From OpenStreetMap Wiki
Jump to navigation Jump to search


  • Write article on how key value pairs work, how stacking values and keys in one pair (e.g using : or , separators) is not really in the spirit
  • Write article on how inheritance key chaining is not always a good idea (link to learning python the hard ways description of how bad inheritance is) suggest using composition instead. e.g. no inheritance tags.

Tagging Farmland

The landuse=farmland tag makes up large segments of the landscape in the area I am particularly interested in mapping. The choice of how to split up these areas into manageable chunks has given me a bit of headache. There does not seem to be a clear convention on this and in its absence I have decided to define my own schema for how to add it for the areas I am mapping. After quite a bit of thought I have come to the following conclusions:
(The reasoning for which can be seen further down the page.)

What should NOT constitute a boundary for farmland

  • barriers such as hedges and fences.
  • highways which are a right over way over the land of a farmer but belong to the farmer and do not have a marked effect over the lands function e.g. public footpaths, tracks over farming land
  • highways that are for the sole purpose of allowing vehicles / people involved in the activity of farming
  • any natural tag. These define what the land is and not the use and can and should occur in tandem

What should constitute a boundary for farmland

  • other polygons with landuse tags e.g. residential, quarry
  • highways for the use by the general public and owned by another body that are not soley used for farming purposes.
  • bodies of water - should be tagged with fishing or perhaps irrigation?
  • woodland that is not being used to grow crops - even woodland belonging to a farmer - should be tagged as forest


Should the landuse=* tag be for tagging:

  1. the current use of land
  2. the current designated use of land

While similar sounding these two definitions have distinctive differences in meaning and also mapping implications. Consider the following example:

  • A field (farmland) of wheat that is bounded on three sides by hedges and a steam on the fourth.
  • Along the field side of the stream a thicket of trees and general scrubland a good 4 or 5 meters thick in some places has grown.
  • The wheat is growing in the field and covers most of the field save a a meter or so around each edge which is patchy grass between the hedge and the crop where the farmer would be unable bring their harvesting machinery to bear and has chosen not to sow.

  1. Should the polygon of landuse stretch all the way to the boundary of the stream or should it end several meters before the bank and then an area of natural=scrub be defined. Should this area be defined with the landuse=farmland or should it be left without a landuse=* tag as it is not being used by humans.
  2. Should the small area of patchy grass found between crops and hedging/fencing be defined as farmland or not?

In the example above the area next to the bank of the river and the area of the hedge is not being used for farming in the direct scene of growing crops or providing fodder for animals to eat however it may be providing other benefits that are almost as important to the produce of the land.

This land has no other discernible use and if asked a farmer would most likely say it was for farming.

What Is Farmland

An area of farmland does not necessarily have to have direct agriculture taking place in or on it (e.g. plants or animals) to be classified as farmland many other things should be considered farmland too. Below are some examples of what I consider farmland and should be tagged within farmland polygon boundaries for the following reasons:

Barriers such as hedges and walls

Barriers used in farmland are in some ways as important for farming as the areas where plants or animals are growing/grazing. They keep livestock contained, reduce erosion from wind and profile a habitat for pollinator insects necessary for crops in farming.

  • Farmland boundaries should not at first be bounded by ways that are hedges or fences if an area of farmland sits on the other side of the fence or hedge unless the user intends to map the areas with a natural tag for the field later on. Even still this may be a bad idea as hedges almost always have gaps to let humans/animals in or out.
  • Should natural tags go up to and join any barrier tag? YES - fields may not go up to the edge when used for crops but do when used for grazing and fields are rotated often - at least in the uk and as such we should not be trying to map the fields use as it changes too often.

Woods on farmland

woods sitting on farmland. Depending on who owns these woods. If they belong to a farmer or anyone not intending their use to be for conservation of nature they should be tagged with landuse=forest they should also be tagged with natural=wood in accordance with the landcover tag. The theory for this is that wood is asset that is owned and if a farmer wishes they can be chopped down for use as timber/firewood e.t.c. If the person who owns the land has allowed them to grow with the direct expression that they are for wildlife then a tag of conservation should be used. managed=yes/no may also be good to use here to give an indication of whether it is lassefair conservation or not. If it is just wilderness with no owner (does this ever happen) then landuse=no could be perhaps be used?


Rivers defined as an area - although farmers often use these as barriers they are not created by humans even if there coarse is often changed by humans (water needs to flow). Although a river may have a landuse by humans such as fishing, irrigation, transport it should not be marked as farmland and such provide a boundary between areas of farmland (this contradicts the barriers argument somewhat but not enough I feel.


See #Highways as Ways vs Highways as Areas

Tagging Direct Agriculture

If the user is concerned about mapping areas that are directly being used for farming production I would suggest they could better define such areas using the product=* or produce=* tag rather than the landuse=* tag which I think is more suited to general use of a land (e.g. not all of a landuse=quarry is used for mining).

Q? Should it be defined as the area owned by the farmer that he could conceivable use for farming if he so wished. In this case we would need to take the legal ramifications of such.

How to tag highways

landuse=highway - for a highway that's purpose is to allow the general public to get from A to B (a highway used for an explicit purpose would not use this tag e.g. a track on a farm would use landuse=farmland

I would not tag highways that where mud tracks on farmland as landuse=highway rather as landuse=farmland,natural=mud(or whatever),car=yes

IGNORE ABOVE - Just tag as highway=* or something (or use highway=highwaybank in the same way as riverbank) and make it a polygon dur... then also tag with landuse=farmland landuse=residential etc a highway is like a wall or anything else it not an explicit purpose. we don't have an area of highway for the purpose of it being a highway - in fact this should apply to all landuse tags if the landuse is to aid the purpose of what happens on them - so are highways for the use of driving - No (raceway maybe the only example of tagging a highway as a landuse) is farmland used for the purpose of farming - yes

another purpose other than to sustain humans then it is not farmland is for the production of food for humans... think about this a bit more. the highway tag is almost a subset to the landuse tag.

Highways as Ways vs Highways as Areas

There is a discussion under way as to whether or not highways should be mapped as areas on top of ways. While given OSM's philosophy this is at the end of the day a matter of the editors discretion I think there are several distinctions between the two ideas that need to be made clear (at least by my understanding) in order for a user to make up their mind on how to proceed.

Firstly we need to define what OSM is mapping. In this world as far as classical physics is concerned objects without areas do not exist. Therefore ways are not actual geographical representations of things rather implied references or vectors. Ways in OSM are viewable to humans when viewing a map only because a renderer gives them an arbitrary dimension of width (usually based on their perceived value to the user rather than the actual real world width.

Originally as the name implies Open Street Map was a intended to show streets with i imagine its main purpose being to allow people and computers to navigate said streets. Computers are the important word to consider in this sentence, as for a computer to navigate a set of streets from point A to point B it does not need to know anything about the actual area of the street, really all it needs to know is the length of the street, what other streets are connected and the rules of travel that apply on this street. For this ways are almost perfect.

Over time the remit of OSM widened and people began mapping more things than just streets in fact they wanted to map everything from swimming pools to sandbanks, not only did they want to map the area they wanted to map many attributes about each area. What a lot of people don't think about is that the mapped representation of a street as a way used for computers (or humans when it is rendered with width) to navigate from point A to point B (path finding) is completely different to that of mapping an area of land such as a field, marsh or street in order to provide a data representation of that of area and its attributes in reference to other areas of land (a land catalogue).

With the addition of an aerial layer in potlatch and the ease of armchair mapping it has become clear that the remit of osm has changed, intended or not, from a database mapping streets as ways just for path finding to a database that also aims to classify the geological make up of the two (mainly) dimensional aerial view of the worlds surface.

Bear in mind the actual end state of these two aims above. One would just be a map of ways of all streets or navigable highways in the world. The other would be a map where there would be no difference between looking at an aerial view in bing vs looking at mapnik view of OSM given the correct rendering colours. Let me elaborate on this: If it where possible the ideal OSM mapnik visual render would look identical to the tile view of Google aerial its just that the data making the OSM version would come from a database of classified things that we could sort, filter show or not show rather than just a photo that means nothing more than a bunch of rgb colour pixel values to a computer. The power of this should be obvious - want the total area of farmland in the world, number of houses - no problem.

Is this all a little bit off the track here? Well yes but no not really because we now understand the difference between mapping highways as areas vs mapping them as ways. If we map them as areas we are saying ok lets map these man made objects on the earth in the same way as we map buildings, fields, tennis courts, etc. When we map them as ways we are saying this is a representation of a humanly navigable route that a computer can understand and use in pathing algorithms.

The distinction we need to make is should we allow the render to add this area for us or should we do it ourselves at the data level. If we let the render do it then it will not be terribly accurate. If we do it, it will take a long time.

Either way if you are concerned about mapping landuse or land cover i would urge the user not to map over a road but map around it as you are mapping area not paths. If it concerns you that you are leaving areas unmapped then map them otherwise feel content in the knowledge that most renderers will hide the area with a road as roads are important landmarks and renders almost always render them larger than their actual real world width.

So really when it comes to mapping a road as an area think to yourself am i mapping this as a landcover / landuse or am I mapping this as a navigational tool.

My personal view is that the navigational tool is more important as due to the fact that area can roughly be approximated and implied from this data. However i do enjoy mapping land area so where i can i will map as an area as well as a way.

It is interesting to note that the river way and riverbank area tag pretty much summarises how these two concepts of ways as navigation aids vs areas as descriptions of the natural environment should coexist.

Highways as Landuse

This tag causes me a lot of problem in this discussion as it sits in the middle of these two definitions of street ways as navigation tools and street areas and land classification. I would not recommend we develop a landuse=highway tag or anything similar to this.

Think about this a bit more.

What about highways such as footpaths or tracks that are not built specificity for the purpose of travel and are not owned by a highway agency designated for public use? In this case i would just map them as a way as the land cover is whatever is underneath (e.g. grass (in a field)) i would consider mapping the land cover as dirt or whatever if it was a regularly used track.

Example a field with a track over it (the track is just over the grass of the field) should this be tagged as an area with landuse= something different from the surrounding farmland tag. I would argue that this land is being solely used for or to aid farming (assuming its not a public track) and it should be tagged in the wider area of farmland. If we start making a distinction we are going to end up being lawyers deciding if landuse is private or public e.t.c

Wikipedia vs Open Street Map

OSM and wikipedia are often banded about together as if they are similar but they are only similar in the scene of who enters data, the laws that apply to that data. They differ massively in how the entered data is interrupted. In wikipedia the data entered is meant to be used by humans not by computers. While there are many projects that try and use Wikipedias data for many other purposes its main aim is to allow people to read and learn about a particular topic. At the end of the day you could dump wikipedia into a massive excel spreadsheet and it would still be pretty useful as long as you could sort and find on title name.

The date in OSM on the other is designed to be used by computers before humans get hold of the data in any meaningful way. If you dumped all the data from osm into a spreadsheet it would be pretty useless to a human for most of the purposes we wish to use it for (navigating and finding where things are).