough plan for an automated methodology::

* assign UPRNs to roads (this is way easier with the lookups, otherwise one
needs to pull more UPRNs than necessary & filter them out). Only UPRNs
which coincide with an OS Open Local building are used (to exclude
substations.post boxes etc). UPRNs in very small INSPIRE parcels could also
be potentially exlcuded. Note most (probably all) of the non-address UPRNs
are not contained in the sequence runs for a road anyway.
* find all sequences with 3 or more UPRNs for the road
* crudely assign proxy addresses directly according to the sequence
* refinements tbd: a) multiple UPRNs at one location (use only the first,
adjust sequence; >3 not usable); b) expect zigzag patterns in a vector from
the sequence of even/odds, divergences indicate missing/additional
numbers); c) cross-check with other available open data (these can be used
to anchor proxy numbers).

Any mapping I've done uses an entirely manual version of this, and I have
multiple examples where I've been able to solve an irritating issue which I
failed to notice when actually surveying.

TLDR: they are potentially very useful, but advice from ODI, Peter Wells or
Owen should be sought.

Jerry


On Thu, 1 Jul 2021 at 21:06, Rob Nickerson <rob.j.nickerson at gmail.com>
wrote:

> Hi all,
>
> I was reading the latest (always detailed and helpful) blog post by Owen
> Boswarva today on addresses [1]. One thing that it mentioned that I was not
> aware of was that there is a lookup between Unique Property Reference
> Numbers (UPRNs) and Postcodes:
>
> "National datasets of UPRNs matched to postcode units and statistical
> geographies are available as open data in ONS's National Statistics
> Address Lookup
> <https://geoportal.statistics.gov.uk/search?collection=Dataset&sort=name&tags=all(PRD_NSUL)>
> (NSUL) and ONS UPRN Directory
> <https://geoportal.statistics.gov.uk/search?collection=Dataset&sort=name&tags=all(PRD_ONSUD)>
> (ONSUD). From late 2020 those datasets also associate UPRNs with postcodes,
> enabling the calculation of notional geographic extents for individual
> postcode units."
>
> Has anyone had a look at this? Do we think the licence is sufficiently
> "open" for OpenStreetMap use?
>
> To save you downloading a huge file, I have downloaded the April 2021 NSUL
> and extracted 10 entries from the North East along with the manual. This
> can be downloaded from the OSM UK site [2]. I also have a bash script for
> processing the data and am happy to share that or the resultant file
> (Lat/Lon/Postcode csv) with you.
>
> [1] https://www.owenboswarva.com/blog/post-addr3.htm
> [2] https://osmuk.org/wp-content/uploads/2021/06/NSUL_Apr_2021_extract.zip
>
> Thank you,
> *Rob*
> _______________________________________________
> Talk-GB mailing list
> Talk-GB at openstreetmap.org
> https://lists.openstreetmap.org/listinfo/talk-gb
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openstreetmap.org/pipermail/talk-gb/attachments/20210701/b04b5610/attachment-0001.htm>


More information about the Talk-GB mailing list