Boundaries for waters which does not seem to have well-defined borders

I could not find a relation for the Black Sea which at first suprised me.
But then I thought it maybe is because there is no distinct boundary since it is connected through other oceans and eventually the Mediterranean.

On the other hand there is a polygon (OSM relation) defining Lake Michigan even though it is connected to Lake Huron.
That polygon seem to be closed at an arbitrary position just to the west of the Mackinac Bridge:

In a similar way, a polygon for the Black Sea could be closed where the river starts (i.e. the river south towards Istanbul)

Are there any general rules for when it is okay to create a “virtual” boundary to be able to close a polygon when there is no exact boundary?
Maybe some local authority (e.g. US government or US state Michigan) has defined an exact border where OSM has closed its polygon,
but Turkey (or United Nations, since the Black Sea is located at many countries) has not defined exactly where the boundaries are for the Black Sea?

Just an opinion… Something mapped approximately is better than something not mapped at all.

In the case of your Lake Huron example, someone chose a perfectly reasonable line between two headlands, and I would’ve done exactly the same, based on nothing other than the general shape of the shores. And… adjusting this likely would make zero difference to any OSM-based map that I can think of. The name renders in the right place, give or take a millimeter.

I’ve mapped lots of bays on big lakes using the “eyeball is good enough” principle. :slight_smile:

PS, If you’re a new mapper (welcome!), please be careful about editing major features; a little mistake can break the map in a big way. I see the Black Sea is mapped as a node, whereas it would make more sense (to me) to map it as a relation combining the shorelines and islands, but even I would hesitate to touch that, given it’s not my area, and it’d be a big change.

Creating polygons for huge water bodies is not recommended. Create a node instead. A multipolygon relation has many disadvanteges:

  1. You cannot map something fuzzy in OSM, you can only map precise things. This will mean that you will have to invent/guess boundaries for your polygon which suggests a precision that does not in fact exist.
  2. The polygon relation will automatically be edited whenever someone splits or joins a coastline segment and it will be downloaded whenever someone downloads a piece of the coastline. This slows down and complicates the editing process for everyone.
  3. Huge polygons like that can even clog processing pipelines (and have in the past occasionally led to tile servers not updating properly). While this admittedly is a technical issue, it is a clear sign that OSM’s tools are not well suited for handling huge polygons like that.

Counter-argument: A node gives a renderer no clue about the size of the water body, thus about whether to render the name at a given zoom level. Some renderers are smart about this, when size can be determined from a relation.

Also, the location of a node is perhaps more arbitrary than, say, a straight line between headlands to demarcate a bay.

Anyway, I’m sure this has all been argued to death (I found a few spots), but haven’t yet stumbled across a definitive rule, eg how huge is “huge”.

TLDR: I wouldn’t mess with it.