User:Alv/Tagging issues

From OpenStreetMap Wiki
Jump to: navigation, search

On this page I have listed things that are already being mapped, but where users have, or might have, different ideas of what the tags really mean, or what they don't mean/include/imply.

Comments welcome on User_talk:Alv/Tagging_issues.

Key: start_date



There are loud voices that "osm should map what's on the ground", i.e. not what was there before. I'm not saying we should draw roads that were built and removed in the past, but to record the attributes of the current roads, or other features.


For some objects the meaning of start_date=* is obvious. For example most buildings are built once, and later renovations don't make the building a new building. A section of a stream that was diverted, was built when the ditch for the new course was dug.

The main concern is highways; we have old (as in old enough to be public domain) maps that show how some roads existed already in the 19th century, or even earlier. Their course and physical state may have changed significantly, or they might be just the same as they were back then, if a newer road was built nearby. Most long distance roads have been widened and straightened, even made into a dual carriage road.

If we come across a way tagged highway=cycleway start_date=1771, we can deduce that it wasn't a cycleway back then - or a pedestrian road or any of the current classifications. On the other hand, if there's two residential street sections, both built in the 1960's, and a section with highway=pedestrian inbetween, and all are tagged with start_date=1961, we can't tell if it was originally pedestrian or a normal road - or if the middle section has start_date=1997, did the road exist in 1960?

Let's first consider single carriage roads only. What's their start_date? Is it always the date when the construction work of the latest layout ended, or is assumed that the shape might have been improved but the road is still the same? What's the threshold?

  1. The center line must be at the exact same spot as it was "back then"
    While some old maps are amazingly accurate, it's impossible to tell if the road was right under the current, or maybe under the right hand side shoulder of the current road.
  2. The old course must have been within the current road surface
  3. The direction of the old form must have been (reasonably) identical to the current road
    Current roads are wide from the 19th century perspective: if the old course followed the landscape to avoid inclines, it can run at a significant angle (0-30°) to the current road while still staying within the land used for the current highway. There might still be remains of the old road within some tens of meters at places, either as driveways or cycleways.

If start_date is always the date of the latest layout, what's the threshold for layout change? Some examples:

  1. An old one lane unpaved road is improved and made into a paved cycleway? No turns are changed, some inclines might be lowered; does it change the start_date?
  2. A oneway two lane road is made into a one lane road with dedicated parking slots, just by erecting one traffic sign and some white paint on the asphalt; does it change the start_date?
  3. A new turning lane is constructed at the edge of a road; does it change the start_date?
  4. A new underpass for cyclists is added, which often is technically a new bridge on the road; does it change the start_date? There's now two real world objects (a bridge and a road) presented with one way in OSM - one tag or explicit tags bridge:start_date=* (or start_date:bridge=*) and start_date:highway=*? If so, why different tags depending on the situation?
  5. A central median/crash barrier is added, but the width required is "stolen" from the paved shoulders - the road needs to be split to two oneway ways in OSM; does it change the start_date? Compare with turning a 2 lane twoway road into 3+3 dual carriage road.
  6. A road is widened significantly at an intersection, say from a 2 lane road to separate 3+1 lanes for both directions of road - effectively the road on other side of the median is at the old location, and the other side of the median is a totally new construct.

And if start_date is always the date of the latest layout, what tag to use for "this road was originally constructed here in ..."?

But we should not forget the double carriage case in general, where the old road followed (some parts of) the other carriage way, but the other is completely of newer origin.

bus lanes

Resolved: Big discussion when-was-it.

lanes=* has always been described as "Total number of physical travel lanes making up the way". Some tag dedicated bus lanes on a road as busway=*, others with the appropriate one out of lanes:psv=*, lanes:psv:forward=*, lanes:psv:backward=*. In some countries other road users are allowed to use these lanes for turning, or at some signposted times of the day.

Tags Notes
busway=opposite_lane Does that allow taxis? Bicycles? What if it's available to all outside rush hours?

lanes:psv:backward=* oneway:bus=no

Yes check.svg



landuse=forest used to describe "area set aside for forestry". A recently full chopped forestry land parcel is still landuse=forest, but with no or very few trees standing: natural=scrub or natural=heath describes the physical condition for at least several years.

natural=wood was from the start "here grows trees". Without landuse=forest on the same way, i.e. as an area not used for forestry, it logically meant "primeval woodland"; with the aforementioned tag, it became a maintained forest.

A bot was used in 2008 to remove natural=wood from all ways with landuse=forest as "redundant", effectively breaking the tagging. That was a result of some discussions on the talk-de mailing list (in German), possibly because nobody replied to this: [1], but the confusion and discussions predate that, for example [2].

Forest trails

Just link this: Consolidation_footway_cycleway_path.


Traditionally large rivers are drawn as a way marking the center line of the river (tagged waterway=river, accompanied by a second way drawn along the riverbanks and tagged with waterway=riverbank. Early 2008 (as far as I remember the date correctly) even the remaining rendering rules were improved to persuade users not to tag the riverbanks with natural=water.

A recently (May 2011 according to this wiki) reintroduced practice is to "allow" users to tag the areas between the riverbanks as

Farmland and other open areas


We used to have landuse=farm for all open areas used for farming, including tillage and pasture and all that and excluding only the farm yard. The distinction between landuse=* and natural=* was clear, but to certain extent only in the minds of the early pioneer mappers. But there are several kinds of open, treeless areas, some naturally occurring biotopes:

  • Tillage; land used for growing crop - landuse=farm for certain
  • Pasture; land growing mostly grass, with domesticated animals grazing
  • Meadows; land growing grass/flowers, harvested for later feeding to animals. Isn't that landuse=farm + natural=meadow or natural=grassland?
  • Perpetual meadows; land growing grass/flowers without any, or seldom human intervention - now we have an approved feature natural=grassland and others use landuse=meadow
  • the strips of land between tillage fields, often encompassing a stream, growing grass and other non-woody plants (is it just a narrow perpetual meadow?).
  • Scrubs; areas with woody but low plants; you'd probably get bruises if you walked through them without clothing - natural=scrub
  • Fell (as in a nordic hill) top grassy areas mixed with some patches of dwarf shrubs and patches of bedrock visible - mix of tundra and moorland
  • moorland; is it just a subtype of heath?

And some clear, distinct cases

  • areas set aside as lawn, grass with (mostly) no other purpose than filling and giving some structure to the area (one use of landuse=grass)
  • heathland; not originally described in detail enough, esp. when no tag for meadows was given - "low bushes", and translations might have recommended for areas more meadow-like.

And some tags in use that might or might not make sense:

  • natural=fell - grass + "often grazed" + "upper laying" - how come it's distinct?
  • natural=grassland - anything that grows just grass or grassy plants

Trying to use Wikipedia or Wiktionary definitions of the words used so far in the osm tags leads us nowhere. I'm also certain there's some biotopes missing from this, esp. in the warmer areas of the planet.

In the end this comes down to other problems with tagging of natural areas, and different characteristics of different forests - a venture no one has yet tried to nail down. Some have drafted their ideas, or usable tags, at (Key:plant_community, Environmental OSM, Key:habitat, Key:taxon and Proposed_features/landcover). Perhaps we should imitate the existing classification systems, but then we'd need some botanists to help us. Just as an example: depending on the soil, the undergrowth of a forest can be anything from mainly lichen to grass to shrubs, or a mixture of them. And such areas can be put to different uses.

used for ploughed/manipulated? grass? small woody plants? trees? tags natural language
tillage ploughed as far as crops are grass-like plants no no landuse=farm
pasture (field) smoothed yes no no landuse=farm
hay meadow smoothed yes no no landuse=farm

Some suggest landuse=meadow

- none or minimal yes no no natural=meadow / natural=grassland
pasture (forest) hazards removed initially, minimal thereafter yes none or minimal yes natural=wood forest pasture
dwarf-shrub habitat none or minimal minimal dominant none or exists depending on trees natural=wood / natural=heath heath or Fi:Kangasmetsä

One tree, or many

Unresolved: No discussion as of yet.

It's obvious when a tree is solitary, or when it has many treetops, but branches at a height. But then there are border cases; does a tiny bit of air between the branches, all the way to ground level make a branched and maimed tree suddenly two trees? Pictures coming sometime.