Key:traffic:hourly

Not directly related to car routing, but information that could be useful for other uses or plain make a interesting visualization. Traffic changes every day, but we can get reasonable estimates by counting the traffic for a know time and by categorizing that measurement.

Your country officials might publish automatic measurements of some trunk or primary routes, but it is very unlikely they even have any measured data on residential roads, cycleways or footpaths.

Measuring

Survey length

We don't need to count the traffic for a full hour to get roughly correct results, keeping in mind that the traffic can vary 10-20% each day.

• One should aim to measure the traffic for 12 minutes or more.
• For high traffic roads anything over 5 minutes might be reasonable, if it is a multiple of the cycle time of any nearby traffic signals. After about 5 traffic signal cycles the estimated numbers gain only little more precision.
• For quiet side roads (less than 100 cars/hour) it can make a big difference whether you survey for 8 or 12 minutes; anything measured in less than 8 minutes is only a rough scale; but usable to distinguish it from roads with considerably higher traffic
• Very low traffic walking/cycling routes need surveys of over 15 minutes for usable figures (other than "the traffic is near non-existent".

In any other cases than the last one above, measuring for more than 30 minutes adds very little value, unless you can do the measurements automatically.

Way or node?

In most cases it is better if you can measure the traffic crossing a single node which is part of only one way, or of two consecutive ways. In some cases, for example a footway crossing a two carriageway road in an area that has sidewalks drawn as separate ways, it can make sense to split the footway to be just the part crossing the road and add the tags on that way.

Tagging hourly traffic on longer ways, where the traffic can turn away to side roads between the end points of the way, should be discouraged as the traffic density is not uniform on the whole length of the way.

Traffic tagged on a node with several roads intersecting can't convey the information on the traffic direction, but is better than nothing. Once the map is locally complete enough for users to even consider tagging traffic densities, there's likely suitable nodes just before most of the intersections (pedestrian crossings, traffic signals, extra lanes starting) - such nodes are a ideal place to tag the traffic density.

Values

Values come in three types:

• =105/12:00: someone counted the traffic for 12 minutes; their observed count was 21, thus they have a reason to believe the traffic is 105 vehicles in an hour.

  This makes it possible to measure the traffic on several days; just add up the original counts and the interval lengths to get a more precise number.

• =2410: someone counted the traffic for a unknown interval. Extrapolating from that they got the hourly traffic count to be 2410 vehicles in an hour.

  If one does have more data to refine any such value, a good estimate (given these guidelines) of the original interval can be six minutes; The new hourly traffic would then be (241 + measurement)/(6:00+measurement time)*60.

• =<5: someone observed a road for roughly 11 to 12 minutes but witnessed no traffic; we can only say that the expected hourly traffic is less than 5 vehicles in an hour.

  For very quiet roads the interval between vehicles can be assumed to uniformly distributed.

Values are usually rounded down to nearest integer.

Key family - namespace

Editing manually long lists of values in one tag value can be slow, so I'm using a "family of keys" - namespace. In addition to starting with traffic:hourly, each key should have, separated with colons, the following qualifiers:

• the hour the measurement was made in 24 hour format.

  If the measurent period was less than 20 minutes and crossed the hour mark, use the hour where it ended. That is, a measurement from 16:50 to 17:05 should be entered at traffic:hourly:17:...=*

• the day of the week (Mo-Su) or (midweek holiday, suggestions welcome). Traffic is significantly different on working days vs. other days.

Very welcome are also:

• seasonal indicators: winter/spring/summer/fall

  often not by calendar but by local conventions

• weather indicators: wet/snow/sunny/hot-as-hell

  a short shower does not make the weather wet, but it's often still wet for half a day after a long rain.

  a sunny day in the fall can be unusual enough

You can also use:

• transport mode names: hgv/bus/bicycle/foot etc.

  If you're interested in the proportion of hgv's or buses

  If the sidewalks are not drawn as separate ways, you s

Examples

• traffic:hourly:08:Mo=10, a footway in a residential area on a Monday morning.
• traffic:hourly:16:Fr:summer:backward=180, a popular walking route full of pedestrian commuters going home in the direction opposite the osm way direction.
• traffic:hourly:07:Tu:forward=2300/12:00, a primary road used in the morning rush hour and counting in the direction of the way.
• traffic:hourly:01:We:fall=150/15:00, a oneway section of the same road in the fall and at night.
• traffic:hourly:01:We:fall:wet=150/15:00, as above, but in wet conditions (is raining or has rained considerably in the past day).

What counts as traffic

Anything using that portion of the road.

• A car towing a trailer is one vehicle.
• A person pushing a stroller, or walking a dog, counts as one.
• A cyclist on a sidewalk counts as traffic on the sidewalk, even if he's not allowed to drive there.
• Remember to count yourself in the figure, too, if you cross the point.
• Even if the sidewalks are not drawn as separate ways, the traffic count for the road should only include those driving on the vehicle side of the curb. If you want to also count the sidewalk traffic but aren't going to draw the sidewalks, add, say, :sidewalk in the key.

Surveying tips

Once you've mapped your neighborhood thoroughly and need to walk the dog, you can often find locations where you can walk around a park but keep one or several spots on the nearby roads visible at all times.

Start your stopwatch or wait till the clock digits on your mobile phone change, and keep counting up to any number of full minutes. It's easier to count two directions if you always make a mental note of the count for both directions. (one zero, two zero, two one, three one, ...)

On longer straights it's often possible to choose a location far enough to keep walking to that point for up to about 6 minutes.

If the traffic's quiet enough and you have a suitable mobile phone or similar, you can scribble the values in a new sms or other note. Or make a sketch of the roads on a paper and use tally marks.

Using video

If there's more than two directions to monitor, most people easily loose track of the count so far. With one video recording it's quite possible to get traffic counts for 10 or even more distinct locations.

If specifically surveying the hourly traffic, you can stop and record the traffic on video for later analysis one direction at a time. 10 minutes of video can take an hour to count, but often you can do other things while counting the traffic for one direction.

It's much easier to watch the video later, if the camera is kept still; a bicycle handle bar makes for a good support.

Location

• An ideal place to record the video would be pedestrian bridges near intersections.
• If possible, choose your location so that the camera sees several cars waiting for a green light, in every incoming/outgoing direction that you're interested in.
• Standing too close keeps the incoming or outgoing directions outside the picture, and faster cars driving very near you might even be only visible in one or two frames.
• Note that buses or other big vehicles can hide vehicles in the next lane(s) if you're shooting perpendicular to the direction of the traffic or stand too close to the road.