River areas modernization

Hi everyone,

We are working on a project that is trying to clean up and update river
data globally. In short, we look to fix errors like disconnected waterways,
overlapping water areas or waterways, correcting broken multipolygons,
fixing mistagged features, removing duplicate objects and importantly, also
moving older river tags (waterway=riverbank) to the newer more numerous
version natural=water + water=river.

More details are written up here:

Let me know if there are any possible issues with running the modernization
in Austria.


Kein Einspruch riverbank nach river zu ändern, rein persönliche Stellungnahme.

No objections regarding riverbank to river transition, just my personal point of view.

The bold statement might though find some opposition on the talk.at mailing list. You MUST post there too.

I just posted there as well. thanks


First of all: Thanks for cleaning stuff up. A few years ago I did some cleaning up of the NHS (? was it called that way) data in the US - it was a pain.

That being said:

  • Tagging of unspecified water areas: There’s more than I would have thought. An effort to add water-keys to them seems beneficial, but I recommend using all available basemap.at-layers. When in doubt - leave untagged or add a note so that people with local knowledge can take a look into it.
  • I have no general objection regarding the retagging from waterway=riverbank to natural=water+water=riverbank. However, it seems to affect around a thousand ways/relations in austria. Most of them have probably not been touched in a longer time and could benefit from a geometry update and a little love around them. Handling these thousand objects seems like something we could handle in an orchestrated effort while generally improving map quality. Would anyone be interested to join on that?
  • Canal search: As far as I can remember this was a problem in the NHS data - but I doubt that’s a problem in austria. I think it’s best to skip this step here or to provide at least some examples beforehand.
  • Duplicate nodes and other validation: (My) JOSM reports around 130 errors, and 3800 warnings after validation of the “water=river+waterway=*” query. Duplicate nodes were a big problem in the NHS data, the ~100 errors of waterway-duplicate-nodes could probably be handled by the local community now that we’re aware of it.
  • Conflicting natural=* and water=* tags: Your wiki pages states: “For example, if an area is tagged water=reservoir and waterway=riverbank, decide whether it is a reservoir or a river area, and update the tagging accordingly.”: There might be cases were the distinction is clear, but in some cases I would have trouble making my mind up on which is which. This should imho not be decided by someone without local knowledge if it’s an edge case.

As a summary:

  • Retagging to water=river is fine with me, but if someone wants to join me I think we could manage these without an automated edit.
  • Adding water=* is fine with me, but I’m wondering how cases where it’s not possible to tell from aerial imagery are handled. Could you explain this further?
  • Modifying existing data or removing conflicting tags… I’m not sure how conflicts would be resolved in case of ambiguities. How many ways/relations have conflicting tags?

Zwecks der Vollständigkeit:

  • Ich denke das Umtaggen von waterway=riverbank zu natural=water+water=river ist sinnvoll. Es betrifft etwa 1000 Wege/Relationen in Österreich. Diese sind wohl wahrscheinlich großteils schon bisschen älter. Wenn wir uns für das Umtaggen entscheiden wäre ich dafür, dass wir das gemeinsam machen und dabei gleich Geometrie und andere Kleinigkeiten mitnehmen, erscheint mir von der Größenordnung her nämlich machbar.

Wäre das als MapRoulette Aufgabe lösbar? Würde mich auch gerne beteiligen, wenn es organisiert abläuft.

Thanks for the reply!
We definitely support involvement of local people so if you decide to do it yourself that is fine by me.

To answer some of your points:

  • Conflicting natural=* and water=* tags: if the distinction is not clear, it is fine to leave just natural=water. That is also one of the advantages of the new tagging scheme.
  • Adding water=: This is pretty much identical case. Leaving natural=water without adding water= tag is acceptable.
  • removing conflicting tags: This luckily happens rarely. But for example I’ve seen objects tagged as waterway=riverbank + natural=wetland. In that case I just leave natural=wetland as that is pretty much a definition of an area that gets flooded and turns into river while other time water might be absent.

I walked recipe 1)

out skel qt;

Adding to area size proved not a good measure to tell lake from river. Manually purged from 25000 on. Remains after some minutes:

  • 17 ways “natural=water”, 11 of which “waterway=riverbank”
  • 1 multipolygon “waterway=river” (only this one errors in JOSM validator)

What is your count? Wondering, how the number of 1000+ for Austria, stated above, came to happen? There are lots of lakes not tagged as such…

I am going to change the ones found, seems small fish, but will wait for your number, in order to learn what else is in need of “mordernization”

Interestingly, a query for “waterway=riverbank” brings up much more objects. I rather not touch those, quite often, these areas include lots of shingle, definitely not a case for “modernization”, but the ~20 items found so far indeed are tightly mapped areas where there is water only.

MapRoulette scheint mir nicht geeignet, eher ein StreetComplete Quest: Ist das ein See, ein Teich, ein Bassin :slight_smile: Aus dem Luftbild kann das manchmal leicht, manchmal unentscheidbar sein.

DIe knapp 1000 kamen bei mir über den Step 2, mit anschließendem JOSM-Filter auf “waterway=riverbank” - am Ende noch unselect nodes und ich war bei etwa 1000 ways.

Regarding conflicting tags/updates: that seems like sensible defaults where no data is being destroyed so I’m good with it.

Ahso, das holt aber ganz schön viel! In Tirol sind 25 Relationen und 66 Flächen Riverbank. Ich werde das sicher nicht machen, die auf Water umschreiben - da sind zu oft Bachbetten mit Schotter usw. erfasst. Das rendert falsch, und mit dem umtaggen wird das nicht besser.

There are 25+66 waterway=riverbank in Tyrol. I advise against mechanical retagging. Step 6 Upgrade tagging will make things worse. Every object has to be examined. Not a few times, shingle is better than water.

One solution to the shingle problem is to have two overlapping geometries: one with natural=shingle and the other with natural=water.

I fixed that; All the riverbanks around streams in Tyrol, that may carry enough water, to fill their bed, over a short period in spring or after heavy rains only, have been upgraded to shingle. Some cases might be up for debate, but the waterway line can make up for the amount of water even then. Feel free to upgrade the remaining ones to water areas. Some particularly hard cases have been left for you to decide :slight_smile:

Update: Days ago, two more “riverbanks” areas were mapped. Both around a stream, where the aerial shows just shingle. Using the JOSM editor. Indeed, when one searches templates and types “Flussbett” one gets riverbank. Some lobbying seems due?

Thanks everyone for contributing to the modernization project and also for bringing the issue to attention of JOSM devs!