[OSM-talk-ie] Estimate of number of building=* in Ireland

[Colm Moore]

Hi,

Inspired by seeing the estimate in the Microgrant application of 5.5 million 
buildings on the island of Ireland, I did some number crunching.

I downloaded the populations of Kilkenny townlands (1,500+) from the CSO and 
analysed the population against the number of buildings per civil parish (100+) 
for County Kilkenny. This is assuming Kilkenny has all or nearly all buildings 
mapped. Based on my inspections, this is largely true.

The CSO data is somewhat distorted for the Kilkenny city area (100+ townlands), 
due to the way the CSO have arranged the townlands and civil parishes. I could 
look at this in more detail, but there would be a few hours of effort (unless 
someone has a simple way of calculating number of buildings per area, for a 
large number of areas).

I calculated the 'number of buildings per civil parish' using the Overpass 
Turbo query [building=* in "civilparishname, Kilkenny"]. Overpass Turbo gives a 
summary of the data in the bottom right corner of the screen, e.g. 

Loaded – nodes: 4261, ways: 867, relations: 2
Displayed – pois: 0, lines: 0, polygons: 866

I took the number of polygons to mean the number of buildings (this might not 
be perfect - I don't know how those numbers add up). Additionally, some 
polygons, e.g. building=terrace represent several buildings, while in other 
cases buildings may have been crudely split or joined-up.

Depending on the civil parish, we're looking at 0.32-2.29 polygons per capita 
(0.44-3.15 people per building). Rural areas ten to have more polygons per 
capita, especially due to farm outbuildings, while urban areas have fewer 
polygons per capita, due to apartments buildings and semi-detached buildings 
(e.g. two square houses joined together might have only six nodes).

I also calculated 4.40-5.83 nodes per polygon. This means some civil parishes 
have predominantly rectangular polygons / buildings, whereas others have many 
L-shaped or other-shaped polygons / buildings.

As I wasn't able to immediately get some 'number of buildings per civil parish' 
numbers (Overpass Turbo had problems returning them, possibly due to duplicate 
names and variations in name spellings), I had to calculate them from their 
component townlands, using the Overpass Turbo query [building=* in 
"townlandname, civilparishname, Kilkenny"].

Depending on the townland, we're looking at 0.23-8.00 polygons per capita 
(0.13-4.37 people per building) and 3.91-6.45 nodes per polygon (i.e. some 
townlands have large numbers of semi-detached or terraced buildings, whereas 
others have a high number of complicated-shape polygons / buildings or 
buildings with too many mapped nodes). It is usual to see more extreme spreads 
when looking at smaller areas.

I'm coming up with about 5.4 million (close enough!) buildings for the whole 
island, assuming the pattern is the same everywhere. However, as shown by 
analysing the smaller areas, there is variation and the 'final' number will 
vary from that. Of course, given that OSM is an ongoing project, there will 
never be a final number.

Colm
VictorIE
___
Talk-ie mailing list
Talk-ie@openstreetmap.org
https://lists.openstreetmap.org/listinfo/talk-ie

[OSM-talk-ie] Estimate of number of building=* in Ireland

[Colm Moore]

Hi,

Apologies, there was something else I meant to mention.

Some locations have a very low number of building polygons per capita (i.e. 
implying high number of people per building). This may be **suggestive** of 
locations that are under-mapped. This bears out in some of suburbs on the 
Kilkenny side of Waterford city, where there are lots of unmappable (tree cover 
and fuzzy images) garden sheds, etc. There are other reasons for this, like 
apartment buildings, terraced houses mapped as building=terrace, etc.

On Wed, 13 May 2020 at 14:38, Donie Kelly  wrote:

> Don’t buildings have tags? Did I see a residential tag? Is it used in all
> cases?

Yes, buildings can be given detailed descriptions, e.g. building=house, 
building=apartments. However, about half of the buildings on the island just 
use building=yes. Only saying building=yes gets the building on the map 
quickly, makes them readily findable and lets other mappers fill in details 
later. Not all buildings are readily identifiable from aerial images and 
details like number of storeys can be even harder to determine.

Note that building=residential can be used for buildings that are residential, 
but the type isn't readily identifiable, e.g. I saw some semi-detached houses 
in Kilkenny that were attached back-to-back instead of side-to-side.

Interpreting the total number of buildings on the island from the total number 
of, say houses, would be more complicated than comparing number of buildings in 
Kilkenny to the number of buildings on the island.

On Thu, 14 May 2020 11:47:29 Tadeusz Cantwell  wrote:

> The residential tag should be used for an area not a house.

Yes, landuse=residential should be used for residential sites / estates / 
neighbourhoods. The iD editor now supports additional sub-tags, e.g. 
residential=rural, residential=apartments and residential=halting_site You can 
also add name=* to the likes of housing estates, which can help detail the 
hierarchy of a town.

Colm


---


___
Talk-ie mailing list
Talk-ie@openstreetmap.org
https://lists.openstreetmap.org/listinfo/talk-ie

Re: [OSM-talk-ie] Estimate of number of building=* in Ireland

[Donie Kelly]

Don’t buildings have tags? Did I see a residential tag? Is it used in all cases?

> On 13 May 2020, at 13:24, Colm Moore  wrote:
> 
> Hi,
> 
> Inspired by seeing the estimate in the Microgrant application of 5.5 million 
> buildings on the island of Ireland, I did some number crunching.
> 
> I downloaded the populations of Kilkenny townlands (1,500+) from the CSO and 
> analysed the population against the number of buildings per civil parish 
> (100+) for County Kilkenny. This is assuming Kilkenny has all or nearly all 
> buildings mapped. Based on my inspections, this is largely true.
> 
> The CSO data is somewhat distorted for the Kilkenny city area (100+ 
> townlands), due to the way the CSO have arranged the townlands and civil 
> parishes. I could look at this in more detail, but there would be a few hours 
> of effort (unless someone has a simple way of calculating number of buildings 
> per area, for a large number of areas).
> 
> I calculated the 'number of buildings per civil parish' using the Overpass 
> Turbo query [building=* in "civilparishname, Kilkenny"]. Overpass Turbo gives 
> a summary of the data in the bottom right corner of the screen, e.g. 
> 
> Loaded – nodes: 4261, ways: 867, relations: 2
> Displayed – pois: 0, lines: 0, polygons: 866
> 
> I took the number of polygons to mean the number of buildings (this might not 
> be perfect - I don't know how those numbers add up). Additionally, some 
> polygons, e.g. building=terrace represent several buildings, while in other 
> cases buildings may have been crudely split or joined-up.
> 
> Depending on the civil parish, we're looking at 0.32-2.29 polygons per capita 
> (0.44-3.15 people per building). Rural areas ten to have more polygons per 
> capita, especially due to farm outbuildings, while urban areas have fewer 
> polygons per capita, due to apartments buildings and semi-detached buildings 
> (e.g. two square houses joined together might have only six nodes).
> 
> I also calculated 4.40-5.83 nodes per polygon. This means some civil parishes 
> have predominantly rectangular polygons / buildings, whereas others have many 
> L-shaped or other-shaped polygons / buildings.
> 
> As I wasn't able to immediately get some 'number of buildings per civil 
> parish' numbers (Overpass Turbo had problems returning them, possibly due to 
> duplicate names and variations in name spellings), I had to calculate them 
> from their component townlands, using the Overpass Turbo query [building=* in 
> "townlandname, civilparishname, Kilkenny"].
> 
> Depending on the townland, we're looking at 0.23-8.00 polygons per capita 
> (0.13-4.37 people per building) and 3.91-6.45 nodes per polygon (i.e. some 
> townlands have large numbers of semi-detached or terraced buildings, whereas 
> others have a high number of complicated-shape polygons / buildings or 
> buildings with too many mapped nodes). It is usual to see more extreme 
> spreads when looking at smaller areas.
> 
> I'm coming up with about 5.4 million (close enough!) buildings for the whole 
> island, assuming the pattern is the same everywhere. However, as shown by 
> analysing the smaller areas, there is variation and the 'final' number will 
> vary from that. Of course, given that OSM is an ongoing project, there will 
> never be a final number.
> 
> Colm
> VictorIE
> ___
> Talk-ie mailing list
> Talk-ie@openstreetmap.org
> https://lists.openstreetmap.org/listinfo/talk-ie

___
Talk-ie mailing list
Talk-ie@openstreetmap.org
https://lists.openstreetmap.org/listinfo/talk-ie

Re: [OSM-talk-ie] Estimate of number of building=* in Ireland

[Tadeusz Cantwell]

The residential tag should be used for an area not a house.

On Wed, 13 May 2020 at 14:38, Donie Kelly  wrote:

> Don’t buildings have tags? Did I see a residential tag? Is it used in all
> cases?
>
> > On 13 May 2020, at 13:24, Colm Moore  wrote:
> >
> > Hi,
> >
> > Inspired by seeing the estimate in the Microgrant application of 5.5
> million buildings on the island of Ireland, I did some number crunching.
> >
> > I downloaded the populations of Kilkenny townlands (1,500+) from the CSO
> and analysed the population against the number of buildings per civil
> parish (100+) for County Kilkenny. This is assuming Kilkenny has all or
> nearly all buildings mapped. Based on my inspections, this is largely true.
> >
> > The CSO data is somewhat distorted for the Kilkenny city area (100+
> townlands), due to the way the CSO have arranged the townlands and civil
> parishes. I could look at this in more detail, but there would be a few
> hours of effort (unless someone has a simple way of calculating number of
> buildings per area, for a large number of areas).
> >
> > I calculated the 'number of buildings per civil parish' using the
> Overpass Turbo query [building=* in "civilparishname, Kilkenny"]. Overpass
> Turbo gives a summary of the data in the bottom right corner of the screen,
> e.g.
> >
> > Loaded – nodes: 4261, ways: 867, relations: 2
> > Displayed – pois: 0, lines: 0, polygons: 866
> >
> > I took the number of polygons to mean the number of buildings (this
> might not be perfect - I don't know how those numbers add up).
> Additionally, some polygons, e.g. building=terrace represent several
> buildings, while in other cases buildings may have been crudely split or
> joined-up.
> >
> > Depending on the civil parish, we're looking at 0.32-2.29 polygons per
> capita (0.44-3.15 people per building). Rural areas ten to have more
> polygons per capita, especially due to farm outbuildings, while urban areas
> have fewer polygons per capita, due to apartments buildings and
> semi-detached buildings (e.g. two square houses joined together might have
> only six nodes).
> >
> > I also calculated 4.40-5.83 nodes per polygon. This means some civil
> parishes have predominantly rectangular polygons / buildings, whereas
> others have many L-shaped or other-shaped polygons / buildings.
> >
> > As I wasn't able to immediately get some 'number of buildings per civil
> parish' numbers (Overpass Turbo had problems returning them, possibly due
> to duplicate names and variations in name spellings), I had to calculate
> them from their component townlands, using the Overpass Turbo query
> [building=* in "townlandname, civilparishname, Kilkenny"].
> >
> > Depending on the townland, we're looking at 0.23-8.00 polygons per
> capita (0.13-4.37 people per building) and 3.91-6.45 nodes per polygon
> (i.e. some townlands have large numbers of semi-detached or terraced
> buildings, whereas others have a high number of complicated-shape polygons
> / buildings or buildings with too many mapped nodes). It is usual to see
> more extreme spreads when looking at smaller areas.
> >
> > I'm coming up with about 5.4 million (close enough!) buildings for the
> whole island, assuming the pattern is the same everywhere. However, as
> shown by analysing the smaller areas, there is variation and the 'final'
> number will vary from that. Of course, given that OSM is an ongoing
> project, there will never be a final number.
> >
> > Colm
> > VictorIE
> > ___
> > Talk-ie mailing list
> > Talk-ie@openstreetmap.org
> > https://lists.openstreetmap.org/listinfo/talk-ie
>
> ___
> Talk-ie mailing list
> Talk-ie@openstreetmap.org
> https://lists.openstreetmap.org/listinfo/talk-ie
>
___
Talk-ie mailing list
Talk-ie@openstreetmap.org
https://lists.openstreetmap.org/listinfo/talk-ie

Re: [OSM-talk-ie] Estimate of number of building=* in Ireland

[Mateusz Konieczny via Talk-ie]

both building=residential and landuse=residential
are valid tagging

(though, if possible then building=house
or other more specific value is preferable)


May 14, 2020, 12:47 by t4d...@gmail.com:

> The residential tag should be used for an area not a house.
>
> On Wed, 13 May 2020 at 14:38, Donie Kelly  wrote:
>
>> Don’t buildings have tags? Did I see a residential tag? Is it used in all
>> cases?
>>
>> > On 13 May 2020, at 13:24, Colm Moore  wrote:
>> >
>> > Hi,
>> >
>> > Inspired by seeing the estimate in the Microgrant application of 5.5
>> million buildings on the island of Ireland, I did some number crunching.
>> >
>> > I downloaded the populations of Kilkenny townlands (1,500+) from the CSO
>> and analysed the population against the number of buildings per civil
>> parish (100+) for County Kilkenny. This is assuming Kilkenny has all or
>> nearly all buildings mapped. Based on my inspections, this is largely true.
>> >
>> > The CSO data is somewhat distorted for the Kilkenny city area (100+
>> townlands), due to the way the CSO have arranged the townlands and civil
>> parishes. I could look at this in more detail, but there would be a few
>> hours of effort (unless someone has a simple way of calculating number of
>> buildings per area, for a large number of areas).
>> >
>> > I calculated the 'number of buildings per civil parish' using the
>> Overpass Turbo query [building=* in "civilparishname, Kilkenny"]. Overpass
>> Turbo gives a summary of the data in the bottom right corner of the screen,
>> e.g.
>> >
>> > Loaded – nodes: 4261, ways: 867, relations: 2
>> > Displayed – pois: 0, lines: 0, polygons: 866
>> >
>> > I took the number of polygons to mean the number of buildings (this
>> might not be perfect - I don't know how those numbers add up).
>> Additionally, some polygons, e.g. building=terrace represent several
>> buildings, while in other cases buildings may have been crudely split or
>> joined-up.
>> >
>> > Depending on the civil parish, we're looking at 0.32-2.29 polygons per
>> capita (0.44-3.15 people per building). Rural areas ten to have more
>> polygons per capita, especially due to farm outbuildings, while urban areas
>> have fewer polygons per capita, due to apartments buildings and
>> semi-detached buildings (e.g. two square houses joined together might have
>> only six nodes).
>> >
>> > I also calculated 4.40-5.83 nodes per polygon. This means some civil
>> parishes have predominantly rectangular polygons / buildings, whereas
>> others have many L-shaped or other-shaped polygons / buildings.
>> >
>> > As I wasn't able to immediately get some 'number of buildings per civil
>> parish' numbers (Overpass Turbo had problems returning them, possibly due
>> to duplicate names and variations in name spellings), I had to calculate
>> them from their component townlands, using the Overpass Turbo query
>> [building=* in "townlandname, civilparishname, Kilkenny"].
>> >
>> > Depending on the townland, we're looking at 0.23-8.00 polygons per
>> capita (0.13-4.37 people per building) and 3.91-6.45 nodes per polygon
>> (i.e. some townlands have large numbers of semi-detached or terraced
>> buildings, whereas others have a high number of complicated-shape polygons
>> / buildings or buildings with too many mapped nodes). It is usual to see
>> more extreme spreads when looking at smaller areas.
>> >
>> > I'm coming up with about 5.4 million (close enough!) buildings for the
>> whole island, assuming the pattern is the same everywhere. However, as
>> shown by analysing the smaller areas, there is variation and the 'final'
>> number will vary from that. Of course, given that OSM is an ongoing
>> project, there will never be a final number.
>> >
>> > Colm
>> > VictorIE
>> > ___
>> > Talk-ie mailing list
>> > Talk-ie@openstreetmap.org
>> > https://lists.openstreetmap.org/listinfo/talk-ie
>>
>> ___
>> Talk-ie mailing list
>> Talk-ie@openstreetmap.org
>> https://lists.openstreetmap.org/listinfo/talk-ie
>>
> ___
> Talk-ie mailing list
> Talk-ie@openstreetmap.org
> https://lists.openstreetmap.org/listinfo/talk-ie
>

___
Talk-ie mailing list
Talk-ie@openstreetmap.org
https://lists.openstreetmap.org/listinfo/talk-ie

Re: [OSM-talk-ie] Estimate of number of building=* in Ireland

[Dave Corley]

The fact that so many buildings are being mapped as building=yes when more
accurate tagging is plainly obvious makes this an utterly frustrating
project.

Saying that others can come along after and correct it is a totally
avoidable requirement by using more appropriate tagging in the first place.
Its also unlike to ever get corrected as evidenced by the volume of data in
Ireland that was created and never improved upon.

The logical approach should be to map it as house/garage/shed etc and ONLY
use building=yes if you cannot make an educated guess.

Doing otherwise puts this data on the level of the tiger import.



On Thu, May 14, 2020 at 1:04 PM Colm Moore  wrote:

> Hi,
>
> Apologies, there was something else I meant to mention.
>
> Some locations have a very low number of building polygons per capita
> (i.e. implying high number of people per building). This may be
> **suggestive** of locations that are under-mapped. This bears out in some
> of suburbs on the Kilkenny side of Waterford city, where there are lots of
> unmappable (tree cover and fuzzy images) garden sheds, etc. There are other
> reasons for this, like apartment buildings, terraced houses mapped as
> building=terrace, etc.
>
> On Wed, 13 May 2020 at 14:38, Donie Kelly  wrote:
>
> > Don’t buildings have tags? Did I see a residential tag? Is it used in all
> > cases?
>
> Yes, buildings can be given detailed descriptions, e.g. building=house,
> building=apartments. However, about half of the buildings on the island
> just use building=yes. Only saying building=yes gets the building on the
> map quickly, makes them readily findable and lets other mappers fill in
> details later. Not all buildings are readily identifiable from aerial
> images and details like number of storeys can be even harder to determine.
>
> Note that building=residential can be used for buildings that
> are residential, but the type isn't readily identifiable, e.g. I saw some
> semi-detached houses in Kilkenny that were attached back-to-back instead of
> side-to-side.
>
> Interpreting the total number of buildings on the island from the total
> number of, say houses, would be more complicated than comparing number of
> buildings in Kilkenny to the number of buildings on the island.
>
> On Thu, 14 May 2020 11:47:29 Tadeusz Cantwell  wrote:
>
> > The residential tag should be used for an area not a house.
>
> Yes, landuse=residential should be used for residential sites / estates /
> neighbourhoods. The iD editor now supports additional sub-tags,
> e.g. residential=rural, residential=apartments and residential=halting_site
> You can also add name=* to the likes of housing estates, which can help
> detail the hierarchy of a town.
>
> Colm
>
>
>
> ---
>
>
> ___
> Talk-ie mailing list
> Talk-ie@openstreetmap.org
> https://lists.openstreetmap.org/listinfo/talk-ie
>
___
Talk-ie mailing list
Talk-ie@openstreetmap.org
https://lists.openstreetmap.org/listinfo/talk-ie

Re: [OSM-talk-ie] Estimate of number of building=* in Ireland

[Mateusz Konieczny via Talk-ie]

May 15, 2020, 00:04 by davecor...@gmail.com:

> The fact that so many buildings are being mapped as building=yes when more
> accurate tagging is plainly obvious makes this an utterly frustrating
> project.
>
> Saying that others can come along after and correct it is a totally
> avoidable requirement by using more appropriate tagging in the first place.
> Its also unlike to ever get corrected as evidenced by the volume of data in
> Ireland that was created and never improved upon.
>
If there is not enough mappers then likely mapping more building as building=yes
is more efficient than using specific building value anyway.

And anyone may map whatever (s)he wants, it is perfectly fine to 
map roads without mapping buildings, map trees instead of mapping roads,
map buildings as building=yes, map bicycle parkings and map no shops at all...

Note also that for example StreetComplete allows newbies and beginners to
build on build on initial not finished data and for example specify 
more exact building types.

> The logical approach should be to map it as house/garage/shed etc and ONLY
> use building=yes if you cannot make an educated guess.
>
> Doing otherwise puts this data on the level of the tiger import.
>
No. TIGER import had plenty of outright invalid, incorrect and insane data.
This is merely missing not very important part of info.

https://wiki.openstreetmap.org/wiki/File:TIGER_fixup_example_before.jpg
https://wiki.openstreetmap.org/wiki/File:Braided-streets-example.jpg
https://wiki.openstreetmap.org/wiki/TIGER_fixup

___
Talk-ie mailing list
Talk-ie@openstreetmap.org
https://lists.openstreetmap.org/listinfo/talk-ie

Re: [OSM-talk-ie] Estimate of number of building=* in Ireland

[Ciarán Staunton]

I don’t understand at all how a community project like osmIRL buildings
could be utterly frustrating, and I wonder about the flipancy of a comment
that the effect is anything like a Tiger import – simply on the basis of
accuracy.



The aim of the task is to have a straightforward activity for people to
work on together, which we haven’t had in a while. Another aim is to have
something that is easy enough to start on as we recruit new people into the
community as contributors. The daily contributions extract shows we have
doubled the contributors to 30. There will be a second phase, when all the
tasks are complete and validated, and there would be even more animations
resolving building use types.



Ireland’s building equals yes tag was 62% of all building objects when the
project started around October 2019 and is 64% only today, was this making
things worse? Admittedly while it increases the number of building equals
yes objects in number, it doesn’t change the proportion, and it completes
the map on a theme the community wanted to go ahead with. The project has a
future phase when there will be more added to the buildings now produced by
this effort.



Right now Kilkenny City 25%, Carlow Town 38% and Waterford City 48% all
have building equals yes tags less than what prevailed on the island at the
outset. In the recent community hangouts and presentations there has been
discussion of techniques already put into practice on how validation
reduces tag generality. JOSM filters as well as a 'Second mapper'
standpoint are blitzing the building equals yes tag before the tasks get
taken down. Nobody seems to think using the filter is hassle to use, or
that to select, scroll and tag numerous buildings – so what have we missed?



Finally, the osm community worldwide has agreed that building equals yes is
not an invalid contribution where the experience level of the mapper is
lower, or the satellite imagery is blurred, or there is a community
decision to map at scale, with planned and committed passes through that
work. Otherwise it wouldn’t exist as a tag, and be the default in the JOSM
buildings plugin. This would be why this tag is 75% of the building objects
in Europe, which are ubiquitous. If this project delivers on a rate much
less than that there ought not to be frustration. There are only two tasks
mapped and validated so far, by the time we finish phase 1 there will be a
lower percentage of building equals yes, so while we can debate the
inaccuracy of the comment about it being akin to a Tiger import, I don't
think that the fact that such an assessment is massively premature needs
much debate at all.


Virus-free.
www.avg.com

<#DAB4FAD8-2DD7-40BB-A1B8-4E2AA1F9FDF2>
___
Talk-ie mailing list
Talk-ie@openstreetmap.org
https://lists.openstreetmap.org/listinfo/talk-ie

Re: [OSM-talk-ie] Estimate of number of building=* in Ireland

[Andy Townsend]

On 15/05/2020 03:37, Mateusz Konieczny via Talk-ie wrote:

May 15, 2020, 00:04 by davecor...@gmail.com:


The logical approach should be to map it as house/garage/shed etc and ONLY
use building=yes if you cannot make an educated guess.

Doing otherwise puts this data on the level of the tiger import.


No. TIGER import had plenty of outright invalid, incorrect and insane data.
This is merely missing not very important part of info.

https://wiki.openstreetmap.org/wiki/File:TIGER_fixup_example_before.jpg
https://wiki.openstreetmap.org/wiki/File:Braided-streets-example.jpg
https://wiki.openstreetmap.org/wiki/TIGER_fixup


+1 to that.

Sometimes people forgot just how piss-poor some of the imported TIGER 
data was.  To be fair, it was a lifetime ago in OSM terms.


Best Regards,

Andy

(not a local but an occasional visitor - next visit was due to be in a 
couple of weeks but alas now postponed)



___
Talk-ie mailing list
Talk-ie@openstreetmap.org
https://lists.openstreetmap.org/listinfo/talk-ie