Finally done! Split the Chesapeake Bay off from the atlantic as its own water relation, rather than part of the ocean. This collaborative project took about a week.
Closed about 1 year ago by ElliottPlack
This change breaks the whole world map! The Chesapeake Bay is a bay but it is still part of the sea, not an inland water area. The bay is clearly tidal, clearly part of the sea, so the natural=coastline tags have to go around it. On many maps this huge area will now not be visible as water any more, because many maps don't render inland water until larger zoom levels. This change is also blocking updates to *all* coastlines around the world, because the automatical coastline tool has noticed this huge change in the world map. See
Jochen, that sounds like an issue with the apps that consume OSM, not the data itself. There are other, larger inland water bodies that are tagged this way, like Lake Michigan, which is 3x as big as the bay. https://www.openstreetmap.org/relation/1205149
I had removed much of the Chesapeake Bay's tributaries about 4 years ago and no one has said anything yet.
I have read that apps like MS Flight Sim don't recognize water relations, but that'd be tagging for the renderer... The coastline is correct, it follows the bathymetry of the Bay.
Lake Michigan is an actual lake, not part of the sea, so it doesn't apply here.
You can not just change what a tag is supposed to mean that has been around for years. Such changes break OSM. You just have to look at a map like the cycle map: https://www.openstreetmap.org/#map=4/43.13/-94.83&layers=C It doesn't show the great lakes in small zoom levels and it will not show Chesapeake Bay in the future any more of this change stands.
I am not sure what you mean with "follows the bathymetry of the Bay". The wiki defines the coastline as "The natural=coastline tag is used to mark the mean high water springs line along the coastline at the edge of the sea." By that definition your coastline is way off.
The Chesapeake Bay is not part of "the sea", it is a defined area of water that is specifically delineated from an ocean. If it is not rendered appropriately, then the renderer should be fixed. Under your definition, because it is tidal, many rivers should be considered seas which isn't correct.
Who are the people who have decided on this "collaborative project" - is it just the two of you, or does this have the support of the US community? Has it been discussed anywhere that has a publicly accessible archive (i.e. specifically *not* Slack)?
Some of these discussions have taken place on changesets, but mostly in the OSMUS Slack, which I thought was a fine place to have such conversations.
The reason we considered this is because new users kept breaking the coastline by drawing it on arbitrary lines or one ways already covered by water. Some of these users claimed to be coming at it from the MS Flight Simulator use case, but I wonder how it works if they fly over the great lakes in that app. We thought that by simplifying the coastline out where the salt/fresh transfer occurs, we'd prevent these kinds of edits.
Here's one of those changesets. https://www.openstreetmap.org/changeset/93828561 The user never responded so I went ahead and rolled it back, then started doing it the proper way, setting up water areas for all the named sections of bay.
Back to the point of the matter, the Chesapeake Bay is an inland waterway and not a sea like an ocean, the Gulf of Mexico, or the Caribbean Sea. It is vastly smaller and surrounded by land.
Sorry, Sparks, but it doesn't matter for OSM what something "is". All that matters is that we agree on tagging, so that everybody can use the data consistently. And the agreed definition on where natural=coastline is tagged is in the wiki and has been used for a long time. You can not change this without breaking a whole lot of usage of the data. You are free to invent new tags that mean a different thing, but you can not change existing definitions. (Well, you could, but in something like this it would take quite a lot of community discussion and a worldwide consensus which would be very hard to reach.)
I'm not changing the definition of anything. I'm defining the "thing" that I am mapping.
Is it possible to have it both ways? I realize it isn’t an exactly analogous situation, but San Francisco Bay appears to be mapped as both a bay and as a series of coastlines, and it seems to have been that way for a long time without breaking renderers and other data consumers:
Nothing about the coastlines explicitly claims that the San Francisco Bay is part of the Pacific Ocean, so the only semantic challenge might be some redundancy between the bay and its coastlines. But that could be resolved by making the coastline ways part of the bay relation.
In the case of the Chesapeake Bay, that would just be a matter of adding the coastline tags back to the ways surrounding the bay, but otherwise retaining the (useful!) bay relation.
I like the idea of having it both ways. That way we could still have this relation and the level of detail around all of the water ways (instead of natural=bay points). Jochen, can you check to see how SF Bay looks on the renderer/data app you are using? I think we can reach an amicable solution for everyone here.
Looking at an application that uses OSM data, the SF Bay is identified as North Pacific Ocean because it is improperly mapped using coastline and only, what I suspect, has a node dropped in the middle of the area, saying that it is the Bay. That's the problem with mapping these inland waters like this. Using "coastline" and then dropping a label in the middle is great if you're making a paper map but if you're using software to actually use the data you actually have to identify the what those areas are and the SF Bay is no more the Pacific Ocean than the Chesapeake Bay is the Atlantic.
Which application did you check? The San Francisco Bay Area relation at https://www.openstreetmap.org/relation/9451753 includes all the coastline ways as members, so it forms a polygon. If the application represents it as a point, it’s probably calculating the centroid automatically.
Ahh, I see the relation, now. However, the use of coastline is still bringing in the ocean as that is what coastline is for, IMO.
I'm looking at YAAC which has the side effect of pulling way identifiers from OSM data and displaying them to the user.
(Correction: the San Francisco Bay relation, 9451753, includes ways redundant to the coastline ways, but not the coastline ways themselves. I’m unsure about the history behind that approach, but it does seem to result in a correct representation of the bay.)
Do the named ways of the relation affect how those apps display the info? See how this coastline has a name? (It was like that before I edited it yesterday) https://www.openstreetmap.org/way/591823494#map=14/37.8278/-75.4774
The coastline ways shouldn’t be named. The name on way 591823494 seems to be an error left over from when https://www.openstreetmap.org/changeset/57065106 introduced https://www.openstreetmap.org/relation/8099409 .
I personally don't have an opinion on any tagging here except that natural=coastline should be where it used to be. If you want other tagging for naming the Bay that's fine with me.
For reference: Previous discussion on the matter can be found on
For several years now this region has required special processing for anyone who wants to derive a consistent coastline data set from OSM data - for map rendering or other purposes. In fact the whole US east coast is now globally by far the area that deviates most from the global consensus regarding coastline placement (as measured by the number of cases where it deviates from that, not by surface area of course, here the Rio de la Plata has the undisputed lead).
The discussion here seems to lead towards a reiteration of the discussion regarding the Rio de la Plata we had a few months ago on tagging. I would therefore like to point to the question i had asked back then already:
These riverbank polygons:
are mapped incorrectly because
(a) the name of a river by broad consensus goes to the waterway line (waterway=river), not to the riverbank polygons.
(b) some of these don't even have a waterway line mapped - which by consensus is mandatory for a river.
To all that are following this changeset, I wanted to point out that there is now an ongoing discussion in the tagging mailing list.
I am reviewing that, and the comments here--I haven't forgotten about this.
The way the natural=coastline tags have been understood for the last >10 years clearly say that the coastlines tags should go around the bay. Some of the details of the discussion on the tagging mailing list do not matter for the situation here. Are you going to revert this now?
No. I disagree with your interpretation of coastline and can only say that it's time to fix this interpretation that you've been propagating for the last ten years. Legally, scientifically, and locally, the Bay is neither ocean nor sea. There are better ways of tagging this and that's what we've done. So, instead of armcharing this from many thousands of km away, how about you let us that actually live here define what's here.
You are breaking >10 years worth of software, style sheets, maps, etc. Everybody who has dependet on this definition. This is not something you can change.
ElliotPlack, regarding the question of Slack: Slack is a private communications medium with no public archive that can be viewed by other members of the community who are not signed up to Slack. Of course mappers may use any form of private communication but if the result is a grand plan like this, it is worth at least outlining it in a form that is accessible and archived even for those who are not signed up to Slack, ideally on the talk-us mailing list.
There are users who use the coastline for rendering (although they do not have the nobility to recognize it) and they care less that the data is used to determine what is ocean and what is not, as shown by the various examples of the coastline mapped tens of kilometers within England or Germany.
They are just tagging the coastline for the renderer!!! JUST FIX THE RENDERER!!! The OSM is data, not a renderer image for the web. Render is one of the uses of the data, but it is not the only only one, nor the most important, but they fails to understand that, and they continue to insist on moving the coastline, as they did with the Rio de la Plata and it is currently very poorly mapped.
The question about what is ocean and what is not is a basic question that OSM data cannot answer with the natural=coastline that some people insist.
I suffered the same fact with the armchair mapping in Rio de la Plata, seems that the local mapper knowledge is less valuable than the opinion of these people. This is a fact right now. The problem has more than one year now.
I'm still waiting since July to see the DWG position on this issue.
The coastline sould be mapped where it should , not where it use to be, if the software is broken fix the software, but the data sould be right.
Log in to join the discussion
OpenStreetMap is a map of the world, created by people like you and free to use under an open license.
Hosting is supported by UCL, Fastly, Bytemark Hosting, and other partners.