Investigate admin data changes in the Netherlands

[stepps00]

The Dutch Centraal Bureau voor de Statistiek source (cbsnl) released updated admin datasets this year. Let's compare with the existing coverage in the Netherlands and update where necessary.

More info: https://github.com/whosonfirst/whosonfirst-sources/issues/171

[emacgillavry]

Please advise how we can assist in progressing this issue. Thanks in advance.

[stepps00]

Hey @emacgillavry - in the linked whosonfirst-data sources issue, you mentioned 15 localadmin have new boundaries in the 2020 dataset, correct? Is this the only placetype with changes expected? Or have geometries changed for other placetypes, as well?

Typically this work involves visualizing all data in the per-country repo (in this case admin-nl), and reconciling it with the source data. But since we're only correcting changes made since the last import (2016), it would be helpful to know what records we expect to update. A list of wof:id values would be a great start, or a GeoJSON file of changed geometries to help visualize the diff between 2016/2020.

[emacgillavry]

Thanks for the pointers. Will start a list of wof:ids that need updating. Here's a first overview of changes per placetype:

• neighbourhoods and boroughs have changed in 15 localadmins since 2019. See attached image to see many more changes 2016-2020: purple is no change: blue (2020) or red (2016): change of boundaries
• localities changed quite a bit
• localadmins have changed, mainly through regional re-divisions, from 390 (in 2016) to 355 (in 2020)

[emacgillavry]

Hi @stepps00. Further to the previous pic, please find attached a GeoJSON file created to show which neighbourhoods require updating because of geometry changes, either through land reclamation (hey, we're Dutch), redistricting or land exchange with Belgium!

deprecated_neighbourhoods_2016.zip

[stepps00]

Thanks @emacgillavry! I'm going to work with @missinglink to investigate this change set.. I'll flag this issue once we're able to take a closer look.

[emacgillavry]

@stepps00 have learned a lot over the last couple of weeks in getting a reasonable WOF.db together for the Netherlands. Through trial and error,I learned that treating wijken as boroughs is not the way to go as these are weighted equally as localities. Will have another go trying to label the wijken as macrohoods.

Also, I found that sometimes localities and localadmins are conflated into one object using the wof:placetype_alt tag. While these may share the same geometry for now, these may digress at some point in time. Therefore, I'd suggest to split these and assigned the current wof:id to the locality as much as possible given the significance WOF puts on localities.

What's the best way to share the custom WOF.db I created using authoritative data so this can be backported into the main WOF stup?

[stepps00]

Thanks for the update, @emacgillavry

Through trial and error,I learned that treating wijken as boroughs is not the way to go as these are weighted equally as localities.

Can you describe what you mean by "weighted equally as localities" a bit more? Do you mean this with respect to the actual administrative structure of wijken in the Netherlands? Or is this comment referencing the way you're using the WOF data?

What's the best way to share the custom WOF.db I created using authoritative data so this can be backported into the main WOF stup?

Feel free to share a download link to a copy, if you're able to host it online. Otherwise, we've had contributors convert the data to GeoJSON FeatureCollections, which they've used to create a Gist file. If either of those options work for you, I can take a look at the suggestions with @nvkelso to propose an import/update plan.

[emacgillavry]

In WOF, localities and boroughs are equally important, I found. Both in lastline and when importing polylines, boroughs and localities that share the same name are picked, (whichever is closest?).

In the Netherlands, wijken are a statistical subdivision a gemeente (localadmin), whereas woonplaatsen are an administrative subdivision of gemeente (localadmin). buurten (neighbourhoods) are a statistical subdivision of wijken. For now, we're working with the following mapping:

• buurten = neighbourhoods
• woonplaatsen = localities
• gemeenten = localadmins
• provincies = regions

There is not a 1:1 relationship between buurten and woonplaatsen, neither administratitve, nor statistical, these buurten tend to be geographically bound within one woonplaats.

[emacgillavry]

In WOF, localities and boroughs are equally important, I found. Both in lastline and when importing polylines, boroughs and localities that share the same name are picked, (whichever is closest?).

In the Netherlands, wijken are a statistical subdivision a gemeente (localadmin), whereas woonplaatsen are an administrative subdivision of gemeente (localadmin). buurten (neighbourhoods) are a statistical subdivision of wijken and as such a statistical subdivision of gemeente. There is not a 1:1 relationship between buurten and woonplaatsen, neither administratitve, nor statistical, however these buurten tend to be geographically bound within one woonplaats.

For now, we're working with the following mapping:

• buurten = neighbourhoods
• woonplaatsen = localities
• gemeenten = localadmins
• provincies = regions

We left out wijken for now, as these sometimes cover several woonplaatsen (localities). We're considering to cookie cut the wijken on the boundaries of the woonplaatsen. Is that a viable option from WOF perspective? Or had we better map wijken to macrohoods?

Full database for the Netherlands available from https://www.dropbox.com/s/k8xi0cxwlydod7i/whosonfirst-data-admin-nl-latest.zip

[stepps00]

We're considering to cookie cut the wijken on the boundaries of the woonplaatsen. Is that a viable option from WOF perspective? Or had we better map wijken to macrohoods?

Let me reply to this issue with more detail once I have a chance to dive into your db file.. but we'd prefer to keep the WOF placetype <> Netherlands admin divisions as-is, whenever possible.

Without diving into the suggestions too deeply, it does look like WOF already maintains the wijken as macrohoods.