Re: [talk-au] Fwd: Classifying settlements (Was Re: Filling in blank space (Was Re: Tagging towns by relative importance, not just population size))
Thanks for the great analysis Andrew. To clarify that I’ve understood it properly, I think your suggestion boils down to the following. Correct me if I’m wrong… City: > 20,000 population. Based on ABS ‘gravity pull’ centres. Some back and forth on OSM but gives similar numbers of cities to currently in OSM. Creates 72 cities in total. Town: 1,000 - 20,000 pop. Based on ABS ‘urban centres’ list. Gives 657 towns across Aus in total. Village: ~250 - 1,000 pop. Based on ABS ‘urban localities’ list. Gives 1,080 villages in total. Hamlet: < ~ 250 pop. Not listed by ABS. All the rest in OSM else that’s not coded as a locality or farm etc. The 20,000 / 1,000 / 250 pop cutoffs work fine by me. Following the ABS groups as best we can is simple, and will accord with patterns that map/database users are used to seeing elsewhere, which is an asset imo. I fully support it. Rather than nail the village vs hamlet distinction immediately, we could tackle the issue in 2 steps by first gaining agreement on the city / town / village cut offs, and then nailing the more challenging village / hamlet cut off in phase 2. Much of the discussion has focused on listing services / amenities that might be used to ‘boost’ places up the village / hamlet hierarchy. Once we can separate all the potential villages and hamlets from the other higher levels (towns and cities), I’m curious to see how often services / amenities just follow the breakdown you’ve suggested anyway. It may well be that most services / amenities are in ABS Urban Localities and above and that very few of the remaining hamlets have many services anyway. For isolated places, this would be relatively simple to test using OSM data. For example, how many ‘hamlets’ vs villages have hospitals etc and how many outlier hamlets are there (which might be boosted up a level). Thanks again for all your work, Ian ___ Talk-au mailing list Talk-au@openstreetmap.org https://lists.openstreetmap.org/listinfo/talk-au
Re: [talk-au] Fwd: Classifying settlements (Was Re: Filling in blank space (Was Re: Tagging towns by relative importance, not just population size))
On 11/10/23 16:22, Andrew Davidson wrote: The first thing to keep in mind is how concentrated the AU population is. Sydney and Melbourne both have 20% of the population living in them. If we add on Brisbane we reach the 50% mark, which means the majority of people live in one of three cities. As a result there is not much to go around for the rest. If we adopted 50,000 as the cutoff for a city we're going to more than halve the number of currently mapped cities. 50,000 might work for the US (and is also the value the UN has adopted for global comparisons) but it's too big for AU. At the other end 15,000 is too small, we'd end up creating an additional 25% of cities. I would suggest that we adopt the ABS's threshold of 20,000. This is the population level at which they consider a settlement starts to have "gravity" and pulls in surrounding urban areas. It used to be 30,000 back in the early days of their methodology but I assume they think people are more mobile so the "pull in" starts earlier now. 20,000 also has the benefit of not changing the number of cities we have by much. 10 currently mapped cities would become towns and 13 current towns would become cities. For towns the US threshold of 10,000 is way too crazy high. There are 1,000+ things currently mapped as towns. If we adopted 10, this would drop to 101. Even 5,000 would only get that to 198. I was thinking that we would just use the ABS's UCL list. This divides settlements into urban centres and urban localities. If a settlement is on the urban centres list and its population is over 20,000, then that's a city, otherwise it would be a town. In effect this is a cutoff of 1,000, which the ABS has used for more than 50 years suggesting that it's getting relatively smaller over time. The urban localities would be villages (a lower cutoff of ~200) and settlements not on the list hamlets. Agree with the basics of the above... The bigger shifts are going to be in the towns and villages. The UCL has (using the rules above): 72 cities 657 towns 1080 villages but we currently have 1,000+ towns and 1,800+ villages. It is hard to be very precise, as these will include place nodes nested inside other urban centres and localities. I looked at the ratio of CTVs from the US/CA/NZ on the assumption that being new world settlements the ratios should be similar. The 9 towns for each city in AU is similar to the others 7/9/8. What is different is the ratio of villages to towns. AU is 1.6 the others 2.4/4.0/2.3, which suggests: 1. There are a lot of villages in CA 2. Settlements in AU are more thinly spread. I'd expect some similar thin spread of settlements in CA too... but I have not looked. 3. 200 might be too high. The problem being it is a lot of work to get population numbers for places too small to register on the UCL. I think that the population figures can be the 'best guide' but where the population figure is not in OSM then a fall back on to the facilities measurement would be a guide lacking anything else... however unfortunate that maybe. If cities and towns all have OSM population figures than it is only towns that need this fall back guide? Oh .. hamlets ... I'd not tag a farm as a hamlet no matter how large it may be. Farm populations fluctuate with activities, I'd just leave them as farms. ___ Talk-au mailing list Talk-au@openstreetmap.org https://lists.openstreetmap.org/listinfo/talk-au
Re: [talk-au] Fwd: Classifying settlements (Was Re: Filling in blank space (Was Re: Tagging towns by relative importance, not just population size))
I think you are on the right track, Andrew. Australian Bureau of Statistics (ABS) is a credible and independent statistical authority that we are lucky to have in this country. I would be confident of their data and analyses. Their Urban Centres and Localities (UCLs) looks suitable for our purpose. Your suggested population sizes to classified as cities, towns etc seem reasonable to me, but others may have different suggestions. However, I am not sure that we have clarified if OSM has permission to use current ABS data. On Wed, 11 Oct 2023, at 4:22 PM, Andrew Davidson wrote: > On 6/10/23 18:14, Little Maps wrote: >> Thanks Graeme, it’ll be great to hear what others think too. Cheers Ian > > The first thing to keep in mind is how concentrated the AU population > is. Sydney and Melbourne both have 20% of the population living in them. > If we add on Brisbane we reach the 50% mark, which means the majority of > people live in one of three cities. As a result there is not much to go > around for the rest. > > If we adopted 50,000 as the cutoff for a city we're going to more than > halve the number of currently mapped cities. 50,000 might work for the > US (and is also the value the UN has adopted for global comparisons) but > it's too big for AU. At the other end 15,000 is too small, we'd end up > creating an additional 25% of cities. > > I would suggest that we adopt the ABS's threshold of 20,000. This is the > population level at which they consider a settlement starts to have > "gravity" and pulls in surrounding urban areas. It used to be 30,000 > back in the early days of their methodology but I assume they think > people are more mobile so the "pull in" starts earlier now. 20,000 also > has the benefit of not changing the number of cities we have by much. 10 > currently mapped cities would become towns and 13 current towns would > become cities. > > For towns the US threshold of 10,000 is way too crazy high. There are > 1,000+ things currently mapped as towns. If we adopted 10, this > would drop to 101. Even 5,000 would only get that to 198. > > I was thinking that we would just use the ABS's UCL list. This divides > settlements into urban centres and urban localities. If a settlement is > on the urban centres list and its population is over 20,000, then that's > a city, otherwise it would be a town. In effect this is a cutoff of > 1,000, which the ABS has used for more than 50 years suggesting that > it's getting relatively smaller over time. > > The urban localities would be villages (a lower cutoff of ~200) and > settlements not on the list hamlets. > > The bigger shifts are going to be in the towns and villages. The UCL has > (using the rules above): > > 72 cities > 657 towns > 1080 villages > > but we currently have 1,000+ towns and 1,800+ villages. It is hard to be > very precise, as these will include place nodes nested inside other > urban centres and localities. > > I looked at the ratio of CTVs from the US/CA/NZ on the assumption that > being new world settlements the ratios should be similar. The 9 towns > for each city in AU is similar to the others 7/9/8. What is different is > the ratio of villages to towns. AU is 1.6 the others 2.4/4.0/2.3, which > suggests: > > 1. There are a lot of villages in CA > 2. Settlements in AU are more thinly spread. > 3. 200 might be too high. The problem being it is a lot of work to get > population numbers for places too small to register on the UCL. > > > ___ > Talk-au mailing list > Talk-au@openstreetmap.org > https://lists.openstreetmap.org/listinfo/talk-au ___ Talk-au mailing list Talk-au@openstreetmap.org https://lists.openstreetmap.org/listinfo/talk-au
Re: [talk-au] Fwd: Classifying settlements (Was Re: Filling in blank space (Was Re: Tagging towns by relative importance, not just population size))
On 6/10/23 18:14, Little Maps wrote: Thanks Graeme, it’ll be great to hear what others think too. Cheers Ian The first thing to keep in mind is how concentrated the AU population is. Sydney and Melbourne both have 20% of the population living in them. If we add on Brisbane we reach the 50% mark, which means the majority of people live in one of three cities. As a result there is not much to go around for the rest. If we adopted 50,000 as the cutoff for a city we're going to more than halve the number of currently mapped cities. 50,000 might work for the US (and is also the value the UN has adopted for global comparisons) but it's too big for AU. At the other end 15,000 is too small, we'd end up creating an additional 25% of cities. I would suggest that we adopt the ABS's threshold of 20,000. This is the population level at which they consider a settlement starts to have "gravity" and pulls in surrounding urban areas. It used to be 30,000 back in the early days of their methodology but I assume they think people are more mobile so the "pull in" starts earlier now. 20,000 also has the benefit of not changing the number of cities we have by much. 10 currently mapped cities would become towns and 13 current towns would become cities. For towns the US threshold of 10,000 is way too crazy high. There are 1,000+ things currently mapped as towns. If we adopted 10, this would drop to 101. Even 5,000 would only get that to 198. I was thinking that we would just use the ABS's UCL list. This divides settlements into urban centres and urban localities. If a settlement is on the urban centres list and its population is over 20,000, then that's a city, otherwise it would be a town. In effect this is a cutoff of 1,000, which the ABS has used for more than 50 years suggesting that it's getting relatively smaller over time. The urban localities would be villages (a lower cutoff of ~200) and settlements not on the list hamlets. The bigger shifts are going to be in the towns and villages. The UCL has (using the rules above): 72 cities 657 towns 1080 villages but we currently have 1,000+ towns and 1,800+ villages. It is hard to be very precise, as these will include place nodes nested inside other urban centres and localities. I looked at the ratio of CTVs from the US/CA/NZ on the assumption that being new world settlements the ratios should be similar. The 9 towns for each city in AU is similar to the others 7/9/8. What is different is the ratio of villages to towns. AU is 1.6 the others 2.4/4.0/2.3, which suggests: 1. There are a lot of villages in CA 2. Settlements in AU are more thinly spread. 3. 200 might be too high. The problem being it is a lot of work to get population numbers for places too small to register on the UCL. ___ Talk-au mailing list Talk-au@openstreetmap.org https://lists.openstreetmap.org/listinfo/talk-au
Re: [talk-au] Fwd: Classifying settlements (Was Re: Filling in blank space (Was Re: Tagging towns by relative importance, not just population size))
Thanks Graeme, it’ll be great to hear what others think too. Cheers Ian ___ Talk-au mailing list Talk-au@openstreetmap.org https://lists.openstreetmap.org/listinfo/talk-au
[talk-au] Fwd: Classifying settlements (Was Re: Filling in blank space (Was Re: Tagging towns by relative importance, not just population size))
& re-sending to the list! Thanks Graeme -- Forwarded message - From: Graeme Fitzpatrick Date: Fri, 6 Oct 2023 at 13:08 Subject: Re: [talk-au] Classifying settlements (Was Re: Filling in blank space (Was Re: Tagging towns by relative importance, not just population size)) To: stevea Cc: Kim Oldfield via Talk-au Thanks Ian & Steve Looking at the numbers from a Qld perspective, I'd go inbetween the two samples! e.g. Hamlet <250 Village 250-1000 Town 1000-15000 City 15000< Which would produce https://overpass-turbo.eu/s/1Bup It also becomes obvious that there are quite a few places with no population listed! Thanks Graeme On Fri, 6 Oct 2023 at 12:35, stevea wrote: > Oops, resending to the talk-au list as a whole: > > > On Oct 5, 2023, at 7:00 PM, Little Maps wrote: > > City = > 50,000 people > > Town = 5000 - 50,000 > > Village = 1000 - 5000 > > Hamlet = < 1000 > > > > This kind of query gives a broad-brush pattern of how we can classify > places into cities, towns etc. If we can gain consensus on broad cutoffs, > we can then explore how services such as health and educational facilities > influence outcomes. > > A great OT query; thank you! > > In USA, and by no means do I mean to be culturally insensitive or seem > like I'm ramming anything down anybody's throat, we use some rough > guidelines at > https://wiki.openstreetmap.org/wiki/United_States/Tags#Places which > overlap somewhat. That wiki, again, deliberately USA-specific (and still > emerging and getting fine-tuned as of 2023) says: > > City = > 50,000 people > > Town = 10,000 - 50,000, though some "incorporated municipalities" which > are smaller than 10,000 (such as the rare state capital which qualifies, > like Montpelier, Vermont, or other VERY significant "towns" with less than > 10,000 but they contain an important "cultural center" like a university, a > hospital or other "major amenity" will get place=town as well, this can > include "major shopping" or something like "the only big box (hardware, > variety...) store around for a long ways") > > Village = 200 - 10,000, though this is flexible (as of 2023), and it is > emerging as consensus that a village contains at least a small commercial > area such as a supermarket, a small market (even a convenience store), a > bank, a gas station (or two, you know, for price competition's sake!) and > perhaps a medical clinic and/or cluster of doctor / dentist / medical > offices. > > Hamlet < 200 people > > Isolated Dwelling = no more than two households / families. (Could be a > sheep / cattle station for you folks down under). > > Trying to help offer perspective, please, though, "you do you" (Aussies do > Aussies). > ___ > Talk-au mailing list > Talk-au@openstreetmap.org > https://lists.openstreetmap.org/listinfo/talk-au > ___ Talk-au mailing list Talk-au@openstreetmap.org https://lists.openstreetmap.org/listinfo/talk-au
