From OpenStreetMap Wiki
Jump to navigation Jump to search

Why not junction=mini_roundabout?

Would not the key 'junction' be more appropriate than 'highway' for mini-roundabouts since this tag is only applied at the intersection of two or more ways? Also it would fit in better with junction=roundabout. --Milliams 18:36, 21 August 2007 (BST)

+1 Gorm 02:06, 20 December 2009 (UTC)

Roundabout Names

Is there any facility to add a name to a roundabout -- the main roundabouts here in Basingstoke, UK are named.

Certainly. Add the 'name' tag to the roundabout with the name in it. Morwen 20:45, 27 October 2007 (BST)

Is there any standard / common practice on naming roundabouts in the middle of a longer road? See here:
==Knossou Avenue====(o)====Knossou Avenue====(o)====Knossou.. etc.
I believe they should have the same name as the bigger road. Zeptomoon 18:07, 1 September 2010 (BST)

As I understood it, the roundabout is a junction and therefore, does not belong to any of the intersecting roads. Consequently it should not carry the name of any road (bigger or smaller). --ALE! 19:49, 1 September 2010 (BST)
The intersection itself might be named, independently and differently from the roads crossing it. For example Prinsenkrysset in Trondheim, Norway. (Intersection of Prinsens Gate and Kongens Gate. If the roundabout is named in such a manner, than name=* is on its place. --Skippern 21:48, 1 September 2010 (BST)
I get the idea! Makes sense to me. I guess, I was tempted to tag for the renderer again...d'oh! Zeptomoon 12:05, 2 September 2010 (BST)
Strong support for this conclusion!--JeffH 21:19, 17 December 2010 (UTC)

Junction Names

A junction does not necessarily have a roundabout and can still have a specific name independent of the road names. For the moemnt I've just put a 'name' tag on the junction. --Le top 09:42, 29 January 2011 (UTC)

The number of points

The example shows some roundabouts and uses "just" 4 points. In the map a typical roundabout has 8 or more points be drawn. It might make sense to add a good hint on how many points are sensible to use (i have seen rounabouts with 30+ points) Flohoff 12:55, 18 May 2008 (UTC)

In my opinion roundabouts should have as many nodes that it takes to give it a round shape. Compare with how you draw a twisty road section. --Erik Lundin 22:53, 17 September 2008 (UTC)
It would be nice if JOSM could create a circle with 12, 16, 24, 32 or 48 evenly spaced points. Alas, it creates paltry 8 points. I have painstakingly (by following, ungluing and rotating) created a 16-pointer which I keep copying and pasting around. T99 (talk) 19:17, 29 December 2013 (UTC) -- BTW, JOSM did add this capability and it was useful while it lasted, but then it disappeared. Now the default is 18 points and there is no documented way to change that. T99 (talk) 08:35, 7 November 2018 (UTC)
iD does circles - and squarifies things.
You can add any number of nodes you need to create correct connections with connected streets, notably when there are traffic islands separating the directions, and get the correct shap. On large roundabouts, this helps navigation by more explicitly saying to drivers when to place themselves on the external lane to exit the roundabout. This may be useful also when there are special service lanes (for cycles, buses, tramways, sometimes subways and railways) crossing the central island, which may have the priority to the other vehicules turning around. Just place additional nodes anywhere, then select the ring and press "O" (in JOSM or iD) and all nodes will be aligned on a circle and equally spaced (except onodes for interconnection which will keep their anggular position on the circle.
So you can just start drawing a roundabout using 4 nodes (or 8 nodes if there are trafic islands for separating oneway access/exit ways: place first these connecting nodes. Press O, and a few nodes may be added (this avoids non realistic sharp angles with more than 45 degrees). Add enouch nodes so that segments will not pass through the central island and will keep the direction within the outer lane. Press O again if you have added a few more nodes. Generally with about 16 nodes you get a much better look in all scales (with angles flattened with less than 15 degrees) and this works for most roundabounts. Very large roundabounts may need mode nodes to fit the outer lane correctly. It may also be useful to draw the central island and tag its landuse=* (such as grass) or surface, or other elements on this island (trees, light poles, statue, fountain, sometimes footways...). 8 nodes is generally suitable only for miniroundabouts. 4 nodes is generally insufficient and not realistic as they produce too sharp angles (more than 90 degrees on at least one node).
Note that a some roundabouts are clearly oval, or are half-circular only on two sides, the two other sides being mostly straight; a few ones may be found that are lightly "8-shaped" (often created by interconnecting two successive roundabouts to create a single one, without needing more land space). These unusual shapes are still useful to represent correctly as they provide hints about the direction to take and where/when to exit the ring. In those cases, the fast "O" keyboard shortcut will not work properly.
For these reasons there's no standard number of nodes, the precision needed largely depends on the ratio between the outer and inner radius and the number of generic lanes on the ring (most roundabouts have two lanes, roundabouts with only one lane are normally only mini-roundabouts, whose central island is still traversable by long vehicles such as buses and trucks).
So just place nodes correctly to fit the outer lane of the ring, and correctly place other features you may want to to locate around the roundabout (signals, crossings, postal boxes, recycling containers, bus stops, trafic calming, sometimes even trafic light within the ring itself...) — Verdy_p (talk) 06:56, 29 July 2016 (UTC)

highway=*-tag of a roundabout

Old text said: "If the larger road ends at this roundabout (so all other connecting roads are smaller), the roundabout should probably get the same highway=* tag as the smaller roads."

No, that was not a very good idea.

If a "larger road" ends at a roundabout then there must still be more traffic on any segment of the roundabout (traffic from the larger road plus traffic from the smaller roads) than on either of the directions on the larger road. So, the roundabout should be of the same class as the larger road.

Not necessarily. E.g. the roundabout at the end of a motorway is usually not part of the motorway itself - and is more likely to be something like trunk or primary. I've changed it to "usually". Richard B 16:22, 4 September 2008 (UTC)
Motorways would certainly be the exception - whatever's at the end of a motorway should generally not be tagged as motorway. Similarly, slip roads should be excluded when thinking of the "larger road" - a roundabout where "trunk" meets "primary" should be "trunk", but if it's grade-separated with slip roads the roundabout should be "primary". Such is my opinion, at least. Chriscf 16:30, 4 September 2008 (UTC)

I think the currently documented rule is overtly simplistic. It does not cover all possibilities. For example, what if there are three tertiary highways terminating at the roundabout and one residential highway going through? Obviously, trying to cover all possibilities could make the rule very complex. Therefore, it has to be taken only as a guideline. The mapper will have to use his/her best judgment to evaluate the significance of not only each of the highways but also that of the roundabout. By the way, I think the guideline for roundabouts should be similar to those for links. T99 (talk) 21:46, 29 December 2013 (UTC)

On a relation

I think we need an example for how to tag a relation (e.g tags needed for the ways and tags needed for the relation)--skyper 13:46, 2 November 2009 (UTC)

I see very little use for adding such relation, maybe someone can enlighten us... Surely it's simple to group together consecutive ways tagged with junction=roundabout, before using them for any purposes needing a single entity for each roundabout? Alv 14:08, 2 November 2009 (UTC)

Ideally a roundabout should be a single closed way: you normally don't need to split the closed way even for creating route=* relations. When driving along the route it may happen that you don't take the right turn or can't take it because you've missed it or another vehicle blocked you by driving in the incorrect lane, or where not sure if this was the correct exit; you should not stop in the middle, all you need to do is to make an additional turn, which just takes a few seconds instead of using dangerous moves or stops; making this turn still does not take you out of the route.

But sometimes a roundabout is split in several segments because it has some parts with more or less lanes (e.g. for bicycles) or because some segments are small bridges or tunnels where you need to tag additional protection barriers. When a roundabout is split you may need a relation to tag its inner island and assigning it a landuse, or just to name it and reference it as a single object. A large roundabout in a relation may even be used as an "associatedStreet" holding its own address for houses around it.

In more rare cases, a roundabout will have a special way passing through it for tramways or uses only: these vehicles, when present have the priority to the other vehicles already on the ring (that must give the way): in these cases, there will be warning traffic lights that will turn on when these vehicle are coming. The rest of the time, the special way is closed and this is a normal roundabout. To place the correct trafic lights and conditional giveways, or connect correctly the tram/bus routes, you may need to split the roundabout into sections and group these sections into a common relation.

So there are some valid cases where roundabouts could be relations. And this is already happening. However for most roundabouts (small enough for the central island containing nothing more than a handful trees or a light or some flowers), they should not be split at all and the central island should be traced separately using a concentric way. — Verdy_p (talk) 01:14, 22 March 2017 (UTC)

It seems there is a mixup in your arguments of roundabouts as a member of a relation, and junction as a relation. I have not yet investigated what type the 900 mapped relations are. Sure a roundabout can be a member of a route. Not so sure what purpose a relation would have that combines the highway with the inner green landuse. associatedStreet is used by very few mappers nowadays as we have a functional addr:* scheme. For the traffic functionality, there is no relation needed; routing engines cope well with recognising the closed loop tagged with juntion=roundabout, even when the loop is split for different lane features. My conclusion, on relation = no.--Polarbear w (talk) 10:33, 22 March 2017 (UTC)
You seem to misread, I was only speaking about using junction=roundabout as relations. For cases where a roundabout is split in multiple segments these relations are in fact intended to describe what is within the island, and I think it's just better to draw a separate circle for this inner use. But for creating ROUTE relations, I've never said we need to add roundabout relations as members of these routes; however, adding multiple ways taken as part of the roundabout just complicates route relations: it is not justified for something else than creating bridge or tunnel sections in the roundabout. For all other case, a single closed way can be tagged and we evidently don't need a relation for the roundabout itself, and not even for routes passing through them when we can just include include the single closed way as a single member for the route relation.
There's still another need you forget: roundabouts are more than just highways and are used as "places" with their own name, and their own address: this requires a relation, but NOT for the junction itself, only for a "associatedstreet" relation to associate the closed way (or the list of ways) as members (with role "street") and addresses around it also as members (with role "house"): this type of relation however should not be tagged with junction=roundabout which only concerns the highway part (and not anything that may be present in the central island). There exists also cases where these "associatedstreet" relations will also need to include shortcut ways passing through the roundabout (at level or in tunnels), or on the side of it: these ways do not stritcly belong to the streets they connect, but they may be the effective highways where the address points are located (with possibly junctions with private access alleys); these also concern footways or cycleways which are part of the addresses located there as they are the effective ways to access these addresses (inaccessible directly from the loop of the roundabout junction).
When roundabouts are used as places these shound be delimited as a standard multipolygon area, not delimited by the highways but drawn on the limits between the public area containing highways/cycleways/footways and the central island and everything inside of it) and the bordering private properties or amenities (public buildings, parks, sports centers, theaters...): these polygons could be landuse=residential polygons; they'll have a name and a place=* tag, such as place=square (that polygon does not have to be circular, and most often it won't be, even of the roundabout inside it is circular, but as well it should not be used as an associatedStreet relation for grouping addresses located around it, but still located outside of it) — Verdy_p (talk) 16:34, 23 March 2017 (UTC)

I analysed a couple of examples where junction=roundabout was used on a relation:

  • type=multipolygon used to create an area:highway with the middle green cut out. Multipolygon use is not considered a relation tag.
  • type=multipolygon containing highway segments as members. Tagging mistake (leading to area rendering in Carto). Fixed where found.
  • type=route used to attach a nameless route just over the highways (already tagged junction=roundabout) forming the roundabout. this is completely useless, as routers already recognise the closed loop with one or multiple ways. In some of these route-roundabouts the highways lack the junction=roundabout and are consequently not recognised by popular routers. Some mappers add route=junction
  • type=junction which is only a proposed type (itself used 1100 times); the proposal mentions junction=roundabout as one of its categories. Some of these relations carry just one members, which makes them pointless.

I still conclude there is no current valid reason to keep the OnRelation flag. --Polarbear w (talk) 19:07, 22 March 2017 (UTC)

ref tag

Is it correct to remove ref tags from roundabouts? I personally think that this is ok as a roundabout is a junction and therefore belongs to two or more different roads. But please correct my reasoning if I am wrong. I want to get a second opinion on this. --ALE! 22:24, 3 January 2010 (UTC)

Is the highways parts of relations? In that case the ref is protected by the various highway relations, and not really needed on each road segment. If not, than maybe all the road refs should be included. But what about roads that starts or ends in a roundabout, should their refs be extended to the roundabout, or is this really only valid for roads crossing a roundabout? I tend to tag only the refs of roads passing the roundabout. --Skippern 08:43, 4 January 2010 (UTC)
I think the situation is analogous to the 'Roundabout names' issue discussed earlier: the roundabout does not belong to any of the roads and should therefore not have road refs on it. To be more concrete: if you're crossing a roundabout from one minor road to another, it just causes confusion to be told that you have to drive some metres on say a primary road with such and such number. And IMHO, adding refs from all classified roads would be just like including all the road names, just leading to even higher levels of confusion! However, as with names, if a country gives numbers to roundabouts themselves, then these can of course be tagged. I strongly agree with ALE, who started this topic, but I see a lot of misleading tagging here so it would be really good to get some clear guidance into the wiki itself. --JeffH 21:19, 17 December 2010 (UTC)
I guess it varies from country to country. F.ex in Norway, an intersection (no matter if its a roundabout or not) is usually funded, named, owned and maintained by the same agency/department as the road 'highest' class of two roads crossing. In those cases, using the same ref=* on a roundabout as the "master road" would be quite correct. --Gorm 19:55, 4 June 2012 (BST)
The ref= should be handled analog to the name. It is not uncommon in the world that roundabouts have a given name and ref. As per definition the roundabout is a junction e.g. interconnects multiple streets. So it is per definition not exclusive part of one of the streets so it should not carry the name or ref of one of the connected streets. Flohoff (talk) 18:03, 7 November 2014 (UTC)

Two editors confused by roundabout instructions

Just a suggestion for improving this article. I was mapping my first roundabout today, and I wasn't sure whether the following applied to the roundabout I was working on:

If a road is split into two separated carriageways (one approaching, one departing), draw those explicitly and tag these carriageways (not the roundabout) with oneway=yes. Pay attention to their correct orientation and use separate nodes of the roundabout for the approaching and departing road.

Does this apply to all roundabouts that connect to two-way streets? What does this look like in practice? I decided to just connect the normal residential roads to the roundabout circle that I'd made, and not split anything, but I don't know if I did the right thing. And then later today, another editor asked a related question on IRC: "If a highway approaches a turnabout junction, do we split the way into two oneways? And if we do it, how to add both the oneways in a relation so that continuity (in the relation) is preserved?"

Since both of us were confused, it sounds like this page could be improved with an example or other additional explanation for that section. Also, the other editor and I were both unfamiliar with the term "carriageway"; we had to look it up while reading this page. Thanks! Brittag (talk) 05:52, 1 September 2013 (UTC)

I believe the intent here is that this applies where there is a significant island, or other protected area, between the two carriageway, i.e. they are physically separated. Although some people like to micromap everything, I think that an ordinary residential road is best mapped as a single way. -- Hadw (talk) 14:22, 17 November 2013 (UTC)

Bypass lanes for nearside turns

I've come across a roundabout where at least some of the exits have a left (nearside) lane that has priority over traffic on the roundabout for traffic that is leaving at the next exit. That breaks the normal UK rule that traffic on the roudabout has priority. I'm not convinced that the right approach is to split that lane off before the roundabout, as lane changes to or from it are possible along its whole length, and tend to be the only safe approach for bicycles. -- Hadw (talk) 14:13, 17 November 2013 (UTC)

On further reading, it turns out the real problem here is that this is actually a traffic circle and not a roundabout! -- Hadw (talk) 15:10, 17 November 2013 (UTC)

Tag:junction=traffic_circle should be explicit

The article currently says that traffic circles should not be tagged at all. However, it seems that people expect satellite navigators to tell them to turn off at the nth exit, but satellite navigators aren't clever enough to infer that a feature is a traffic circle and therefore has exits. This, combined with a general failure to read and understand the section on what is not a roundabout, is almost certainly resulting in mis-tagging of traffic circles as roundabouts. -- Hadw (talk) 11:54, 5 January 2014 (UTC)

this has been sorted with the introduction of junction=circular after discussion on the tagging list. junction=traffic_circle is ambiguous and could easily be confused with regular roundabouts. --Polarbear w (talk) 17:06, 24 November 2016 (UTC)

Circular and widened road features vs Tag:junction=roundabout

There overlapping content, is there option to cleanup pages in less duplicate manner? Xxzme (talk) 18:29, 10 November 2014 (UTC)

The "Circular and widened road features" were written as a template that had been embedded in the pages of the different tags involved.--Polarbear w (talk) 17:06, 24 November 2016 (UTC)

On a node?

The definition allows junction=roundabout on a node, which makes no sense to me, it could always be drawn as a way. The node should be reserved for the mini_roundabout case. Currently it is used 4000 times, however that is less than 1% of the 450000 tags on ways.--Polarbear w (talk) 21 March 2017

Isn't it same logic as allowing a building on a node? If you are in a hurry or you focus on something else while mapping, you can just tag it on node and later you or someone else might return to it and map it properly. Chrabros (talk) 00:01, 22 March 2017 (UTC)
The nodes tagges are possibly mini-roundabouts, but most often, people don't want to take the time to draw a real circle and reconnect the ways (without breaking bus routes at the same time) It's just so simple for them to just mark a node: it is probably OK if it's with a FIXME to allow others making the geometric changes, but in my opinion they should just add a FIXME. A junction=roundabout by itself has no meaning except with lots of default assumptions.
  • Sometimes this is because they know there's a roundabout, but it's still not visible on the imagery and instead of creating an approximative form for ways, they just tag a node, knowing that this will be detected for later review. Notably the position of crossings for footways, the rection of ways by traffic islands, the exact measurement of the central island, the number of ways in the ring, or if cycleways are interrupted, or of bus stops are moved: they don't know how to place these details.
  • Sometimes this also happens only for juctions created temporarily during works nearby. This won't be on aerial photos when the works will be finished, and it's not very easy to draw these from photos taken from the ground.
If the junction is new or has been recently changed, tagging the node only is just an interim solution that is best then nothing, as if it was still a standard crossroads.
More and more roundabouts are being built worldwide to replace standard junctions: they replace standard crossings, by removing trafic lights, reducing accidents because they are also efficient speed limiters, they are still making circulation more fluid, reduce the noise and pollution, they allow planting trees, reduce the bad effects of winds on long boulevards, they are decorative, and don't use any energy for traffic lights. They're a bit e'xpensive at construction time, but then municipalitites save money each year. And driver's life is simpler: they only need to look for vehicles from a single direction, and spend less energy by burning fuel waiting in long lanes of vehicles near traffic lights or in trafic jams that pollute everybody around. Only US still does not have a lot or roundabouts, but now they are spreading worldwide (even if European countries have many more, you can find them including in African cities or in China). US has a few ones but only for very large roundabouts, or for roundabouts with sections passing through bridges above uninterrupted motorways. — Verdy_p (talk) 00:43, 22 March 2017 (UTC)

Thanks Verdy_p for the favourable speech for roundabouts, that is correct, but we are discussing the drawing style. Even for a new roundabout, all you need to do is drive the full circle with your GNSS, and you can map it precisely even without aerial imagery. Thus this is no reason for me to draw a node only.--Polarbear w (talk) 10:21, 22 March 2017 (UTC)

I agree that this tag should not be used on a node - either it's a mini_roundabout node when flat or a junction=roundabout way when it has a non-traversable island. Though as has been mentioned, some mappers might use it as a placeholder - would be good if QA tools like keepright or Osmose highlighted it as requiring fixing. GinaroZ (talk) 16:08, 22 March 2017 (UTC)
Osmose already signals incorrect roundabouts (including those that are traced in the incorrect direction, and it also detects circular oneway highways that should be roundabouts; it also signals names or references given to them to match only one of the roads/streets connecting to them: roundabouts normally have no reference or their own reference, and have no name or use their own name distinct from the public roads/streets connecting to them; The rules about these are country dependant and may be tuned per country). — Verdy_p (talk) 18:57, 22 March 2017 (UTC)
Users should be careful then verifying Osmose flags, since not every circular oneway is a roundabout. --Polarbear w (talk) 17:06, 24 November 2016 (UTC)
Yes but in almost all cases, Osmose detections are correct. If this is not the case we should be able to tag these explicitly with another value for junction=*, probably like juction=crossroad, or with an explicit note=* affirming it. These cases are almost inexistant in France, and Osmose detects the case where the circular way has only one connection for motorvehicles: this is efecetly not a roundabout but an "internal loop": a goosd sign is these cases have parking lots around the end, while stopping or parking is normally forbidden around roundabouts.
Note that a roundabout is still a roundabout if there are multiple connecting cycleways, we can only ignore the footways and service ways for alleys to private properties
But beware that real roundabouts also exist within private areas with public access (e.g. in commercial centers between parkings separated by trafic islands or other barriers or footways).
Roundabouts are also not necessarily purely circular, they exist with square shares, or ovals, or were created by joining two touching toundabouts in an 8-shaped loop and closing traffic on their touching borders. — Verdy_p (talk) 19:39, 22 March 2017 (UTC)

Implicit restrictions at roundabouts

I'd like to add to the wiki that a roundabout implies no_u_turn at its entrances, since I've seen such errors. for example, Any objections? -- Zstadler (talk) 11:38, 4 April 2017 (UTC)

I don't like too many implicit assumptions in OSM since they are difficult to follow for data consumers. (I even prefer to explicitly tag a roundabout with oneway=yes). A roundabout allows an easy turn by driving the full circle. I don't know why the mapper in your example tagged the restriction of turning into the same way - is that occurring more frequently? A U-turn always requires extreme care from the driver, but I don't know a rule in the driving legislation that declares it forbidden at the place where a road joins a roundabout. So, unless there is specific signposting, I see the relation in the example unnecessary, but I also do not see a need to describe such implicit assumption.--Polarbear w (talk) 13:07, 4 April 2017 (UTC)
I'm confused. If you claim that there is no implicit restriction on a U-turn on the entry to a roundabout, then such a turn is allowed and adding an explicit restriction is not "unnecessary". Moreover, performing such a u-turn is effectively driving against the traffic direction in one-way road. --Zstadler (talk) 05:08, 5 April 2017 (UTC)
In Germany, for example, there are currently more than 200 restrictions on roundabouts. See . It seems that users in different countries see them as "necessary". -- Zstadler (talk) 06:12, 5 April 2017 (UTC)
@Polarbear w: I think Zstadler is speaking about the restriction that applies from going between the two oneway branches of the Y connected to the same roundabout (i.e. coming from the roundabout circle back to this circle): this does not go against the two oneway restrictions.
However, these restrictions apply when the common long branch of the Y (the road connected to the roundabout via this Y) has a continuous separator (a line or zebra painted on the road, or a traffic island or barrier) between lanes for each direction.
  • When there's an island or barrier, this common branch should however be drawn with two parallel but separate ways, and such restriction is not needed.
  • The restriction is useful when this is just a painted line or zebra in the middle of the road (effectively on these roads, U-turens are disallowed, but not just on the central node of the Y, but all along the dual-direction way. Here again a restriction relation is not needed (just tag the way)
  • If the common dual branch has no island/barrier/continuous line/zebras, U-turns are allowed starting from the central node of the Y and all along that way: we also don't need these restrictions. But some countries may prohibit suchg U-turns if the next road crossing is less than 50 meters away or if the connection node of the Y is less than 50 meters from the roundabout (this case happens within urban areas, where U-turn are probibited almost everywhere, including on dual-direction ways: drivers just have to join the next crossing ou roundabout. The circlular loop of roundabouts offers the safest way to perform a U-turn.
  • U-turn at end of noexit ways however are always permitted where there's a "turning circle" area, or there's some parking (provided that there's at least two lanes, otherwise you'll need driving backward, or entering in a private access way or in front of a garage door to perform the maneuvre).
For all theses reasons, I do not see any benefit of addding any explicit restrictions relation on the central node of an Y (only simple U-turn restriction tags directly on bidirectional ways with just a painted central line).
And I've never seen any roundabout that did not allow to perform a U-turn using their circle: you can even perform multiple rounds on it (notably if for some reason you could not take one of the exits, because normally you must not stop in the middle of the loop where you just have to drive slowly, and to carefully give way to vehicles from the inner loop lane that want to exit and have their lateral warning lamps flashing, but sometimes you have to make another turn on the loop if the exit is busy). — Verdy_p (talk) 12:44, 5 April 2017 (UTC)
Finally about the 200 roundabouts on Germany, their geometry is not correct. Adding these restrictions on nodes directly placed in the circle is not the correct way to do it. It is MUCH simpler to just draw the separate oneway lanes of the Y connecting to the I instead of drawing a birectional way directly to a node of the circle: this is poor/lazy drawing that forgets also to mark the position of giveway nodes on one of the two oneway branches. — Verdy_p (talk) 13:01, 5 April 2017 (UTC)

Roundabout flares / Names on connecting Y-junctions

Just as point of clarification as I can't get a solid answer from reading this page. When you have a road that splits into a Y-junction you keep the classification of that road until it connects to the roundabout, meaning they do not become links. Does this also apply to the name of said road? For example, if a primary road named "Don Quito Avenue" became a Y-junction, the proper tags for those split roads would be highway=primay, name=Don Quito Avenue, oneway=yes correct? - Te-ika 23 July 2018

Not sure what you mean with "Y-junctions". I you mean that the road splits for a few meters around a small traffic island ("splitter island"), that is called a "roundabout flare". For a small one, there is no need to split the road in OSM, just tag a node with traffic_calming=island. If they are larger, they are part of the road, thus they carry the same classification, name and ref as the other part of the road. --Polarbear w (talk) 09:09, 24 July 2018 (UTC)
I would rather remove “ it is sufficient to set a node tagged traffic_calming=island” because this is not a clear tagging scheme. Either this is put to the node that the incoming road shares with the roundabout (than the meaning to which feature this applies is not clear) or it is put on a different node on the incoming way, but near to the roundabout (this would be misleading, because it would mean that the road is going around the traffic island, but the carriageways join again on the other side of the traffic island – which does not reflect the reality here). Sommerluk (talk) 13:31, 24 July 2018 (UTC)
I mean indeed to set the node separately; it must not be set onto the roundabout itself since it does not affect the traffic within. I cannot see the implication that the carriageways need to join again. Otherwise these splitting island have typical properties of a traffic_calming=island, e.g. often the footway routes over them. Of course we could invent traffic_calming=spitting_island, but I see no specific need for it.--Polarbear w (talk) 20:04, 24 July 2018 (UTC)

Roundabout flares / How to map

There is the chapter about roundabout flares offering the node with traffic_calming=island. Unfortunately the "How to map" chapter still insists on splitting the road. This is erroneous if people focus on "How to map". I therefore added some words in the "how to" explaining, that a split can be avoided if it is a small island. I'm aware this might be a little bit redundant, but a "how to" has to be complete. I also tried to give some guidance what a "small island" is, stating that crossing of footway/cycleways does it make "not small". Please let me know if somebody feels this not being appropriate. --Bicycle tourer (talk) 16:42, 26 September 2018 (UTC)

Connecting ways to roadabouts

Re: "Two ways which enter and exit a roundabout should never connect to the same node on the roundabout. This is required to allow routing applications and software to provide correct directions ..."

Well, sometimes the truth is stronger than fiction and the off ramp from one road actually meets the on ramp to another ramp at the roundabout, almost like a squished intersection between 2 one-way roads (except the traffic already in the roundabout has the right-of-way). The junction ought to be mapped as it is, and the routing applications will have to adapt their algorithms accordingly.

T99 (talk) 08:20, 7 November 2018 (UTC)

Not a roundabout addition

To make it more clear, I suggest adding a bit to the last line on Not A Roundabout, the part in italics. This is how it's described on the Throughabout page. Suggestion: "In this situation the tag junction=roundabout has to be avoided strictly. These cases should instead have a junction=circular tag applied to their circular traffic.” --Awiseman (talk)

Suggestion for lanes that bypass a roundabout

Currently the "How to map" section says incoming and outgoing ways should not share a node -- I think this is correct, but there are examples of when there is a lane that bypasses the roundabout, in which case I think it would be ok if the road shares a node. I suggest adding this after the second to last bullet: "There are scenarios where a road entering a roundabout has painted markings or road signs indicating the outermost lane bypasses the roundabout. In these scenarios, an exception can be made to the separate node policy above and the entrance/exit roads should share nodes, as vehicles making this maneuver would not enter the roundabout or yield to roundabout traffic." One example is here, the southern road can bypass the roundabout: For these examples it would be the most accurate representation of ground truth. --Awiseman (talk) 23:05, 10 January 2020 (UTC)

Interesting case, but why not put lanes=2 to this short way segment? --MalgiK (talk) 01:56, 11 January 2020 (UTC)
This is not accurate. It is a non-divided version of slip lanes, which should be represented by the short segment on the roundabout, as it is still part of the circular carriageway. As mentioned, use lanes=2, with at least change:lanes=yes|no; and turn:lanes=left|through on the approach. To be the most accurate, use connecitivity relation. For yield control, perhaps something like give_way:lanes=minor|no on the highway=give_way? Kovposch (talk) 11:50, 13 January 2020 (UTC)
I agree that adding additional lane tagging makes this junction more accurate than before, but it would not eliminate the incorrect navigation/modeling that is created by separating the southern entrance/exit nodes. If the southern entrance/exit nodes are separated, all west to southeast traffic will be forced to enter the roundabout, when in reality that traffic will never enter or yield to the roundabout. Lane tags would be helpful to add, but in order to provide accurate navigation for west to southeast traffic (and all other directions of traffic), the southern entrance/exit nodes would be shared. What do you think? --Awiseman (talk) 19:48, 15 January 2020 (UTC)
It really depends on what is meant by "enter the roundabout". But first, are you sure you don't have any misunderstandings on your example? According to 2019 Mapillary imagery, circularizing vehicles are allowed to cut into the outermost lane, with broken line on the left side. That's why I said change:lanes=yes|no) at first. This would be a strange and weaker case. There's even a yield sign on the right side of entry. It looks confusing,Kovposch (talk) 11:06, 16 January 2020 (UTC)
Concerning definition, currently OSM doesn't map individual lane areas. There's no solution to tell one lane from another. If routers/navigators don't support by-lane features, it's their fault, and they should accept a lower level of accuracy. We are trying to represent the roundabout as it is. Kovposch (talk) 11:16, 16 January 2020 (UTC)
Ah, it looks like the example I cited might have some construction. Here's another one similar to the issue I'm referring to: example and on Mapillary here -- currently there is not a shared node even though all three approaches would allow avoiding the roundabout. Let me know what you think. Thanks. --Awiseman (talk) 21:19, 23 January 2020 (UTC)

roundabouts with trafic lights


  1. The circular traffic has the complete right of way (at every node an at every time) -> Tag:junction=roundabout
  2. The incoming traffic has always the right of way (at every node an at every time) -> Tag:junction=circular
  3. The traffic is ruled by traffic signs, marks and/or lights
    1. The circular traffic has the complete right of way (if traffic light is of) (-> mostly tagged as Tag:junction=roundabout, but may not match with law (German STVO))
    2. The incoming traffic has the complete right of way (if traffic light is of) (-> Tag:junction=circular)
    3. some but not all of the incoming ways have the right of way (-> Tag:junction=circular)
  4. Special cases e.g. with crossing trains (-> tagged like "ignore the train")

contradictory description

  • Introduction of Tag:junction=roundabout: The tag junction=roundabout is used only on road intersections where traffic on the roundabout has right of way. That is, the roundabout itself should be free from all intersection controls including traffic signals, stop signs or stop markings, give-way (or yield) signs or give-way markings. (this is only case 1)
  • Tag:junction=roundabout#Special_cases Signal controlled: Even when all junctions are signal-controlled, it is suggested that junction=roundabout be retained. (case 3.1 -> but does not fit to Introduction and with German law)

What is the reason to mark a circular traffic / roundabout?

  1. To generate driving instructions like "In the roundabout/circular take the second exit"
  2. To mark the driving direction (also possible with Key:oneway)
  3. to mark the right of way (works only for roundabout): (otherwise Key:priority_road)

What are the arguments for tagging a false roundabout (with traffic light or other traffic signs) as roundabout?

  • its easier to tag (aggregating type, direction and right of way in one tag)

What are arguments against

  • may not match with law (traffic rules). there may be small differences in rules for roundabouts and false roundabouts.
  • at junctions with traffic lights nobody has always right of way


  • The right of way is importend. o.k. You may use Key:priority_road for pseudo roundabouts.

--Langläufer (talk) 08:19, 26 February 2020 (UTC)

Truck Aprons?

Roundabouts that are 1 lane typically have a raised, paved area located inward. This is designed so that towed trailers or other long vehicles can navigate the sharp turning radius by having their end trail over the center area. This is not a mini-roundabout since there will still be a landscaped area in the very center. See

There doesn't appear to be any documentation on how to tag this feature. I am wondering if anyone has feedback?

Maybe you can consider it as some sort of shoulder in the meantime? -- Kovposch (talk) 14:48, 28 July 2020 (UTC)


How to deal with cycleways around roundabouts? A section walking newbees like me through this topic would be great! Martianfreeloader (talk) 11:33, 8 November 2020 (UTC)

Where to map if several lanes?

In a roundabout which has more than one lane, where should I map the highway? On the centre of the outermost lane or on the centre of all lanes? Martianfreeloader (talk) 10:56, 5 December 2020 (UTC)

Turbo Roundabouts

Starting a section here. I've preliminarily converted some roundabouts to turbo, using lane connectivity Relation:connectivity. Note that it requires quite a lot of the relations in one roundabout and some splits. Also so far I did not find any router usable to verify the lane routing after the junction is all set.

Here my 2 examples:

Details with some explications will follow.

Splitting roudabouts

To me it's common practice to split roundabouts that need splitting to accomodate the mapping of routes or where the number of lanes differ in the roundabout to be able to tag that. In December 2020, someone made an (otherwise good faith) edit that added "** Do not split the roundabout itself.". A recent discussion led me to change the wording of this to "** Avoid splitting the roundabout itself unless necessary for tagging different properties or route relations.". This was reverted by a user using a very threatening tone. I therefore seek to affirm the notion that splitting roundabouts is allowed. Please sound off your views on this. --Riiga (talk) 15:54, 18 February 2022 (UTC)

Yes, this is fine. Cases for splitting roundabouts include route relations, public transit relations, turn restrictions, different numbers of lanes, and probably all sorts of other roadway attributes that may differ through a roundabout. --ZeLonewolf (talk) 15:56, 18 February 2022 (UTC)
Do not mix up splitting them when required compared to splitting them for fun. The problem was especially raised when it is was not required, exacerbated by the fact that route relations are also completely messed up. (There may be a whole section about splitting, and what to watch for, but without that it is extremely harmful in general, especially when using ID editor, especially when ID editor splits up without notifying the editor.) The page currently says when to split and what and how; the specific paragraph is talking about connecting ways, which must be be used as an excuse to split. That's the problem. --grin 16:00, 18 February 2022 (UTC)
Why would you split a roundabout for fun? If iD does, that's a problem with iD, not the wiki or common practice in OSM. I exclusively use JOSM and only split roundabouts when needed, just what my edit to the wiki page was trying to suggest. --Riiga (talk) 16:05, 18 February 2022 (UTC)
I already explain the problem here and here. This is about a bug report for iD editor which splits the roundabout automatically breaks the relations. --Urbalazs (talk) 17:12, 18 February 2022 (UTC)
I still don't understand how they are broken by this. If they are mono-directional, the opposite side should not be included in the first place. If they are bidirectional, eg JOSM is capable of showing forward and backward, and could easily be extended to showing they form a loop (as in rings for areas) regardless. --- Kovposch (talk) 17:54, 18 February 2022 (UTC)
iD explodes roundabouts if you set turn restriction on the entrance/exit branches. ITineris (talk) 16:16, 18 February 2022 (UTC)
@Grin: If you understand that splitting may be required in some cases, why you made edits claiming that it should be never ever done, even when required due to only part of roundabout belongs to a relation or where some property (turn lanes, surface, lane count, sidewalk etc etc) changes? Mateusz Konieczny (talk) 11:09, 19 February 2022 (UTC)
Roundabout entrances/exits are often mapped incorrectly: drawn with two separate (oneway) ways instead of a single way with a `traffic_calming=island`. Would it be drawn correctly, one shouldn't need to exclude some parts of the roundabout from the route relation. ITineris (talk) 16:16, 18 February 2022 (UTC)
The basic rule of dividing a highway in OSM is "physical separation". A traffic island is clearly "physical separation", so if anything, what you suggest is incorrect. Even if done your way, there are good reasons to split a roundabout: bus routes that only go in one direction, different number of lanes in a roundabout, etc. --Riiga (talk) 16:39, 18 February 2022 (UTC)
Have you read the section about Roundabout Flares on the page? ITineris (talk) 17:49, 18 February 2022 (UTC)
Have you read the clause "For small one" and "no need"? Not "all" and "must not". --- Kovposch (talk) 17:51, 18 February 2022 (UTC)
I agree with strongly. Non-splitted ones are more wrong / less correct from the onset. --- Kovposch (talk) 17:44, 18 February 2022 (UTC)

It sounds like you seem to have difficulty in coming up with consensus. If you would like to change how roundabouts are mapped, it would be beneficial to gather more public mapper input by starting a new thread about this on the tagging mailing list. -Bkil (talk) 19:09, 18 February 2022 (UTC)

As I mentioned in the iD issue, this edit in 2020 has been taken out of context. The page had referred to the longstanding debate about whether to "split" a flared approach into a dual carriageway, per the physical separation principle, or keep it as a single carriageway for simplicity. The edit was merely clarifying that this guidance about "splitting" did not refer to splitting the roundabout way itself just because of the flare. In fact, other parts of the page have always referred to "roundabout way(s)", acknowledging the potential for multiple ways. Unfortunately, the diff makes it difficult to discern the nature of that minor edit. Relation:route discusses this issue in more detail, describing the tradeoffs between two possible mapping styles. More broadly, the PTv2 scheme implies that route relations follow a linear, well-ordered progression to the extent possible. – Minh Nguyễn 💬 02:18, 19 February 2022 (UTC)