Individual residential landuse for each house or street: good or bad?

[JamesBelchamber]

@LordGarySugar

How are you going to calculate and then subtract the area of roads? They are very rarely mapped.

Length * Width - where road width is measured, use that, where it isn’t, use a sensible default based on the road type.

@LordGarySugar

Why would you need to calculate the total area including roads?

To calculate the proportion of a residential area that is made up of roads? I dunno I wouldn’t make assumptions about how the data is used, I’d just make it as useful as possible. What are you using the data for?

Note that traditionally residential roads are considered to be part of residential land use, too - at least in the UK - so you would be deviating from that norm EDIT: apparently that changed

Residential areas, including all dwellings, gardens and outbuildings, but EXCLUDING all access roads, pavements paths and verges and any other surrounding infrastructure

Seems to mostly align with cost-saving measures rather than data quality, but aligning with the government’s definition would be a good reason to map this way. EDIT2: and now they seem to have stopped collecting it at all? Wonderful

EDIT3-Final.xlsx: but the change in approach did seem to then align with the original edits :

[LordGarySugar]

@JamesBelchamber

Length * Width - where road width is measured,

I believe width= refers only to the physical carriageway width - not including sidewalks and green buffers, so you’d still have to approximate those parts

@JamesBelchamber

To calculate the proportion of a residential area that is made up of roads?

Well I believe you can do this now. If you define an area, you can see the exact percentage of land dedicated to residential retail etc… (can’t remember the name of the site that shows this with pie charts) currently any land dedicated to roads will show up as unknown… maybe I need to add lots of landuse=highway

@JamesBelchamber

apparently that changed

Thank you David Cameron for backing me up on this one

Although I certainly wouldn’t divide up landuse with as much granularity as in that screenshot - things like shared paths, shared driveways and the grassy areas around apartment blocks/council houses I would all include within residential areas.

[JamesBelchamber]

@LordGarySugar

maybe I need to add lots of landuse=highway

Unironically, if you’re mapping like this, I would yeh. But then you probably don’t want to stick the polygons together, so you’d still have gaps. Hmm.

@LordGarySugar

Thank you David Cameron for backing me up on this one

I can’t believe you’ve brought austerity into our fair map smh

[LordGarySugar]

@JamesBelchamber

you probably don’t want to stick the polygons together, so you’d still have gaps.

Why would I not connect them together? I wouldn’t leave any gaps between landuse=residential and landuse=highway… (not the same as area:landuse=residential

[SomeoneElse (Andy Townsend)]

@LordGarySugar

landuse=highway

I was surprised to see that mentioned - I think that it’s fair to say that the use of that tag is at best “fairly uneven” - as an example look at usage in the southern UK.

[UtterClutter]

Even landuse=railway is a bit sparse

[ndrw6]

There is nothing wrong with not mapping land uses (leaving blank spaces) - we are trying to provide useful information, not colour the map. If there is something of interest for you to map - go for it, but don’t feel compelled to put something there just because the area is empty.

To me the whole discussion is about de jure zoning regulations vs actual land uses. In my opinion calling a residential road a residential landuse is a stretch. It does not function as a residential area and it is serving residents in the same way as nearby supermarkets, sewage processing plants or district power plants do.

From purely practical point of view, it is rare to find truly residential areas. Even in new-built areas there are exceptions (publicly available parks etc). Typical villages, towns have hundreds retail/commercial/religious/administrative amenities all over the place, often served by “residential” roads.

BTW, coastal areas are fun: OpenStreetMap. Patchwork of residential, commercial and anything in between areas. The only way of mapping them is sticking to on the ground truth principle and map each property as it is.

[JamesBelchamber]

@ndrw6

To me the whole discussion is about de jure zoning regulations vs actual land uses. In my opinion calling a residential road a residential landuse is a stretch.

I don’t think it’s really about de jure vs facts on the ground - rather, it’s about whether you consider roads, which only/mainly provide access for residents, to be “land used for residential purposes”. A good measure may be “if we knocked down the surrounding houses, would we still keep this road”.

@ndrw6

From purely practical point of view, it is rare to find truly residential areas.

I think this might undermine the “residential is only houses” argument.

[JamesBelchamber]

I should add that I don’t feel super strongly about this because ultimately I don’t think anyone is using this data for anything; it certainly does not cause problems e.g. with routing. My strongest opinion is reserved for retail areas, which I think should include all the related infrastructure (parking, roads, etc) that form a single, cohesive “shopping area”. But even then, it doesn’t seem to have a meaningfully degrading effect (the map might look a bit confusing and not really match the facts on the ground but people can put two and two together).

Are there consumers of this data (other than for pretty rendering) which do have opinions on it?

[Mateusz_Konieczny]

@ndrw6

To me the whole discussion is about de jure zoning regulations vs actual land uses.

Not really?

We definitely do not map de iure planning zones. Farmland zoned for residential is landuse=farmland not residential

And this has widespread support, with opposition below lizardman levels.

[UtterClutter]

I am often bemused by the common rhetoric which describes cartographers as providing “too much detail”

[LordGarySugar]

Me again - I hope it’s ok to do this but I’d like to continue this thread to try to find a consensus with how to re-tag or correct residential landuse that has been split down to the level of individual properties. We seem to agree that residential landuse should not be used for mapping individual residential plots, but is it right to simply merge them all together and get rid of them, or should they be retagged place=plot to preserve previous mappers’ work? Currently there is only ~800 uses of place=plot in the UK, so I wanted to discuss what different approaches could be taken. The question is essentially, Should we be mapping individual properties as place=plot in the UK?

Personally, I would not go to the effort of mapping place=plot, but I want to know if I should be preserving what has been split already or if it’s ok just to merge it all.
As an example, here are four residential properties which could in theory be tagged as place=plot as they are owned by different people and have visible barriers separating them:

Some points to explore: what would be the benefits and drawbacks of mapping individual residential plots, and what tags would be used (if any) to associate a plot with a particular address or building. Also, how would apartment complexes be mapped as plots, if at all?

I raise these questions as it has been shown in the thread above that there are a number of places where residential landuse has been excessively subdivided. (That’s my opinion, but I believe there is a strong consensus in this thread not to split residential landuse any smaller once it has been divided up to exclude residential roads - note some mappers would not split residential landuse even to this extent). I’d like to find a solution for how to re-tag these areas, and harmonise their mapping with other places around the country. The two main options are 1) Merge subdivided landuse together and do not map place=plot or 2) Retag subdivided landuse as place=plot, and re-add larger landuse=residential ways surrounding them

There are 2 main ways residential landuse has been subdivided, and would require a different amount of fixing, depending on what solution was preferred.
The first way is when residential landuse has been divided into individual properties, similar to my example image. This example is in Cambridge.

These could be relatively easily retagged to place=plot, but would require some work drawing new landuse=residential polygons around them. (Maybe there’s a josm plugin that could do that?)

The second way is where landuse has been split into chunks of a handful of houses, in this example seemingly according to which way the houses are facing. This is in Windsor.

It would not be possible to retag these area as place=plot, so what to do with them?

Poll time! For the purposes of this poll, only the residential areas the size of individual properties would be retagged as place=plot. The residential areas in Example 2 cannot be retagged as place=plot, and would need to be split further before they could be tagged as place=plot.

Retag these areas as place=plot, or just merge them back into regular sized landuse?

• Merge subdivided landuse together and do not map place=plot
• Retag subdivided landuse as place=plot, and re-add larger landuse=residential ways surrounding them
• None of the above!

[Jarek]

@LordGarySugar

but would require some work drawing new landuse=residential polygons around them. (Maybe there’s a josm plugin that could do that?)

there is indeed: JOSM/Plugins/Shrinkwrap - OpenStreetMap Wiki

[LordGarySugar]

Ah, that’s exactly what I was thinking of! Would make it easy to sort this stuff out, if we decide to convert them to place=plot areas.

[Friendly_Ghost (Casper)]

We should refer people who are interested in place=plot to authoritative sources, because there’s no way we can maintain the data and guarantee the correctness of it in OSM.

[ceirios]

The cadastral parcels in my area here are about 14 years out of date at least. IMO, we shouldn’t blindly use these and especially not to waste time mapping place=plot.

Thankfully imagery makes it obvious that the parcels are out of date but the same can’t be said for an average row of terraces

[Robert_Whittaker]

The way the Land Registry cadastral parcels work is that parcels can be subdivided but never merged. So what looks to be out of date is actually the correct (albeit not very helpful) current representation. The knock-on effect of this is that we don’t have an open data source for the actual current parcel/plot boundaries where there’s been redevelopment.

In the interests of saving the previous mappers work (and since there’s a JOSM plugin to make adding the actual landuse polygon much easier) I’d say we should keep the plot boundaries and just retag them as place=plot.

In the second example, I think I’d just do whatever is easiest to more reasonable landuse polygons out.

In both cases, it would probably be polite to invite the original mappers to this thread and give them a chance to comment first (if they haven’t already).

[LordGarySugar]

I’ve put a couple of changeset comments, but having used overpass to look through recently edited (within last year, England 3 months) residential landuse areas in Eng, Sc, Wal, NI, I couldn’t find any other users splitting landuse to this extent. (As would be expected, it’s not a standard way of mapping).

Certainly I don’t think people are blindly tracing cadastral boundaries, they are following physical barriers where they differ from the cadastral overlay. (In reference to Ceirios’ screenshot)

As for the ‘just use official data’ argument raised by Casper, that’s problematic in the UK because as RW says we don’t have open data for ‘real’ property boundaries. So people could be forgiven for wanting to create this data themselves. I raise the following theoretical scenario. What if I want to know the land area occupied by all the houses along Pudding Lane? Is that a question we should be trying to solve with OSM data or not? (we could if all the plots were mapped and associated with the street in some way). I don’t personally have a need for that information here, I’m just devil’s-advocating to see if there are any possible uses for place=plot data.

[LordGarySugar]

In the meantime I’ve done an initial edit here, retagging and shrinkwrapping to see what the area would look like with place=plot retained. This edit compressed 1000 landuse=residential into just 100. The main problem I have with the data as it is now is that there is no identifiable information on each plot, so what is the point?

[ndrw6]

Please don’t experiment on live data and revert this changeset before further edits make that difficult.

If these landuses were intended to be mapped as place=plot that would have been done already. Clearly this was not an intention of previous mappers. There are many reasons for mapping landuses in greater detail, mainly centred around drawing boundaries in greater detail and making mapping of mixed-use areas easier. If you do care about specifying a specific area is a plot - go ahead and add place=plot but don’t remove work of others.

We’ve been going through this discussion already (several months ago IIRC). There was no consensus for a specific level of detail that should be used for mapping landuses, certainly not to the point that one solution should be considered wrong and in need of fixing. I don’t track these discussions daily but that does not mean my position has changed.