Proposed features/boundary=forestry( compartment) relations

From OpenStreetMap Wiki
Jump to navigation Jump to search
boundary=forestry(_compartment) relations
Status: Draft (under way)
Proposed by: Penegal
Tagging: boundary=forestry
Applies to: relation
Definition: Tagging scheme for areas used for forestry
Drafted on: 2020-12-27


Overhead imagery of a forestry compartment
A forestry compartment within a forestry area. The cut lines are visual indications of forestry activity

A forestry area is a mostly-wooded land with defined boundaries, which has at least one of the following characteristics:

  • The area is used for production of lumber, firewood, bark, cork, sap, and related forest products like leaves, mushrooms, wild berries or herbs for human consumption and use
  • Management policies are employed to alter the number, species or shape of trees with activities such as planting, thinning, felling or pruning trees; mulching; enrichment; protection against damage from animals; or coppicing.

A forestry compartment is a numbered sub-division within a forestry area, physically materialized with cleared, visible boundaries.

Proposed are that:

  • New tagging is adopted for forestry areas and forestry compartments, and competing tagging schemes are deprecated, as described in the table below.
  • Definitions for forestry areas, compartments, and associated terminology are adopted.
  • Tagging techniques for modeling forestry boundaries and land cover are adopted as a community recommendation to mappers.

Proposed tagging changes

Summary of Tagging Changes
Tag Action Description
boundary=forestry Approve Describes the boundary of a managed, mostly wooded, forestry area. The forestry area may include multiple disjoint parts and may encompass non-wooded areas such as glades, screes, or ponds as long as the collective area is administered and managed as a whole through forestry. This tag is combined at a minimum with name=* and operator=*.
boundary=forestry_compartment Approve A forestry area may be divided into forestry compartments. A forestry compartment is a referenced area wholly within a boundary=forestry, which has visible boundaries evidenced by the removal of trees along the boundary line or the presence of physical markers. The tag compartment_ref=* is used to indicate the compartment's unique identifier.
landuse=forest Deprecate This tag has been used for forestry areas as well as unmanaged wooded areas. As a result, data consumers have interpreted this tag to mean a wooded area with no specific forestry meaning implied. As a result, this tag is effectively a synonym for natural=wood. Therefore, it is proposed to formally deprecate this tag in favor of a gradual replacement with natural=wood.
Deprecate These tags are synonyms of boundary=forestry and boundary=forestry_compartment for tagging forestry boundaries. The term "forest" in English refers to any wooded area, as opposed to "forestry" which means "the science or practice of planting, managing, and caring for forests". In order to firmly disambiguate forestry tagging from ordinary wooded areas, it is proposed to adopt the forestry variants of competing tags.
natural=wood No Change This tag continues to be used to tag tree-covered areas, regardless of whether they are located within forestry areas or whether they are located within unmanaged wooded areas. This proposal maintains this tag's "de facto" status in order to remain neutral on the question of whether to use landcover=trees as a replacement of or in addition to natural=wood to describe wooded areas.
landcover=trees No Change This proposal takes no position on the usage of landcover=trees. For simplicity and readability sake, this proposal will describe the tag natural=wood throughout this proposal to refer to wooded areas as it is the more popular tag.


  1. There are currently 6 different ways to tag forests and forestry areas. Not only is this situation confusing to mappers, it also prevents data consumers from having a reliable mechanism to differentiate forestry areas from other features.
  2. The tags boundary=forestry and boundary=forestry_compartment are already in use with approximately 6,000 usages. In addition, there are 8,000 usages of synonyms boundary=forest and boundary=forest_compartment. Thus, with 14,000 existing usages, this tagging scheme is already in active use. This proposal establishes a community consensus for standardizing the usage of these tags consistent with OSM practices and norms.
  3. The word forest means "a large tract of land covered with trees" and does not by itself indicate that there is forestry activity happening on the land. By adopting terminology that uses the word forestry, and deprecating tagging that uses the term forest, it will be clear to mappers that the tagging is intended to apply to forestry areas and not necessarily any forest.
  4. Existing objects tagged landuse=forest may be forestry areas, or they may simply tree-covered areas. Deprecation of this tag provides an indication to mappers to that a feature has not been examined to determine whether it is a forestry area or simply wooded.

TODO - content below being reworked

Some definitions


A given, limited area is "managed" when its existence, as a physical area with defined, precise boundaries, necessitate a significant human work, with perceptible effects; "management" is about the land itself, its trees, its glades… but also about its limits.

Please note that this only works for delimited, verifiable areas: the boundaries must be verifiable. What it exactly means for a given country/region depends on local laws and customs; forestry areas boundaries may be materialized with signs or painted marks along the area borders, but local laws and customs may prescribe carved marks or other methods.

This proposal is not about any wooded land: to be considered a forestry area, a (mostly) wooded land must have defined, finite limits; it must be significantly managed as explained below, and this management must be verifiable (distinguishable consequences of forestry work, specifically materialized boundaries…).


"Forestry" is the management of forests, in a broad sense.


A wild wooded area can be considered "managed" when its limits are cleary materialized, for instance with paint, boundary stones, cutlines… These show that humans have the will to consider this area as managed, at least at the very basic level of marking its boundaries; given the effort and the money needed for the materialization of the boundaries (some hundreds of €/$ per km at each pass as of 2020; consider about a pass every ten years), doing it on an unmanaged forest, which can't be expected to make money, would be simply throwing money away. On the other hand, marking boundaries, even simplistically, is the very first act for forest management:

  • it tells neighbours that here starts an area potentially used for wood extraction (with related dangers and property rights);
  • it allows the forest workers to know where their work starts and stops.

Consequently, materialized boundaries are enough to mark a wooded area as "managed", as a "forestry area".

Forestry area boundaries must be distinguishable of standard land lot boundaries: if the forestry area is only delimited with land lot boundary stones, without any signs, marks or paint highlighting the difference between its boundaries, as a forestry area, and a mere land lot boundary, the area should not be modelled as a forestry area, as its boundaries are not verifiable. That also prevents treating each single wooded land lot as a forestry area: a wooded land lot is not necessarily managed nor used for forest-related resource extraction; conversely, marking its limits in a specific way to highlight it being the place of forestry work, instead of a mere land lot, would be the very first step for forestry, before starting to work on the trees or the land.

Forestry works

Forestry is also, of course, the management of the trees. The number, species or shape of trees are altered by human by any of the following, with or without extracting wood for human needs: trees undergo a selection for biodiversity, phytopathologies, wood quality or to regulate the respective proportions of present species; they can be planted, selected, fallen or pruned; the area undergoes forestry works such as mulching, enrichment, protection against damages from animals, coppicing, thinning…

Forestry area

Consequently, a "forestry area" is a mostly wooded land with a distinct existence (that is to say, it is considered, as a whole, one and only one), with verifiable, distinct and precise boundaries, different of standard land lot boundaries; it must also be in one of these two cases:

  • the area is significantly used to produce lumber, firewood, bark, cork, sap, and related forest products like leaves, mushrooms, wild berries or herbs for human consumption and uses; this may be one of the official goals of the modelled area, or common knowledge. The extraction of resources may not take place everywhere on the area: some places within the area may be subject of extraction prohibition, permanently or "temporarily" (note that, in forestry, "temporary" may mean for some decades); if such places are considered part and parcel of the managed area itself (a prohibition is an act of management), they are not to be treated apart;
  • the area is subject of forestry works, that is to say alteration of number, species or shape of trees by humans by any mean, including the following, with or without extracting wood for human needs:
    • trees undergo a selection for biodiversity, phytopathologies, wood quality, maintaining the wooded state of the area or to regulate the respective proportions of present species,
    • they can be planted, selected, fallen or pruned; the area undergoes forestry works such as mulching, enrichment, protection against damages from animals, coppicing, thinning…

Again, such operations may not take place on the whole area: some places within forestry areas may not need such operations for decades or may be voluntarily excluded of them; if these places are considered part and parcel of the managed area itself, they are not to be treated apart.

By exception, if the goal of the management applied to the wooded area is totally and verifiably caused by the area being granted an official status of protected area (for instance, its purpose is only maintaining the biotope of protected species), the area is to be considered a protected area, not a forestry area, as the first notion describes it more accurately. Using this exception should be justified using the source tag, to prevent dispute over the status of the area.

Forestry areas and protected areas

Forestry areas and protected areas can overlap: in this case, they should be modelled separately, with a boundary=protected_area relation and a boundary=forestry relation. When this happens, in the overlapped area, the protection status of the protected area prevails over the resource extraction rules of the forestry area.

A forestry area itself, with logging or other resources extraction, may also be considered a protected area by local authorities and laws; in such cases, the deciding factor is the purpose of the use of forestry for the management of the area. As told above, if the goal of the management applied to the wooded area is totally and verifiably caused by the area being granted an official status of protected area, the area is to be considered a protected area, not a forestry area, as the first notion describes the area more accurately. Then, use boundary=protected_area.

If the protection is not the sole goal of forestry within the area (typically, when the sale of wood is also a declared goal of the area), modelling the area with separate * and * relations would imply that the area being protected and being used through forestry are separate notions. To prevent that, the area would be mapped with a single boundary=forestry relation, and accordingly tagged with protect_class and other relevant tags related to protected areas, to reflect the dual status of the area.


"Compartment" designates a single piece of a given forestry area, which is used as a whole by the area manager and passers-by, and which is given a reference, often a number, which is unique inside the given forestry area; it does not mean a land lot or a cadastral parcel, which are about taxes and ownership. Though boundaries of forestry area/compartment and land lots frequently overlap, a given forestry area or compartment may be composed of multiple land lots, and a single land lot may be divided in multiple forestry areas/compartments. This proposal is not about land lot boundaries, but about forestry boundaries: if a given land is managed by the forestry area manager as a single forestry area/compartment, it is to be modelled as such, whatever the land lot boundaries.

There may be no compartments in a forestry area; some are too small to justify creation and maintenance of compartments, other may be managed following methods which make compartments essentially useless. Consequently, the modelling of compartments is not required. They should be modelled if present, but are in no way expected to be present in all forestry areas.

Please also note that in some countries like Poland, though compartments exists, they are often not materialized on the ground. Their limits and existence may be announced in public GIS data, but they basically fail to have verifiable and distinguishable existence and limits on the ground. Consequently, in such cases where compartments are not materialized, they are not to be mapped.

Current tagging limitations

Land cover

Rendering example of a forest including a grassland which is not an role:inner member of the forest multipolygon
Rendering example of a swamp whose trees are used for wood production

The main issue with current tagging practices is that landuse=forest confuses forestry areas and managed wooded lands. Though forestry areas mainly covers wooded lands, this is not always the case; forestry areas may contain screes, ponds, scrublands, grasslands… which are not covered by trees, and are consequently not supposed to be tagged as wooded areas. Such non-wooded areas are still considered by foresters or general public as part of the forestry area, as they may be:

  • managed by the manager of the surrounding forest;
  • under reforestation;
  • lands with a distinct, wood-related ecosystem (glades for instance, which essentially exist because they are in a wooded area, else they would be called grasslands);
  • legally considered part of the surrounding forestry area, i.e. they are subject to the same subset of laws, rights and obligations (Régime forestier in France, for instance).

An example to explain: let's represent a forestry area with a broadleaved part and a needleleaved part, with a pond in the middle.

Example forestry area

Because of the current confusion between managed wooded lands and forestry areas, the forestry area will most likely be mapped with landuse=forest. Regarding the pond, such non-wooded areas are currently mapped as separate features, often as an inner limit of the surrounding landuse=forest. In cases where this area is considered part and parcel of the surrounding forestry area, this modelling creates an erroneous separation from the surrounding forestry area.

Mapping the pond outside the forest, with non-overlapping polygons, and different unrelated entities for one forestry area.

Some mappers may be tempted to bypass this limitation by voluntarily not modelling the pond/scree/scrubland… as an inner limit of the wooded land entities. In such cases, upon rendering, the rendering engine interprets this situation as follows: the pond/scree/scrubland is part of the limits of landuse=forest (multi)polygon without being excluded of it; as this polygon is for wooded lands, I render in it the trees pattern I use for forest/wood. This tagging issue thus leads to erroneous renderings which incorrectly tell the viewer that the area is covered with trees.

Mapping the pond inside the forest, with overlapping polygons, creating misleading rendering; there is still no logical link between the different entities composing the forestry area.

To summarize, the problem with the current mapping practices is that mappers will use landuse=forest (or natural=wood) for forestry areas as they confuse forestry areas and wooded lands. These are different notions, and these tags were designed for wooded areas, whereas forestry is not always about wooded areas. Consequently, using these tags makes renderers display the tree pattern in landuse=forest/natural=wood polygons, even if they were used with forestry area in mind. Oblivious of this confusion, the mapper trying to map such scree/pond/glade has a difficult choice. When encountering a glade/pond/scree…, considered part and parcel of the forestry area, the mapper can either:

  • exclude it of the surrounding landuse=forest/natural=wood entities: the rendering is correct, but the mapping does not reflect the reality, as the feature is in fact considered part and parcel of the forestry area; or
  • include it in the surrounding landuse=forest/natural=wood entities: the mapping reflects the reality, but provokes a misleading rendering.

Please note that none of these possibilities removed the confusion between managed wooded areas and forestry areas.

Some forestry areas may even be something else than natural=wood or landuse=forest: for instance, they can be a scrubland (natural=scrub), which often forms after a clearcut followed by natural regeneration (i.e. no plantation); in this case, the land will not be again tree covered until the seedlings grow, which can take decades.

A forestry area can also be a swamp whose trees are felled for human consumption. Such area is tagged with natural=wetland and landuse=forest, confusing their rendering by overlapping their texture. This rendering logic allows consistent rendering of some areas where, in reality, a wooded area physically overlaps with another feature (for instance, a wooded residential area); anyway, for forestry areas such as this one, this double rendering troubles viewers and fails to describe the reality in an understandable way.

Different properties for different parts of the same forestry area

Forestry areas, especially those spreading on hundreds or thousands of hectares, are typically not of a single type: they can have broadleaved areas and needleleaved areas, sempervirent and deciduous areas… Currently, such areas are each individually modelled with their own landuse=forest/natural=wood (multi)polygon with the relevant tagging.

In this situation, as shown in examples above, the forestry area, instead of being modelled as a unique entity as it is named and managed on the ground, is in OSM split into in multiple entities, without a logical link between them for modelling their belonging to a single, unique real-world forestry area.

One could use a multipolygon landuse=forest/natural=wood relation with the relevant tags, for instance leaf_type, applied on its role:outer members. Nevertheless, such tagging is not currently taken into account by the renderers, and chances are it will never be, as it would be very complicated for a renderer to compute the rendering of an area by querying multiple overlaping entities. Besides, multipolygon role:outer polygon members are not supposed to have common borders, as multipolygons weren't designed for such uses, but for separate polygons belonging to one unique entity.

One could also use a site relation, but it is designed for multiple nodes belonging to a unique entity, not for polygons, neither separated nor adjacent.

Forest compartments

The other main problem is about compartments. Usually, but not always, wide forestry areas are divided in compartments, each with a reference which is unique within the parent forestry area — for instance a number —; this allows referencing a part of the forestry area as "Forest of X, compartment #12". There is a relation tagging scheme for compartments: one can create a boundary=forestry_compartment relation – or boundary=forest_compartment, which is essentialy a duplicate – to model such compartments. These relations are not widely used, though, and there can be many reasons for that:

  • this kind of relation has not been formally approved yet;
  • the tags to be used with it are not clearly stated;
  • there is no clearly stated way to link compartments and their reference in it to the forestry area they are part of.

This leads to the current situation: although most forestry areas — at least those big enough, or in countries with a significant development of forest management — are divided in compartments, there is no standardized way to map them in OSM. Consequently, most users simply don't bother with compartments, and map forestry areas as a big landuse=forest/natural=wood entity without bothering with their subdivisions.

Besides, as compartments often have their limits materialized on the ground — painted markings, cutlines, signs… — with their reference to allow passers-by to find their bearings, in some countries they are displayed on the publicly available maps. This is not currently possible in OSM: even if this information is in the database, it is not modelled in a standardized way, and the renderer cannot know where to find this information. That prevents users to use compartment data to find their way, even if they are available in OSM; it also discourages contributors to add this data in OSM if they want to, and renderers to bother about them.

This created inconsistencies in the OSM database, in areas with compartments, where users tried to model forest compartments or tagged them in order to have them rendered with their reference:

In all these tries, one can see an additional problem: the trees symbols used for wooded areas get rendered twice, once for the compartment, and once for the forest it belongs to. This creates a buggy rendering and decreases the readability of the map, by preventing the viewer to correctly see the leaf_type symbology.

Physical vs. logical

Overall, as told above, a major part of the problem is that there is a confusion between what is physically a forest or a wood — a wooded area — and what is considered a forestry area, by general public, forest workers or administration; even the naming is confusing, as forestry areas are commonly called "forest" or "wood", without differenciating forestry areas and unmanaged woodlands. Consequently, these notions (wooded lands, aka. "physical forests", and forestry areas, aka. "logical forests") should be treated separately, with different tagging.

Forestry areas and protected areas

Some forestry areas are currently also considered as protected areas, as their legal status allows to consider them as protected areas under IUCN classes Ia, Ib, II-VI; however, some of these areas are managed for wood production among other goals, and should also be considered forestry areas. In addition, legal statuses and their connections with IUCN classes differ: US National Forests and French public forests are managed following similar goals (wood production has the same level of importance as environmental protection and access by the general public; the area is considered multifunctional), but US ones are typically considered protected under IUCN Category V[1], whereas French ones are under the legal status of all French public forests[2], which does not consider them as protected areas despite offering the same level of protection as US National Forests.

Typically, a US National Forest is not a forestry area by itself; only some parts of it are. That being said, the proposal must acknowledge that a forestry area can also be a protected area, should a US National forest or similar forest be a forestry area by itself.

In cases where the forestry area overlaps a protected area, one must ask oneself: "Is the forestry area subject to the general restrictions of the surrounding protected area, or is it some sort of enclave, out of these regulations?". In the first case, one may simply map the forestry area without bothering about the surrounding protected area: the area encompassed by the forestry area will then be mapped as also encompassed by the protected area, modelling the fact that the rules of both areas apply. On the other hand, if the forestry area is not subject to the regulations of the surrounding protected area, then it is a de facto enclave in the protected area, and should be mapped accordingly as an role:inner member of it.


Examples were given above; others could be found in the OSM database, for instance here. This forestry area spreads on hundreds of hectares of wooded land but, to be modelled as a single landuse=forest entity, the contributor needed to pretend that the whole forestry area is needleleaved, whereas it is not: there are broadleaved areas and mixed areas, which do not necessarily follow compartment or forestry area boundaries. One cannot draw the needleleaved, broadleaved and mixed areas of the forest or compartments in their own entities, then merge them in a single entity representing the forestry area or its compartments: the site relation is designed for separate node elements, not for polygons; the multipolygon relation is not what's needed either, as it is not for adjacent polygons. Furthermore, even if compartment boundaries were described with one of these, there is currently no way to represent their belonging to a single forestry area.


To allow modelling of forestry areas, one should no longer rely on landuse/natural tags, which describe the physical world. Instead, the proposed tagging creates a boundary=forestry tag for modelling forestry areas, and formalises boundary=forestry_compartment for forest compartments while deprecating the duplicate boundary=forest_compartment.

In the example case above, with a forestry area partly needleleaved, partly broadleaved, and a pond in the middle of it, the polygons would be modelled this way, with a surrounding boundary=forestry relation:

Example modelling, with polygon limits and boundary=forestry relation.

Physical/land tagging

As told above, the proposed tagging scheme is not intended to replace landuse=forest and natural=wood tags. These tags should still be used to describe the actual, physical state of the lands. That being said, the notion of forestry area encompasses and enhances what is generally mapped with landuse=forest: the land is used (managed) as a forest. Consequently, for consistency, landuse=forest will be deprecated; its role of describing wooded lands would be taken over by natural=wood, which is already widely supported, so no rendering disruption is expected from this deprecation.

As told, natural=wood currently describes a wooded land, and interpretations differ about whether such entities are necessarily or possibly unmanaged. The proposal being about the human management of lands (the so called logical land description), there must be a tag, and preferably only one, to describe the physical land; natural=wood seems clearly relevant for this role of describing a wooded land. To prevent confusion, it will be clearly defined as describing a wooded land (land covered by trees), whether managed or not (note that, if the land, wooded or not, is better defined with another tagging, such as natural=wetland+wetland=swamp as the example above, this better tagging is still to be preferred over natural=wood.)

This definition of natural=wood is related to landcover=trees, which describes a land covered by trees. This tag is far less used than natural=wood, but many mappers use it, maybe for slightly different uses than natural=wood, so the proposal does not deprecate landcover=trees. Still, this tag is, at the time of the proposal, neither rendered by the main OSM renderer nor taken into account by editors like JOSM or iD. Therefore, its use for describing the physical land in replacement of landuse=forest cannot be recommended, as it would for instance break the rendering of areas where the proposed tagging would be applied. Consequently, natural=wood is recommended instead for the retagging of existing landuse=forest entities.

Following the deprecation of landuse=forest, entities with this tag would be considered for migration to the new tagging scheme: they would have to be retagged with natural=wood, and a boundary=forestry relation would have to be created if the area is managed, i.e. it is a forestry area. This should not be a fully automated operation: deciding if a wooded area is managed or not will often be done on a case-by-case basis, or need local insight.

Logical/management tagging

The proposed tagging scheme creates a superimposed scheme, which is about management-related divisions and boundaries (the so-called "logical" level), not about the underlying land state, as these are different, although related, notions. Consequently, the proposed relation types must not be used to tag characteristics, especially physical, of the enclosed lands, as leaf_type or water; such informations belong to the underlying (mostly natural) entities, and the proposed relation types must only contain data about themselves, i.e. about the compartment or the forestry area themselves, as management-related entities.


This relation is used to describe a forestry area as a whole.

Please note that:

  • a forestry area is not always divided in compartments: boundary=forestry relation must not enclose any boundary=forestry_compartment relations if their are none or if they are not physically materialized;
  • a forestry area may include smaller forestry areas, themselves either divided or not in compartments, so a boundary=forestry relation may enclose smaller boundary=forestry relations:
    • it may be the case when a forestry area is composed of separate pieces of managed wooded land, each with a different name, which are still part of a larger forestry area; one could use landuse=forest+name=* polygons, but that would come with the same issues than described above, like the lack of overall coherence between pieces of the smaller wooded areas with different tags (leaf_type, for instance);
    • conversely, and whatever the physical features composing it, there can be large forestry areas divided in adjacent forestry areas with different names, operators, owners… — for instance, historical forests which were later divided in many owners, but are still generally known under a single name —; in this case, the bigger forestry area AND the smaller, adjacent ones, having both distinct and as-a-whole existence, would each have their own boundary=forestry relation, and the link between them would again be a spatial one.

Finally, as explained above, a forestry area can also be considered a protected area under UICN classes Ia, Ib or II-VI by local laws and authorities; for instance, US State Forests are considered protected area under IUCN class VI, which allows significant extraction of ressources. Consequently, boundary=forestry applies to all forestry areas, either considered as protected or not, as long as protection is not the sole goal of the usage of forestry. If protection is the sole goal of the forestry applied in the area, use boundary=protected_area; else, use boundary=forestry, and some protection-related tags if applicable.

The above paragraph means that, should this proposal be approved, if some forestry areas are currently tagged as protected areas, they are to be re-tagged as boundary=forestry+protect_class=*+… as described below.

Key Value Discussion
type boundary Mandatory
boundary forestry Mandatory
ref The forest ref Optional. Such reference is probably country-based, so one should use a subtag, like ref:FR:ONF or ref:US:USFS, whenever applicable; the choice of this subtag for a given country is up to the local OSM community.
name The forest name Recommended. If applicable, the forest can be named with this tag, and, maybe, its subtags, like name=Forêt de Fontainebleau or name:de=Wirlinger Wald.
operator The forest operator Recommended.
owner The forest owner Recommended.
admin_level The level Recommended. Use for giving the admin level of the owner if it is a public or government-related entity; for instance, a French forêt communale would be tagged admin_level=8.
source The source used Optional.
protect_class Protection class ID Optional. Use if the forestry area is also considered as a IUCN-compatible protected area by local authorities or laws.
protection_title The forest owner Optional. Use if the forestry area is also considered as a IUCN-compatible protected area by local authorities or laws to give its local protection title, like "National Forest".
related_law The related law Optional. Use for giving the title of the edict or legal text which establishes or regulates the forestry area, mainly if it is a protected area.
leisure nature_reserve Optional. Use if the area is also considered a nature reserve.


This relation is used to describe a physically materialized forestry compartment, when the forestry area is physically divided in materialized compartments. In that case, there must be a 1:1 correspondence, that is to say one boundary=forestry_compartment relation by compartment, and one compartment by boundary=forestry_compartment relation. Consequently, boundary=forestry_compartment relations should never overlap; edition softwares should warn against such cases. The modelling of compartment limits, as for forestry area limits, is alike the one of all boundary relations, especially regarding outer/inner limits and enclaves/exclaves.

The link between a forestry area and its compartments is spatial, as for all boundary relations: a compartment is considered part of a forestry area if and only if its boundary=forestry_compartment relation is spatially fully encompassed by the boundary=forestry relation. It is not useful, and may be counterproductive, to include boundary=forestry_compartment relations as role:subarea members of the corresponding boundary=forestry relation; such modelling is disputed and widely considered redundant.

A boundary=forestry_compartment relation should have the following tags:

Key Value Discussion
type boundary Mandatory
boundary forestry_compartment Mandatory
compartment_ref The compartment ref Recommended. According to the Map what's on the ground principle, this tag's value must be the one displayed on the ground; if official databases records a more detailed ref — for instance one including a forest ref —, one should remove these additional details and only keep the compartment ref as displayed on the ground. This is not a simple ref tag, as such tags are already used on existing boundary=forestry_compartment without standardization, so their current value cannot be assumed to be what compartment_ref means.
name A name Optional. If and only if applicable, the compartment can be named with this tag. Don't confuse name and ref! A name is a word or group of words with a intrisic meaning, whereas a ref is a sequence of numbers/letters whose meaning exists only in the context of a larger forestry area. If you want to use this tag, think twice: it is likely you're trying to tag a ref, not a name, as typical compartments have only a reference, not a name.

As a forestry area is supposed to be managed as a whole, it is not useful to tag each compartment belonging to a forestry area with tags applying to the whole forest; such tags should be applied on the enclosing forestry area entity. For instance, the operator of the forestry area is most probably unique, so an operator tag should not be applied on each compartment, only on the forestry area entity.

Migration of existing entities to the new tagging

The following table explains what to do upon encountering entities which do no use the new tagging scheme. It is ordered in a specific way: when you encounter an entity which matches several tagging cases, use the first one in table order to know what to do.

Note that current landuse=forest entities may in fact encompass several distinct forestry areas, with different names, compartmentalisations, operators… A simple re-tagging of the entity, with the creation of a boundary=forestry when applicable, is technically enough, but, if you know that the area is in fact split into several distinct forestry areas, please map these with distinct entities.

Current mapping To be re-tagged? How to re-tag?
natural=wood+managed=no Yes natural=wood
natural=wood+managed=yes(+operator) Yes natural=wood, enclosed in a boundary=forestry relation, tagged with operator if applicable (and known)
natural=wood not encompassed in a boundary=forestry relation Yes, if you know that the area in fact describes a forestry area natural=wood, enclosed in a boundary=forestry relation with applicable tags
landuse=forest encompassed in a boundary=forestry relation Yes natural=wood
landuse=forest not encompassed in a boundary=forestry relation Yes natural=wood, enclosed in a boundary=forestry relation

Tagging examples

US National Forest

A timberstand improvement area in a US National Forest would be tagged this way:

Forestry area in a UK National Park

In such case, the forestry area would be subject to National Park laws/regulations, so it would be mapped as a separate entity, but not as an enclave in the National Park:

Being encompassed in a protected area and not an enclave in it, the protected area laws/regulations would apply to the forestry area.

French forêt communale

The forêt communale de Chauffecourt would be modelled by a relation tagged this way:

Compartment #6 of this forêt communale would be modelled as a relation tagged this way:

Australian state forest

The Putty State Forest relation would be tagged this way:

What about this area? Is it to be modelled?

Here are some special cases where the applicability of the proposed boundary=forestry relation may need clarification.

Case Eligible? Why?
Illegal agricultural deforestation area No It most likely lack precise boundaries, which are a prerequisite to be modelled as boundary=forestry. In addition, once the deforestation is started, the area likely does not keep its wooded state for long and is thus out of OSM scope.
Legal forestry concession with illegal logging Yes Even if the logging is illegal, the area is wooded, and forestry is the goal of the concession which also explicits the limits of the area, making them verifiable. Of course, once the area is totally, permanently deforested, and the concession expired, the area is no longer a forestry area.
Wooded land, whose trees are managed (pruned, for instance) without border markings nor open GIS data No If the area limits are not materialized in a way different of standard land lots, and if the area does not have precise limits available in an open source forestry data, it being managed as a whole, unique entity is unverifiable and it is not to be modelled as a boundary=forestry relation.
The Black Forest No Even if it is a mostly managed, mostly wooded land, which is commonly known as a whole, single wooded area, with local forestry works, its boundaries are neither distinct nor verifiable, hence disputable, so it is not to be modelled with a boundary=forestry relation. That being said, the underlying physical, wooded lands themselves are still eligible to modelling with natural=wood.
Mostly wooded area used for mushrooms or sap collection It depends There is extraction of resources which are associated with wooded areas, so the only missing criteria is boundaries: are they distinct of standard land lot boundaries and verifiable, either on the ground or with open data? If so, you can use boundary=forestry relation; if not, it is not applicable, as the boundaries are not verifiable.
Wooded area used for logging (one can see stubs) but without materialized boundaries nor open data sources No Wood is extracted, but the boundaries of the extraction area are unverifiable.
A unique, whole forestry area with wood extraction and verifiable boundaries, but fragmented in distinct, separate areas Yes If the fragmented area is still considered in its entirety as a unique forestry area, then it is to be modelled with a unique boundary=forestry relation, which would enclose all the separate pieces of wooded land; use natural=wood for the physical wooded land.
A wooded area with verifiable and distinct existence and boundaries, used for wood extraction, which contain screes and glades Yes The boundaries verifiablity and wood extraction qualifies it for a boundary=forestry relation. As this relation is about forest management, if the screes and glades are considered part and parcel of the forestry area, the relation will not treat them as exclaves of the forestry area. These non-wooded areas are to be mapped by their own entities, and the wooded land around with natural=wood entities.
A wooded land lot, without traces of forestry work on the ground, which boundaries are mere boundary stones No Without traces of forestry work, the materialization of boundaries in a way which is distinct of standard land lots is required.
A wooded land lot, without traces of forestry work on the ground but with boundaries marked in a way distinct of adjacent land lots Yes The materialization of boundaries in a way which is distinct of standard land lots is enough (it is, as explain above, technically forestry), unless you are sure the specific marks used are unrelated to forestry.
A US National Forest Likely, no Typically, in US National Forests, forestry area covers only specific parts of the national forest. Thus, the forestry areas may be mapped by themselves, but the National Forest itself is not to be considered a forestry area.

Applies to

This tagging scheme applies to relations only, as for all boundary objects.


As for all boundary relations, such tagging should not be rendered by covering the enclosed area with a dedicated texture, but only have its limits rendered. How precisely is a complex question inside an already complex proposal, so further precisions will not be given.

In legends or editors, a forestry area could be labelled "Managed forest: mainly wooded, defined area managed by humans (border marks, wood extraction, management of forest stand…)". A compartment may be simply labelled "Managed forest compartement".

 Transition process 

Aside from the tagging transition, the whole OSM ecosystem will need adaptation to take the new tagging scheme into account.

As mentioned before, as natural=wood (the replacement tag of landuse=forest) is already taken into account by renderers and editors, the re-tagging of existing entities will most likely be transparent. The new boundary=forestry(_compartement) will however need adaptation from editors and renderers. Here is how this adaptation might best take place:

  1. request editors (mainly JOSM and iD) to support boundary=forestry and boundary=forestry_compartement; these editors' "Validators" should warn at upload when encountering one of the following situations:
  2. without waiting editor support, start the re-tagging of existing entities, and particularly the creation of forestry relations wherever applicable;
  3. when enough boundary=forestry(_compartment) exist to justify their support by renderers (several hundreds at least), request support of them by the standard tile layer stylesheet, Openstreetmap Carto.

Features/Pages affected

External discussions

  • Tagging list discussion initiated by author of proposal begins here.
  • talk list, 2017: Woods vs Forests

Please use the discussion page for further, "internal" (to this wiki) discussion.


Please comment on the discussion page.