NOTICE: is no longer in use from 1st March 2024. Please use the OpenStreetMap Community Forum

River-lines (waterway=river), per definition, should never be circular (simple closed polygons). Yet, there are thousands of them in some strange combination with river and lake areas. These lines usually form a long thin polygonal shape, sometimes overlapping lake and/or river area borders but often with no matching objects in the mentioned area classes. While in the raster map-making these lines doesn’t matter too much (eventually some area brakes, strange name positioning…) in the vector map-making it does. Note that a very similar case we have with the canal-lines vs. canal areas. So, my question is:
Is there any OSM data preparation tool that uses these circular river-/canal-lines for (vector) data corrections?
I would be thankful for any references. Thanks in advance

asked 10 May '15, 20:05

sanser's gravatar image

accept rate: 5%


Do you have any examples

(10 May '15, 23:11) DaCor

Of course. Otherwise I would never write/asked something like I did. If you want, you can get the shp files with large number of cases (7536) from the dump from some weeks ago. In my experience, changes/corrections there go pretty slow.

(11 May '15, 09:55) sanser

A link to a circular river way in OSM would be useful.

(11 May '15, 09:59) SomeoneElse ♦

If I understand correctly, the question is about what Osmosis calls "Closed waterways":

Some are just bad NHD imports:

A few I looked at were ditches:

(12 May '15, 22:13) maxerickson

Not quite. The question was about using these anomalies to correct the vector data in a data preparation chain. There were questions about examples. Your list shows many, thanks. I have also uploaded many thousands to my G-repository for the interested here: If interested, please start with the "Circular_river_lines.docx". Thanks.

(13 May '15, 07:22) sanser

@sanser that document has lots of text and pictures but no links. Can you provide a link to a circular waterway=river in OSM (preferably to the one of the ways, like maxerickson did above)? The "osmose" issues seem to be a mix of false positives (osmose not understanding the tagging), circular parts of drainage networks with no obvious direction, and simple tagging errors.

(13 May '15, 15:59) SomeoneElse ♦

one of the first in the document is the river ways around the island linked in maxerickson's answer (

(13 May '15, 16:26) neuhausr

That's a riverbank, not a river. One of the river ways there is not circular:

(13 May '15, 17:23) SomeoneElse ♦

which is why I said "river ways around the island". sorry, should've not been lazy and linked directly to those ways (3215427 and 48856760 on the north side, 30587867 on the south)

(13 May '15, 18:45) neuhausr
showing 5 of 9 show 4 more comments

It still isn't clear what exactly you're asking. If you're asking if closed-way waterway=river is correct tagging or not, the answer is that in most cases (if not all) it would not be correct tagging. The same would apply to the other waterway classes which are assumed to always flow in a single direction (stream, brook). If you know of incorrectly mapped waterway=river, stream, or brook, the answer is to fix the data. If there are a lot of them, it could be a candidate for a MapRoulette challenge.

Drains and ditches often don't have a flow-direction and mapping them as closed ways is common.

Canals are a grey area that could go either way but could conceivably exist in a closed loop.

A closed way of any of the above waterway=* values should never be interpreted as an area. They are always vector ways.

permanent link

answered 29 May '15, 17:11

alester's gravatar image

accept rate: 28%

I think you're misunderstanding waterway=river. These are vector lines representing the channel of a waterway, so they might be traced to go on both sides of an island. If the direction of one of these ways was opposite, I could understand thinking of them as circular, but since they both head "downstream," that makes topological sense and follows OSM practice.

As pointed out separately, the island is accurately modeled in the waterway=riverbank version of the Thames.

I'm not sure how you're doing your analysis, but it might be convenient to use river's waterway relations when they exist, such as

permanent link

answered 13 May '15, 15:13

neuhausr's gravatar image

accept rate: 21%

I believe it is more a misunderstanding between us than what you have suggested about such a simple/basic notion as the riverline. Essentially, my question was related to something else, as you could see. But, you have a point regarding the analyses and the riverline relations. I have added some more/new notes under P.S. in the referenced document. These notes are related to your two points but also to some earlier useful comments. If you or others are still interested in the circular river-/canal-lines you are welcome to revisit the document.

(28 May '15, 23:11) sanser

Your document is really long and dense--I think you'll have better luck getting help if you can summarize your issue so that it is clear. For example, "the data seems incorrect at this place for this reason". Or, "when I try to render the data at this place in this way, I am having these problems". I'm sorry but I'm not going to read pages and pages.

(29 May '15, 16:15) neuhausr

FWIW I tried reading "pages and pages" but communication did not occur.

(29 May '15, 16:44) SomeoneElse ♦

I just tried reading the "pages and pages" again, and I think the question is really, "How should my vector renderer deal with incorrectly-mapped waterways?" There's a ton of other stuff in there, though, so it's nearly impossible to decipher an actual question.

(29 May '15, 18:14) alester

waterway=riverbank should generally be an area tag, here is the missing island from one of your early examples:

It appears to be the same object as your anomaly.

The riverbank has the tags on the outer part of the multipolygon, which is maybe deprecated, but you will want to interpret it correctly:

The surrounding waterway=river ways look like they make sense.

It is also the case that a long thin area might be correctly tagged as natural=water.

I think if you account for waterway=riverbank being an area, many of your anomalies will go away. I fear many of the remaining anomalies will be more nonsense than anything else (of course depending on what you have included as anomalous).

permanent link

answered 13 May '15, 13:07

maxerickson's gravatar image

accept rate: 32%

What makes sense or don't depend on the criteria and criteria is subjective (somebodies). When a river-line is circular, surrounding a lake area object, between two river-area sections (even aving "river" in the name) to me makes sense to call it for anomaly but probably not to you. But my question/help-request was something totally different. If you have time, you can find some more/new notes under P.S. in the referenced document.

(28 May '15, 23:11) sanser

Follow this question

By Email:

Once you sign in you will be able to subscribe for any updates here



Answers and Comments

Markdown Basics

  • *italic* or _italic_
  • **bold** or __bold__
  • link:[text]( "title")
  • image?![alt text](/path/img.jpg "title")
  • numbered list: 1. Foo 2. Bar
  • to add a line break simply add two spaces to where you would like the new line to be.
  • basic HTML tags are also supported

Question tags:


question asked: 10 May '15, 20:05

question was seen: 4,279 times

last updated: 29 May '15, 18:14

NOTICE: is no longer in use from 1st March 2024. Please use the OpenStreetMap Community Forum