[Talk-GB] Postcodes

Adam Snape adam.c.snape at gmail.com
Fri Nov 9 17:07:43 UTC 2018


Hi

To clarify the question I was asking earlier, this is what the OS say:

"Code-Point Open is created by taking the average of the coordinates of all
the individual addresses in a postcode (provided we have any of sufficient
quality), then snapping to the nearest of those addresses. Code-Point Open
then delivers the coordinates of that address, as representative of the
whole postcode, to a resolution of 1 metre.

The accuracy of a Code-Point Open record could be expressed as, that the
coordinated position will always be within the notional geographical extent
of the postcode."

They do also note that centroids for new postcodes where the buildings
themselves have yet to be surveyed will be given a temporary approximate
position which should be noted as such in the metadata.

My conclusion from this is that we can safely map postcodes to the building
where their centroids are placed, perhaps avoiding doing so (or adding
FIXMEs) on brand new developments.

Kind regards,

Adam

On Fri, 9 Nov 2018, 14:45 SK53 <sk53.osm at gmail.com wrote:

> I'm pretty sure that the "centroid" is allocated to the nearest delivery
> point in the postcode which places it over a building. See my (now rather)
> old blog post
> <http://sk53-osm.blogspot.com/2013/12/british-postcodes-on-openstreetmap.html>
> from 2013, and the note by Jenni Tennison. A caveat is, of course, that the
> Land Registry Prices Paid data proved to be an open data mirage.
>
> Please remember that Nominatim has a table (not recently updated) of all
> postcode centroid which are used for searches. These usually show as AB10
> 2## or similar and are at a lowish zoom level.
>
> Judging by taginfo stats we now have around 8-10% of all postcodes mapped,
> and Robert Whittaker's site suggests
> <https://osm.mathmos.net/addresses/pc-stats/> over 10%, so better than in
> 2013, but nowhere near the level we could get if we adopted a sustained
> campaign to use what information we have.
>
> Personally, I add addr:postcode to streets when: a) it is clear that all
> properties share a postcode, but individual properties have not been
> mapped; and b) when the local authority includes the full postcode on the
> streetname sign (e.g., Gedling & Rushcliffe). In the former case this
> should be regarded as an iterative step towards the desired position of
> individually mapped addresses; in the latter it reflects an on-the-ground
> rule.
>
> The available sets of open data which can be used to resolve postcodes
> are: Food Hygiene (the best, easiest to resolve, coverage of the whole UK -
> even Rutland); Companies House Open Data (surprisingly useful even in areas
> of social housing); the National Register of Social Housing (NROSH, not
> updated since 2011, but still very useful); CQC (medical practices, care
> homes etc). I haven't looked to see how many postcodes are covered by these
> in total, but it should be a reasonable proportion of the total. If you
> aren't aware Will Phillips OSM-Nottingham site does allow searching of
> various open data sets across the UK (I would recommend searching only in
> the viewport, so you need to zoom out and in to the target area). The
> quickest way to ensure at least one address is mapped for a given postcode
> is using Greg's FHRS tools.
>
> Jerry
>
> On Fri, 9 Nov 2018 at 13:44, Adam Snape <adam.c.snape at gmail.com> wrote:
>
>> Hi,
>>
>> I ask because the 'centroids' do not appear to be centroids in a pure
>> mathematical sense, they always appear to be placed on a building, never in
>> open space. Now, if this were merely been done by attributing the centroid
>> to the nearest building regardless of whether it actually belongs to the
>> postcode or not, it would serve no purpose. It seems far more likely that
>> it would be attributed to the nearest building belonging to that postcode.
>> If this is the case then it gives us a way of tying these centroids to an
>> actual building within each postcode area and that gives us something
>> tangible to map. Can anybody suggest whether I'm onto something here?
>>
>> Kind regards,
>>
>> Adam
>>
>> On Fri, 9 Nov 2018 at 13:27, David Woolley <forums at david-woolley.me.uk>
>> wrote:
>>
>>> If centroid has the plain (mathematical) meaning of the word, it will
>>> only fall exactly on the building centre if there is only one building
>>> in the postcode area.
>>>
>>> In practice the building nearest the centroid might have its own
>>> postcode, so you can't rely on the nearest building to the centroid
>>> having that postcode.
>>>
>>> There are, at least theoretically (e.g. a C shaped postcode) where the
>>> centroid is in an adjoining postcode.  I imagine you would get this if
>>> there was a cul-de-sac projecting into a crescent that was small enough
>>> to have one post code.
>>>
>>> On 09/11/18 13:12, Adam Snape wrote:
>>> > Hi,
>>> >
>>> > I agree with not mapping the centroids but...
>>> >
>>> > Is it the case that the centroids are always placed on a building
>>> which
>>> > falls under that postcode? If so, wouldn't it be okay to tag the
>>> > building with the appropriate postcode?
>>> >
>>> > Another idea: Given that postcodes (with few exceptrions) apply to
>>> only
>>> > one street, would it be acceptable to add the postcode tag to the
>>> street
>>> > where there is only one centroid on the street?
>>> >
>>> > Kind regards,
>>> >
>>> > Adam
>>> >
>>> > On Fri, 9 Nov 2018 at 12:26, Tom Hughes <tom at compton.nu
>>> > <mailto:tom at compton.nu>> wrote:
>>> >
>>> >     On 09/11/2018 11:44, David Woolley wrote:
>>> >      > On 09/11/18 11:34, David Woolley wrote:
>>> >      >> if you are only dealing with centroids, I think many have been
>>> >     mapped
>>> >      >> already,
>>> >      >
>>> >      > <
>>> https://taginfo.openstreetmap.org/search?q=uk_postcode_centroid>
>>> >      > indicates that at least 2500 have been mapped.
>>> >
>>> >     Yes, but it's a stupid idea, so please don't...
>>> >
>>> >     Tom
>>> >
>>> >     --
>>> >     Tom Hughes (tom at compton.nu <mailto:tom at compton.nu>)
>>> >     http://compton.nu/
>>> >
>>> >     _______________________________________________
>>> >     Talk-GB mailing list
>>> >     Talk-GB at openstreetmap.org <mailto:Talk-GB at openstreetmap.org>
>>> >     https://lists.openstreetmap.org/listinfo/talk-gb
>>> >
>>> >
>>> >
>>> > _______________________________________________
>>> > Talk-GB mailing list
>>> > Talk-GB at openstreetmap.org
>>> > https://lists.openstreetmap.org/listinfo/talk-gb
>>> >
>>>
>>>
>>> _______________________________________________
>>> Talk-GB mailing list
>>> Talk-GB at openstreetmap.org
>>> https://lists.openstreetmap.org/listinfo/talk-gb
>>>
>> _______________________________________________
>> 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/20181109/339f69e4/attachment-0001.html>


More information about the Talk-GB mailing list