[Tagging] Sample tagging for highways with no lane markings

Tod Fitch tod at fitchdesign.com
Thu May 24 03:47:26 UTC 2018

> On May 23, 2018, at 7:57 PM, Paul Johnson <baloo at ursamundi.org> wrote:
> On Wed, May 23, 2018 at 10:34 AM, Tod Fitch <tod at fitchdesign.com <mailto:tod at fitchdesign.com>> wrote:
>> On May 22, 2018, at 12:48 PM, Paul Johnson <baloo at ursamundi.org <mailto:baloo at ursamundi.org>> wrote:
>> In the case of your typical bog standard American residential street, I'm strongly disinclined to agree that this is a two lane situation.  I'd be inclined to mark unpainted lanes in the cases where channelization regularly occurs without the pavement markings anyway.  This isn't the case on residential streets, as people will tend to drive right up the middle of such streets, only movingly right to meet oncoming traffic and maybe when approaching a stop sign.
> Hmmm. I guess driving culture may vary from place to place in the US. I always keep to the right regardless of the existence of a lane markings. I will admit, however, that traffic studies indicate that the average driver will be a bit more to the center of the pavement if there are no lane markings. Similarly, at least in residential areas, it has been found that drivers will generally go slower if there is no center marking. At least that is the rational my local government is using to remove the center divider marking for traffic calming purposes.
> While this may be true, most people will shy towards center (and perhaps even stay in center) for most of their trip down a standard width street (which, while typically 40 feet, this is inclusive of all features including sidewalks, making the effective width of the roadway closer to 25 feet, a random pull from Mesa, Arizona's design guide <http://www.mesaaz.gov/home/showdocument?id=1044> blindly from Google confirms this, with their design guide being 27 feet across between curbs), means that two full size pickups can (barely) pass two cars parked on opposite sides of the street at once.  That's also generously wide compared to a lot of places, many suburban and small town residential streets I've encountered are open-edged with parking off the paved area, and the paved area being maybe 20 feet on a particularly wide street.  New urbanist street designs are similarly, deliberately, narrow as a traffic calming measure, as parked vehicles will tend to provide de facto ad hoc chicanes.  As such, if lanes are marked at all, it's usually at the very ends of blocks only, where parking is prohibited, as a confirmation that the street is indeed two-way and provide a hint as to the default passing rule.

I have noticed that newer developments, especially infill development, have narrower residential roads than where I live. And I admit I did not look up current design standards. I simply took a tape measure to a number of residential streets in my neighborhood. The one in front of my house is 40’0" +/- 1" between the curbs. There are sidewalks but I excluded them from my 40’ number. Subjectively my current street seems about the same as others in the area and the same as my in previous neighborhood in a different city. Both neighborhoods are older, laid out when accommodating the automobile was high on the list of design criteria. It would be interesting to pull out the design standards that were in effect in the 1950s, 60s and 70s when much of our current suburbia was created. I would not be surprised if a lot of our current stock of residential roads are wider than the current standards specify.

By the way, I don’t see a way to tag the accuracy or confidence level for a measurement. Seems like we ought to have something like *:confidence=*, similar to the *:lanes tagging so we could, for example tag the width of a road as:


If you are only estimating from the most likely source (allowable imagery) then you probably are not going to be much closer than 0.5 meters or a couple of feet.

A confidence/accuracy tag would probably be another can of worms. How are you determining it? Statistically? One sigma? Two sigma? Or assume a single measurement but with a technique known to some typical error pattern?

But I digress.

> I know that road design varies over the world and even, to a certain extent, within different states in the United States. So this discussion is showing different regional points of view. A typical, or to borrow the UK slang  “bog standard”, American suburban residential street is wide enough for parallel parking on each side and space for trucks/lorries to get past one another [1]. Typical parking lanes are about 8 feet (2.4 meters) and a typical traffic lane is 12 feet (3.7 meters). So a total pavement width is typically around 40 feet (12.2 meters). In some parts of the world, even in older crowded US cities, a road of that width might be striped for four lanes of traffic. But a typical US residential street has no lane markings.
> US tends to favor 9 feet per lane and 6 or 7 foot parking strips for a full size residential street (and combine with 6 feet being the minimum, 7 becoming common, and even wider in some places for the bike lane, this will feel quite clausterphobic and many, if not most, drivers who will yield the entire space to a vehicle passing a parked vehicle first to stay out of the door zones).  Per federal guidelines, a boulevard would be at least 10, preferably 11 foot lanes (and this will still feel quite narrow to most American drivers).
> I can see the logic of only using the lanes tag if there is paint on the pavement. But that leads to another issue: It is pretty easy from experience to glance at a photo of a road and say it is wide enough for two lanes of traffic. But it is much harder for me to determine a width accurate to a couple of feet. I don’t see a way to show a measurement error estimate [2] and listing something as width=40'0" implies much more accuracy than a guess based on a quick visual survey or imagery actually provides.
> Look for the wear marks, these will be quite prominent in sun-prone areas and where concrete is used.  Generally speaking if there's defined lanes that are just worn off, there will be wear marks where passing motorists have rolled the same spot repeatedly.  This can often be confirmed with your favorite license-compatible street-level imagery or a survey.  Though if you're using JOSM and have suitably high resolution aerials available, you can use JOSM to draw a line perpendicular to the way from curbface to curbface to find the width.

My current favorite license compatible street level imagery is from my dash cam. :)

I am considering getting another dash cam and rigging up something so that it faces out a passenger side window. I figure that would be a reasonable way to capture a bunch of shop details that I can’t get with a forward facing camera.

Fortunately, or unfortunately depending on how you look at it, the lane markings in my area are kept in fairly good repair so seeing where they are worn off is often not possible.

> I am rambling. To the point, if I were to add my photo [1] to the urban highway tagging examples page of the wiki [3] what tags should it have. My current guess is:
> highway=residential
> parking:lane:both=parallel
> sidewalk=right
> surface=asphalt
> width=40'
> For the specific example given by the photo, what tags would you suggest.
> Probably closer to 34'0" wide (we're still in agreement on customary units that the inches should be included even when not necessary when tagging?), since those vehicles are narrower than a full size pickup (typ. 7 feet) and up against the curb, and so I'm reasonably sure there's not more than 22 feet between them, but more than 18 feet.  I'd still leave off the lanes.

Only issue with fully specifying feet and inches is that it implies inch level accuracy with I doubt any of our road mapping achieves. See above for a digression on tagging accuracy/confidence levels.

I guess the question is: Would you leave off the lanes=* regardless of the width as long as there is no painted center line? If it is width dependent, at what width would you add a lanes=* tag even if there was no center line painted.

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openstreetmap.org/pipermail/tagging/attachments/20180523/1fcdf79f/attachment-0001.html>

More information about the Tagging mailing list