[Tagging] Rivers classification

Daniel Koć daniel at xn--ko-wla.pl
Sun Aug 6 22:55:51 UTC 2017

W dniu 07.08.2017 o 00:15, Lukas Sommer pisze:
> Both, Wisła (http://www.openstreetmap.org/relation/34392) and Czerwona
> (http://www.openstreetmap.org/way/269561538) have the classic stream
> order “1” just before they enter in the ocean. Likely, the former you
> would like to render on low zoom levels, the latter not. So I don’t
> understand how the stream order number could help here.

order:classic=1 can be a basic filter - if there's a river with "2", we 
can safely skip it and take the next one link - river that goes directly 
to a sea, because it's (by definition) bigger. It's a relative system 
which can be fine tuned with some other data.

> Note that for
> Strahler that’s a similar problem: “Higher number” means “yet more
> other rivers that have joint the current river and now it’s bigger
> than before”, but that says nothing (in absolute terms) about how
> large the river is, neither about how important it is neither about
> how long it is and neither about it can be used by big ships. For each
> river system, you would need an own threeshold to make useful
> rendering. I see this awkward…

We already use some custom thresholds on osm-carto for rendering cities 
or filtering out small areas for example.

But the bigger problem with "top down" approach is that it may need 
recomputations and data about the waterways network can be hard to find.

> Something more similar to the roads hirarchy is CEMT, because it has a
> fixed and limited number of allowed values and is directly related to
> (ship) traffic, while stream order numbers by definetion are integers
> that do not have a maximum limit and don’t tell is if big (or small)
> ships can use this river. Wikipedia says, Shreve correlates more or
> less with the amount of water, so this might be the only stream order
> number that might help us. On the other side, it will likely correlate
> also with the yet existing width=* value.

CEMT is limited to Europe, as you know, but we can also use this as an 
additional data. There are also some other data I think we could use:
- width at the river mouth
- length/area of the basin (they are correlated - see: 

It all needs some testing, but we need some values to be tagged first.

"Like a halo in reverse" [M. Gore]

More information about the Tagging mailing list