[Talk-GB] UK coastline data

[Borbus]

Hello again,

I've finished integrating the VectorMap data all the way along North
Norfolk and now the Blackwater estuary too. I did a survey in Maldon before
I made changes near the town to make sure nothing was wildly different. It
looks like the new Maxar imagery is quite recent in that area and,
helpfully, was taken at low tide. I made minor adjustments to the VectorMap
data in places where the more recent imagery showed significant differences.

I wanted to fix up some other coastal areas I'm familiar with, such as
around Folkestone and Dover, but these areas consist of features other than
beaches and mud flats that are less clear to me how to map (like "beaches"
consisting of large rocks from the cliffs).

There has been a change merged into osm-carto which will, in my opinion,
degrade the rendering of many of these tidal areas. You can see the change
here, along with my concerns at the bottom (I am georgek):
https://github.com/gravitystorm/openstreetmap-carto/issues/3707

Doing this work, and seeing the above changes to osm-carto got me thinking
that the time is ripe to sort out how we do water in OSM. There are
fundamentally two concerns which are currently conflated in OSM:
* The water area, ie. "you will not be dry if you try to stand here",
* The water boundary, which could be a river/sea bank, a building, a mean
high water level, and maybe other things.

The area and the boundary are both important for different reasons and
convey very different information. At the moment we handle boundaries very
poorly in OSM. This makes actually rendering the boundary very difficult,
because it will actually cross water areas in at least two cases: where
rivers meet oceans, and where river areas join together.

If these were all done using multipolygons we could separate the concerns
completely. So I propose to do this: map areas as multipolygons and
boundaries by ways which define multipolygons.

Water boundaries could be any of:
* natural=coastline, this should roughly correspond to the legal boundaries
of the country, ie. the coastline is the edge of the country,
* natural=riverbank, for rivers, usually these are managed separately from
sea banks, (we can't use the unfortunately named waterway=riverbank,
because that actually defines a water area, not a river bank),
* natural=shore, for lakes and anything else?
* anonymous ways, for cases where the bank is not very significant, this
can just be mapped as a normal polygon, for example a pond,

These boundaries can coincide with orthogonal tidal features which I also
propose:
* tidal:mean_high_water_spring=yes,
* tidal:mean_low_water_spring=yes,
* tidal:... for other levels of interest like astronomical low (for
nautical charts),

Note that these can exist where no bank is present, e.g. on beaches.

These boundaries can also contain many other useful pieces of information
such as: material of bank, man made or natural, mooring, ownership,
management and probably many more things.

It could be quite possible to have a way in which all of the following
features coincide:
* natural=coastline,
* tidal:mean_high_water_spring=yes,
* tidal:mean_low_water_spring=yes,
* man_made=sea_bank,
* material=concrete,
* barrier=wall,
* mooring=private,

It's also quite possible to have a completely distinct seabank, high water
level and low water level, e.g. in areas of reclaimed land where a man made
sea bank is a flood defence and is well above mean high water level.

Sorry for such a long post. I would like to identify any problems with what
I'm thinking quickly before I proceed to make proposals for these on the
wiki. So if I could ask people to just give a rough thumbs up or thumbs
down to this kind of approach I would be grateful.

Thanks, and happy mapping,

Borbus.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openstreetmap.org/pipermail/talk-gb/attachments/20190804/6de4493e/attachment-0001.html>