Proposal talk:Waterway=riverbed

From OpenStreetMap Wiki
Jump to navigation Jump to search

Desert rivers and streams

These can have no water in them for years. Yet the river bed could be mapped. It this tag is to be used for such rivers and streams then the wording needs to be changed to accept these normally dry water ways. Warin61 (talk) 07:42, 8 September 2018 (UTC)

Not completely decided, historic riverbeds that may be visible centuries after the river ceased to exist should perhaps get an own landform tag. RicoZ (talk) 18:49, 25 September 2018 (UTC)
Not historic. The rivers still exist but only have water ephemerally. Possibly once every 5 to 10 years. So still an 'active' riverbed. Warin61 (talk) 23:12, 25 September 2018 (UTC)

What about engineered rivers?

"encompass the whole area that is regularly flooded and visibly formed by the river" - what about engineered rivers, with levees? Is it supposed to be also used here? Note that "visibly formed by the river" is a poor phrasing, it includes usually at least entire valley (but I have currently no idea for a better) Mateusz Konieczny (talk) 21:00, 9 June 2019 (UTC)

The area within the levees would probably constitute a riverbed and if the water is frequently filling only a part of it I would use it there. However not in the case of a mostly natural river with flood protection dams that are projected to protect against 100-1000 year floods. Agree about the wording. RicoZ (talk) 21:46, 9 June 2019 (UTC)

regularly flooded/regular flow

I would expect riverbed to cover area that is regularly covered by river, and not area that is regularly flooded. There are some regularly flooded areas (including residential areas built in stupid places) that are not parts of riverbed. In other words - riverbed for me would be something covered by river after big rain, not during a large flood Mateusz Konieczny (talk) 08:57, 10 June 2019 (UTC)

I have added "A classification of natural rivers, 1994 David L. Rosgen" as reference to the page and hope to make a useful definition according to "4.1. Geomorphologic characterization" therein. I would simply take Fig 1 and use what is depicted there as "flood prone area". As apparent here the exact definition slightly varies by stream type/section but it seems useful and intuitive. Also the classification developed in that section could be a nice additional benefit.
If we could agree on that the only problem would be to get a "free" picture that could be used in this page. RicoZ (talk) 22:00, 10 June 2019 (UTC)
Is it feasible to find some photo of river valley and mark there zones from that image? Can you maybe publish this image temporarily somewhere (on your web site, upload to some cloud drive, upload to image sharing site and later delete it etc)? Mateusz Konieczny (talk) 05:33, 11 June 2019 (UTC)
Re: ""flood prone area" - how are mappers going to verify the flood-prone area? Is this the highest average annual water level, or a "10-year" flood level, or the "100-year-flood" line used by some insurance companies and governments? All of these are statistical properties based on historic records, modeling and assumptions that individual mappers aren't going to be able to do themselves. This definition would make the tag non-verifiable.--Jeisenbe (talk) 11:08, 9 July 2019 (UTC)
By Rosgen, the floodprone area is defined as "the width measured at an elevation which is determined at twice the maximum bankfull depth" and is linked with 50 or less year floods "by observation". In addition the entrenchment ratio (width of floodplane / bankfull river width) and riverbed type make it possible to estimate how "well developed" (or rather well defined) a flood plane probably is. This is of course all fairly academic but there are field guides out there how map those things. Also, no matter how well defined it is on an academic level it is quite foreseeable that mappers will map it rather arbitrarily - but should we abandon it because of this? Anyway thanks for asking questions.. I noticed a few more important details after rereading Rosgen. RicoZ (talk) 18:20, 11 July 2019 (UTC)

Much of my mapping is done in Alaska where the riverbed tagging scenario is very helpful because most Alaskan rivers are wild, that is they have not been channelized nor do they have flood control. Consequently, their volumes change dramatically between the seasons. The larger rivers have massive flood plains with extensive gravel or silty areas that are quite dry in the winter (when they're frozen) and summer but totally covered with fast-moving glacial water at other times. I started out mapping the Delta River in central Alaska using the waterway=riverbed tag but later changed those to use the much more common natural=riverbed tagging. But now I'm wondering if that was a good idea.

I much prefer the waterway=riverbed tagging because that allows one to use natural=sand and other natural=* tags to better describe the riverbed when it is not covered by water. The waterway tag seems more consistent too because while they are natural features the riverbed is also an important part of the waterway. Anyway, which high-level tag is it to be? I'm in the process of mapping or remapping other Alaskan rivers but would like to resolve the question of which tag will be accepted for future use.

I think either would be workable because if one wanted to define areas of sand or gravel in a riverbed, the surface tag would also do that job. Opinions, ideas? AlaskaDave (talk) 07:30, 9 June 2020 (UTC)

waterway=riverbed would be better for the reasons you mention and also for the reason that natural=riverbed has been used in two inconsistent ways before it has been documented.
I wish I had the time to finalize this proposal - the main issue is that many people will cry "not verifiable" which should be fixed by a more detailed reference and description of the widely used Rosgen classification with a practical guide... someone who has academic or practical experience with this would highly welcome. However if you feel confident to use it as is, do it. --RicoZ (talk) 18:15, 9 June 2020 (UTC)
Many waterways in Alaska are so irregular that verifiability is a practical impossibility. Their maximum extent (that is, the extent of the riverbed) is fairly easy to see in satellite imagery but the waterway itself, the part that would be tagged waterway=riverbank, might change wildly from year to year. If mapped in the standard way, in which the entire riverbed is tagged with waterway=riverbank, many rivers in Alaska would appear to be hundreds of meters wide in the summer on your map or GPS when in reality the waterway is only flowing in a small stream or streams in the distance, often completely out of sight of the highway or trail. This not only looks weird but it doesn't represent the reality on the ground.
Putting aside the issue of verifiability, the major problem I have with the waterway=riverbed tag is that it has few so uses compared with natural=riverbed. (68 ways vs 484). If I include the riverbeds I've added in Alaska so far, that would add 320 more ways along with 50 relations to the totals. So, in a sense, my choice will add weight to the argument for using one or the other alternatives going forward. AlaskaDave (talk) 00:06, 10 June 2020 (UTC)
Look closer: a very high number of the uses comes from imports in Italy such as https://www.openstreetmap.org/changeset/16733635. Furthermore this import is mapped "the wrong way".. in a braided river every little "island" got its own "natural=riverbed" entity, resulting in a relatively high but meaningless taginfo count. Tried to reach the mapper years ago and was hesitant to fix it without consent but seeing that I get no reply I may do something with it now. RicoZ (talk) 20:14, 10 June 2020 (UTC)
Last looked at that particular edit in 2017 - now it looks like OMG.. what could be done to fix it other than delete all of it? RicoZ (talk) 20:55, 10 June 2020 (UTC)
I agree, that's quite a mess. Unfortunately, many braided rivers show this sort of variation, especially over a period of years. That's why I said that verifiability is a practical impossibility. I've worked on braided rivers a lot and can see changes every time there's new imagery available on OSM. You can sometimes see differences even when switching between Bing, Pixar and Mapbox imagery. Aside from that, and ignoring user:DarkSwan_Import's use of natural instead of waterway for the riverbed, it looks like the sort of thing I've done on the Delta River https://www.openstreetmap.org/changeset/83697700#map=11/63.8955/-145.8620 Keep in mind that my tagging will likely change to back to use the waterway=riverbed tag as a result of this discussion. Thanks for the helpful comments. AlaskaDave (talk) 01:30, 11 June 2020 (UTC)
Looking at your example I am wondering if we have the same idea how to define and map the riverbed: for me the riverbed covers all of the area covered by the river and the surrounding sand/gravel areas whereas in your example it appears you exempt the actual river from the riverbed area?
Unfortunately I just noticed that some helpful editor "corrected" my examples of waterway=riverbed usage in the Valle Maggia so I don't have good examples right now :( RicoZ (talk) 21:37, 11 June 2020 (UTC)
I have tried to recreate the examples : https://www.openstreetmap.org/changeset/86535599#map=15/46.2855/8.6462 . It has been somewhat damaged by subsequent editing and I couldn't fix that yet. RicoZ (talk) 22:06, 11 June 2020 (UTC)
I thought it was the same idea but I decided to "exempt" the riverbank (that is, the river proper) from the riverbed because I thought that was the only way to map it. I had been thinking of the riverbed as only the parts that were intermittently flooded, which I now realize isn't really correct. The way I mapped them in my work involves complex relations and is a ton of work. Your model wherein the riverbed extends from one bank to the other makes a lot more sense and is much simpler to map. Thanks for the example; it looks good. I will rethink my methodology going forward. AlaskaDave (talk) 07:57, 12 June 2020 (UTC)
Yes, making it easy, and the results well defined was an important design goal. So the riverbed should cover the complete area including the actual water flow(s) and other areas such as vegetation or bare_rock. Only "true islands" should be exempted by multipolygons. Should I somehow amend the description to make it clearer? I realize a picture would be better but don't have one right now.. do you have an idea how to make such a picture from JOSM or otherwise? RicoZ (talk) 10:55, 16 June 2020 (UTC)
Well, the problem with making a representation of the larger Alaskan braided rivers is that there is no way that I know of to overlay anything other than a simple waterway=river way on top of a riverbed multipolygon. These rivers typically have a wide channel of flowing water in all seasons, in other words, they need a riverbank relation as well as the riverbed to fully show their properties. If one tries to draw a riverbank multipolygon on top of a riverbed multipolygon there's going to be problems. That's a major reason I chose to map them the way I did, with separate multipolygons sharing their edges for each feature. Of course, if using that method one must give up the requirement in your model for the riverbed to extend from bank to bank. It's a real problem. AlaskaDave (talk) 12:10, 16 June 2020 (UTC)
Not sure I understand the problem. It was indeed intended that there is waterway=riverbank covering part of the area defined by waterway=riverbed (if it would be covering the complete area than the riverbed is probably pretty useless to map). It may feel "wrong" to do but is nothing new in OSM and would work even better if it is defined properly in this proposal. As of rendering, right now eg OpenCarto will draw the smaller area on top of the larger area so the desired result will usually just happen although we might be lucky and get some official support sometimes. RicoZ (talk) 17:00, 18 June 2020 (UTC)
That's interesting. You said: "It was indeed intended that there is waterway=riverbank covering part of the area defined by waterway=riverbed." If that's true it's not at all clear from the text in the How to Map section. It would appear I was wrong to assume there would be a problem superimposing one multipolygon on top of another. JOSM complains if any part of one multipolygon overlaps any other so I automatically assumed OSM wouldn't render such overlaps either. This knowledge puts a whole new spin on our conversation. I'll try mapping a braided river that way at my next opportunity. And I think you should amend the How To Map section to discuss and clarify the multipolygon usage. AlaskaDave (talk) 07:23, 21 June 2020 (UTC)
Can you suggest a wording.. even if it isn't perfect it would help me to understand where the text needs most improvement. Regarding the JOSM issue, not sure I understand what you are trying to do. The riverbed and river can be mapped with simple areas without any multipolygons, would the problem manifest even without the use of multipolygons? RicoZ (talk) 18:30, 11 July 2020 (UTC)
I added some new wording for the How to Map section to your personal page. Take a look and get back to me. We can work out more details as needed. I have used the tagging scheme as you suggested earlier to map portions of the Tsina River in Alaska. The relation is here: https://www.openstreetmap.org/relation/11269704#map=14/61.1942/-145.6415 You should look at it inside JOSM to get the full picture because the OSM slippy map doesn't render riverbeds correctly. In my experience, simple areas almost never work for either riverbanks or riverbeds. Also, a short aside. I'm a proponent of avoiding shared nodes so when I add a natural=wood that adjoins a water body, a very common situation, I don't double up nodes but create a relation so ways can be shared rather than duplicated. AlaskaDave (talk) 00:31, 13 July 2020 (UTC)

Additional attributes (riverbed shape etc)

The Rosgen paper (and others) provide additional classification/characterization of the streams and riverbeds based on various characteristics:

  • stream slope
  • channel shape
  • channel pattern (braided, meandering, straight)

So I would propose attributes like riverbed:slope, etc. The values proposed by Rosgen are A-G for most attributes which although widely accepted in the academic literature seems counter intuitive for OSM? What values would be useful? RicoZ (talk) 22:46, 15 June 2020 (UTC)

Hi there. Your proposal is interesting. I was just thinking the other day that I wished I had a way to tag alluvial plains. A question though, why bother with riverbed:slope when there is already the incline tag that works perfectly well on ways? also, riverbed:type seems unnecessary when you can just as easily use riverbed. my own experience adding the extra word "type" just over complicates things for lay people and leads to a lot of miss-tagging or confusion. Since the word type is pretty ambiguous. It's actually something I've look a lot into and that's what I've found. --Adamant1 (talk) 15:25, 16 June 2020 (UTC)
So instead of "riverbed:type=*" you would suggest "riverbed=*" ? I usually prefer the other variant in order to not introduce too many new top level tags and if there would be more attributes (like pattern etc) it makes more sense to used the variant with prefix.
The idea with incline is a good one, however having a dedicated riverbed:slope would allow people to put there values like "Aa+" or "D" where they don't have a good possibility to measure the slope in degrees but by appearance it looks like a river fitting into the "Aa+" category etc. RicoZ (talk) 17:36, 18 June 2020 (UTC)
I've looked at the original paper. Most of the classification system has to due with the incline of the river, which is available by using a DEM (digital elevation model) in concert with OSM data, or with the shape of the river, which is available from the geometry of the objects in the database. These characteristics do not need to be tagged. Furthermore, these are characteristics of the waterway, rather than the riverbed. --Jeisenbe (talk) 16:46, 16 June 2020 (UTC)
Indeed much of the classification is driven by, or strongly correlated to the slope/incline of the river. However I think current freely available DEM data is nowhere near precise enough at the moment to replace the classification everywhere in practice.
As of riverbed type, this one might be more or less dispensable if instead of riverbed:type we would use eg an additional natural=bare_rock or similar.
One of the reasons of this proposal was to avoid overly detailed mapping of fast changing river flows in wide riverbeds, hence the proposal of riverbed:pattern instead of drawing detailed river flows according to many year old Bing pictures.
So in some cases other data may provide sufficient or better information but in some cases this attributes may come handy. They could be applied to the waterway as well, although this would duplicate information where river arms and such are drawn. RicoZ (talk) 17:36, 18 June 2020 (UTC)
If DEM isn't sufficient to calculate the incline of the river, we already have the tag incline=* which can be added.
While we should seek to map the current, accurate course of the river, it is much better to map the precise course and extent of the river as it appeared 12 months in aerial imagery than to map only the maximum possible area which is sometimes covered with water, since that area will also change every year and it is much less helpful for routing (e.g. canoeing), rendering and analysis of water-cover area, river flow and so on. I would recommend that mappers do their best to map the precise extent of the river at high water level based on the best available survey and imagery data, and in the long-term the goal should be to keep rivers up to date with changes every year, or more often if necessary. So, mapping the total extent of the "riverbed" is also a good idea, but it isn't a substitute. This means that tagging the "riverbed:pattern" should not be considered a sufficient alternative. --Jeisenbe (talk) 16:58, 19 June 2020 (UTC)
While you can estimate the incline over sufficiently long stretches of rivers with GPS I think it might still be helpful to be able to write something like "Aa+" to mean ">10%" or A for "4-10%". I do not oppose the use of incline if someone wants to use it.
Regarding detailed mapping of waterflows in riverbeds versus specifying riverbed:pattern and the main stream I believe both should be possible. I would be against painting a braided river flow with many side arms in great detail when it is highly likely that the exact pattern will look very different next year. Some details tend to be very stable and those can be certainly mapped. Another aspect is that the detailed mapping simply takes a lot of work, sometimes the river would change faster than it can be mapped and the satellite imagery may not be available at all. So if someone want to use rivrbed:shape instead let him do it. RicoZ (talk) 18:40, 20 June 2020 (UTC)

How do we distinguish this from intermittent=yes river water areas?

Currently there are a moderately large number of waterway=riverbank and water=river areas which have the tag intermittent=yes. Some of these are intermittent rivers, where the river runs completely dry during some seasons (or most of the time), but others represent areas which are intermittently covered by water seasonally or rarely. E.g.: https://www.openstreetmap.org/#map=15/61.3362/72.0177

How should mappers determine when to use waterway=riverbed instead of waterway=riverbank + intermittent=yes?

I would think for most intermittent rivers a combination of both methods would be best. Normally I would expect there would be a visibly larger riverbed area and intermittent waterflows within it. Only in cases when the riverbed isn't easily recognizable or entrenched like type "F" it would make little sense to map the riverbed in addition to the riverbanks. RicoZ (talk) 18:53, 20 June 2020 (UTC)
Did you check my example link above? Do you consider that an incorrect use of intermittent=yes? --Jeisenbe (talk) 06:04, 21 June 2020 (UTC)
Very interesting example, creative mapping but I think nothing wrong with it. Riverbed has more possibilities and should be easier to achieve the same effect. With riverbed you can not only define areas of intermittent water flows but also how they look like when no water flows (sand, gravel, shrubs..). RicoZ (talk) 18:45, 11 July 2020 (UTC)