I recently came across a river with its riverbank tagged as natural=coastline for 12.2 miles (19.6 km) upstream of the actual sea. I spent some time and retagged the riverbank as a `waterway=riverbank', (this edit), only to be told that it was wrong for me to have done so.

To clarify, there is no dispute over the location of the mouth of the river, as the way tagged waterway=river is unchanged, including its termination point. Previously the waterway=river way passed, as it was mapped, through a narrow sea channel for its final 12.2 miles. I have changed it so that the waterway=river is within waterway=riverbanks.

Tagging riverbanks as coastlines is not something I've come across before, although I'm told it is the convention in the UK. In what situations is it appropriate? What are the advantages?

edit: If this is a UK convention, is it documented anywhere?

asked 29 Aug, 19:05

keithonearth's gravatar image

accept rate: 20%

edited 29 Aug, 23:31


The WikiProject United Kingdom front page has a link to "mapping guidelines", but it appears that the link has never actually pointed to anything since it was added in 2009. There is United Kingdom Tagging Guidelines, but it doesn't mention rivers or coastlines. I'd ask whether this "convention" is documented anywhere. As for the method they want to use, I could agree with it if the tidal effect reaches that far upstream, but only if the waterway=river way ended at the coastline. It shouldn't extend 12 miles "out to sea".

(29 Aug, 23:01) alester

@alester, thank you for your input. The tidal effects do indeed reach that far upstream, just as they do reach far upstream of any river. On larger rivers the tidal effects reach far more than 12 miles upstream. To me it makes more sense to differentiate between the sea from tidally influenced river, by using the tidal=yes tag on the riverbank trace.

Keep in mind that the tidal influence isn't sea water coming 12 miles up the river, but rather the fresh water being slowed by the sea water, and draining more or less quickly due to the tides.

(29 Aug, 23:30) keithonearth

I agree with everything you said. My main point was that one needs to choose where the coastline is and have the river end at that point. Having it end elsewhere is simply nonsensical.

(29 Aug, 23:47) alester

It was me who made the original changeset comment. I have no problem with the riverbank tagging, for that is what it is. However removing it from the coastline is IMHO a separate issue. Almost all other rivers in the UK have the coastline tagging up to the tidal limit, and changing this one river made it follow a different paradigm to the rest. Personally I prefer consistency, and a paradigm that follows existing conventions where they exist.

(30 Aug, 17:19) csmale

Hi @csmale thanks for the comment.

Prior to making this edit, I had not realized that it was common to use the natural=coastline tag for rivers in the UK. The two rivers I checked before making this edit both used the standard waterway=riverbank tag on their bank traces. I've done a slightly more thorough survey of the local tagging, and see that 7 out of 10 do tag the riverbanks as natural=coastline.

I also prefer consistency, but I do not think that having the bank tagged as sea-shore, and the centre line as a waterway=river is consistent.

I think the fundamental issue here is that I think the natural=coastline is for mapping the sea-coast, and not other things. Do you not agree that natural=coastline is for the sea-coast exclusively?

I think it would be helpful for you to add your point of view as an answer to this question, preferably with a link to some documentation for your suggested usage, if one exists.

(31 Aug, 06:23) keithonearth

natural=coastline should only ever be used for the coast of the sea, not rivers, not lakes. This is a very special tag that is handled differently than other tags in the processing chain that creates the maps. If you use it somewhere else, it will likely break this processing.

permanent link

answered 30 Aug, 07:12

Jochen%20Topf's gravatar image

Jochen Topf
accept rate: 31%


There is some debate on the wiki page about whether the coastline, which should be at mean high water spring level, should be used strictly at the mean high water spring level (so may include the tidal part of rivers where they reach the sea), or should just cut across river's mouth at some arbitrary point. Near here, the coastline includes the tidal parts of the rivers.

(30 Aug, 08:49) EdLoach ♦

There is also discussion on talk-gb (https://lists.openstreetmap.org/pipermail/talk-gb/2018-August/021885.html), some of which concerns a Ria (Kingsbridge Estuary) currently marked as a river.

(30 Aug, 10:40) SK53 ♦

As you have found, in some places the river area is mapped with natural=coastline up to the end of tidal influence. In other areas, saltwater estuaries are marked with coastline up to the approximte limit of salt/fresh water mixing. And in some places the coastline is way out at the farest mouth of the river (eg, Rio de la Plata https://www.openstreetmap.org/relation/3474227 has the coastline marked far out towards the open ocean, due to local cultural and linguistic factors, though most people from outside of South America would think of it as part of the sea)

Personally, I would support a more consistent method following this proposal, which suggests the coastline should end no higher than the limit of tides, and no lower than the point where water flow is seaward for significantly more than half the day and where there is a consistent current: (which would mean changing the coastline around Rio de la Plata and other estuaries) https://wiki.openstreetmap.org/wiki/Proposed_Features/Coastline-River_transit_placement

Following the local mapping preference for consistency with nearby rivers is a good idea in the meantime, until we have a clearer standard.

permanent link

answered 01 Sep, 08:04

Joseph%20E's gravatar image

Joseph E
accept rate: 11%


Current at what depth? Salinity at what depth? At what state of the tide? At what stage of the lunar cycle? It is more complex than many people realise. Phenomena like seiches screw with your mind if you think tides are a nice neat sine wave. I know, from work I have done for a major port authority. How are we, as simple OSM mappers, going to find out where the right place is? In general, we are not going to be able to go out in our own boats with our own doppler sensors... we will have to adopt a system from some other authority.

(01 Sep, 12:39) csmale

This answer focuses on determining the boundary between sea and river. That is a difficult question, and I do not believe that it has a definitive objective answer. The solution you suggest is clever, sophisticated, though difficult to implement. However, it isn't the question I was asking.

The river that I first noticed, and 7 out of 10 rivers local to it, have two contradictory boundaries between river and sea. One for the central way, with a separate one (12 miles, 19km distant in one case) for the riverbank. This contradiction seems wrong to me, and I was surprised that my fix, making the boundary between river and sea consistent for the central way and the riverbank, was disputed.

It seems that some people like to use the natural=coastline tag for tidal riverbanks. It seems pretty clear to me that this is not correct.

(01 Sep, 20:22) keithonearth

The bank of a tidal river can be part of the coastline as well. It is not either/or.

The central way to indicate a (named) river is also not mutually exclusive to the coastline.

An interesting example is the mouth of the River Parrett; the river actually runs parallel to the shoreline for quite a distance. If you travel at right angles to the boulevard (at low tide) you will have to cross the river before reaching more flats and eventually the sea. I would suggest that you cannot deny that it is a river, and at the same time, below the high water mark and therefore beyond the coastline.


(03 Sep, 00:34) csmale

keithonearth, Based on the wording of your original question, I thought the debatable issue was if the edges of the river should be tagged as coastline or not. I believe it is widely agreed that the waterway=river should extend all the way down to the river mouth, where the tidal river or estuary opens out to the ocean. That's the way I have seen all rivers tagged, including ones with very large estuaries here in Indonesia and the Saint Lawrence in Canada. The idea is that the waterway can be used for routing by boaters, and also helps determine the main course of waterflow, and the total length of the river. See:

https://wiki.openstreetmap.org/wiki/Talk:Proposed_Features/Coastline-River_transit_placement "Yes, the river line can and in practical mapping frequently is extended outside the coastline. This makes a lot of sense especially if there is a dug deepwater channel or a marked shipping lane. See for example [1]. There should always be a shared node but as said they can extend beyond."

But this whole discussion has reminded me that these issues have never been cleared up. I think we should discuss this on the tagging mailing list or the forums to get some other opinions.

(03 Sep, 13:33) Joseph E

We discussed this on the Tagging mailing list: https://lists.openstreetmap.org/pipermail/tagging/2018-September/038912.html

One comment from Christoph Hormann: "The coastline closure there is both below the lower limit of the proposal and below the the range I can imagine a meaningful coastline closure rule to allow." ... "That is largely not really an estuary but more of a ria. I have no data for this at hand but you can likely see an abrupt change in the elevation profile near Totnes where the submerged section of the former river valley starts. So in this case it would make a lot of sense to place the coastline closure near the upper end of the tidal section because this is much better defined in terms of physical geography."

(06 Sep, 05:47) Joseph E

Keithonearth, in my opinion, you did a correct edit on the “River Dart”. In Cartography traditionally, a river ends with its moth entering into an other (larger) water object and here should be the coastline separation line. This is the case for the vast majority of the rivers in OSM. Actuallt, the exact position of the river-sea separation line is not an issue for most of the mapmakers but it is a serious issue when the two area objects overlap. These overlaps (there are many of them) may cause confusing errors in the related maps and GIS systems. Here are some examples:
-Missing holes/islands, like here https://osm.org/go/cLLcgvwI?layers=D&way=111147606, overwritten by a river section.
-Missing river section like here https://osm.org/go/0G1VkYEQ-?way=219185874. Only the coastline defined isolated water body/sea exists between two river sections.
-The coastline hole/island is not the same as the empty space between the riverbank/river sections in the sea–river overlaps like here https://osm.org/go/0HieoA5B. So the island is neither the mentioned empty space nor the coastline island but the conjunction, the common part, of the two.
-Land/sand is on the both sides of the coastline (land over the sea water object) like here https://osm.org/go/evYuBFR-- . Besides, wet(LAND) over water is also confusing and may cause diverging interpretations like here https://osm.org/go/evYuBFR--?layers=H.
-A mapper uploaded the "Bunnefjorden" fiord area by reusing the coastline data like here https://osm.org/go/0TuJDUV-?relation=6812187. While the lucky mapmakers that ignore the fiord objects have correct maps (like in the former link) others are making unacceptable errors like here https://goo.gl/kbt547 or here https://goo.gl/B9okpH. Just to mention some.
So, as J.T. suggested, do not use the coastline for rivers and lakes. Even more strictly, do not make water area object overlaps. Especially if you do not know the consequences. Besides that these overlaps are adding huge data redundancy to OSM, you will probably cause some interpretation missmach.

permanent link

answered 07 Sep, 15:30

sanser's gravatar image

accept rate: 5%

Sanser, isn't OSM supposed to be about the data, so the consumers/renderers can make their own decisions? This discussion is (or should be) about how to store information in the data such that the distinction can be made. If someone wants to use the "tidal-limit" definition, how do we tag the data such that they can do that? What about if someone wants to use the a different definition of "the coastline", why not? What they want to get out, determines/shapes/informs what we have to put into the data.

In all the cases you cite, can you think of what would be a better way of tagging, such that the appropriate distinctions can be made by a hypothetical cartographer?

(07 Sep, 17:32) csmale

Yes @csmale, OSM is about the data, and we should be storing the data in a way so that the consumers/renderers can make their own decisions about how to render the data. However your suggestion that we should fail to differentiate between a tidal river and a sea does not benefit the accuracy of how the data is stored.

If someone wants to use the "tidal-limit" definition, how do we tag the data such that they can do that?

There seems to be two approaches to this, 1) to tag the sea and tidal rivers identically, and 2) to tag the sea with the natural=coastline tag and tidal rivers with waterway=riverbank and tidal=yes.

What about if someone wants to use the a different definition of "the coastline", why not?

Well, if we take the first approach, they are stuck using your definition of "coastline", and stuck w/o being able to differentiate between sea and tidal rivers. If we take the second approach they have the option to treat tidal rivers and the sea identically, or differently. The first approach gives data consumers less freedom.

In the end we want the same things, we are just disagreeing on the most effective way to get there.

(07 Sep, 18:25) keithonearth

My point is that I don't want to impose "my definition" on the world. As a data consumer. I want to be able to rely on the data to mean what it is defined to mean. I might have a preference, but I really don't care about "winning". If we fail to agree on a robust tagging system, we are all losers. Modelling means compromising. You cannot encapsulate every last detail in a model, because then it would have ceased to be a model and would simply be a copy of reality. It's about making choices about what details to leave out, and what to include. And that depends on what you want to do with it. My model of what we need to be able to get out, includes: the difference between tidal and non-tidal waters; low water and high water marks so as to reveal the foreshore; limits of administrative jurisdictions (land-based: LWM, seaward extensions etc plus maritime-based: baseline/12nm/EEZ/median line)...

(07 Sep, 18:41) csmale

sanser, it's interesting that you mentioned fjords in Norway. A fjord is a U-shaped valley, formed by glacier erosion, that is now flooded. The lower River Dart is a ria, which is a v-shaped valley, formed by stream erosion, that is now flooded.

Shouldn't a ria and a fjord both be considered part of the marine environment, beyond the coastline?

(08 Sep, 04:00) Joseph E

@csmale: I agree, it is necessary to leave out many details. I agree that it's important to think about what to leave out and what to include. I agree that it's important to differentiate between tidal and non-tidal waters.

Additionally, I think it's good to differentiate between tidal rivers and the sea. By failing to make this differentiation, we are imposing one definition on data consumers: tidal rivers and the sea are identical. My suggestion allows data consumers the choice to treat them the same (riverbanks with the tidal=yes tag can be treated identically to the seashore trace), or differently by treating the differently tags differently.

I do not think that including this data is excessive, nor do I see anything written here to convince that it is excessive. After all I'm talking about tagging a section of water 12 miles long, potentially much longer, not talking about blades of grass or pebbles on a beach.

I think that your argument that "You cannot encapsulate every last detail in a model" is weak in this case.

(08 Sep, 08:07) keithonearth

We discussed this edit (https://www.openstreetmap.org/changeset/61959067) on the Tagging mailing list: https://lists.openstreetmap.org/pipermail/tagging/2018-September/038912.html

One comment from Christoph Hormann: "The coastline closure there is both below the lower limit of the proposal and below the the range I can imagine a meaningful coastline closure rule to allow." ... "That is largely not really an estuary but more of a ria. I have no data for this at hand but you can likely see an abrupt change in the elevation profile near Totnes where the submerged section of the former river valley starts. So in this case it would make a lot of sense to place the coastline closure near the upper end of the tidal section because this is much better defined in terms of physical geography."

Also, on the GP tagging list this issue was discussed recently: https://lists.openstreetmap.org/pipermail/talk-gb/2018-August/021885.html (and following messages):

"The wiki suggests the coastline should be the high water line going up to the tidal limit (often a lock or a wier) but this can be a substantial distance inland. This is AIUI the general scientific approach." - Colin Smale

"The River Dart [Way: 194211894 waterway= riverbank] is a really arbitrary line across the river from Dartmouth Castle, this offends my view of what a coastline is." -TonyS999

So it sounds like the majority believe the coastline should include rias (indented areas of coastline due to a flooded river valley), estuaries and tidal rivers. This is the general practice in Great Britain, in particular.

permanent link

answered 06 Sep, 05:47

Joseph%20E's gravatar image

Joseph E
accept rate: 11%

My question was "Should riverbanks be tagged as natural=coastline?", could you answer that question more directly?

(07 Sep, 02:05) keithonearth

"Should riverbanks be tagged as natural=coastline?"

Most inland "riverbanks" (by which we mean water area polygons) should not be tagged coastline, rather natural=water and water=river is best, though waterway=riverband is also used.

However, estuaries, rias, tidal channels in mangroves, and tidal sections of rivers near their months should be tagged as natural=coastline, because these waterbodies are part of the marine environment.

These water features can also be tagged with natural=water and perhaps a tag such as tidal=yes, though you may need to make a multipolygon relation to make sure that the coastline is drawn properly, since the coastline consistes of ways which much meet end-to-end

(07 Sep, 04:09) Joseph E
Your answer
toggle preview

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](http://url.com/ "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: 29 Aug, 19:05

question was seen: 317 times

last updated: 08 Sep, 08:07

powered by OSQA