Add and correct logical axioms to environmental zone

[cmungall]

See also #136 Zones and layers. I agree layers should be material entities. I am not convinced zones should be immaterial, but assuming they are here are some refactorings and additions I proposed

• [ ] create a DP for designated areas. I propose a shortcut relation and a simple pattern environmental-zone and designated-for some <process>
• [ ] classify the zone hierarchy; currently lacks logical defs so cannot be done automatically
• [ ] Come up with a strategy for ORs. We need to be able to classify with Elk. We can use the el-shunt pattern
• [ ] consider candidates for moving to material entity branch:
  • coast
  • iron-reducing zone of petroleum contaminated sediment
  • marine park
  • rhizosphere (I feel strongly about this one; microbiome people will be confused if they cannot say microbes part of rhizosphere)
  • tectonic plate boundary (these have
  • anything under zone with "geographic feature" in the genus of the definition
  • or anything that has an ENVO material entity as the genus of the definition. E.g. Ramsar site: "A wetland protected by..."
• [ ] text definitions should be aligned with logical definitions [see my blog preview]
  • 'marine environmental zone' EquivalentTo ''environmental zone' and ('part of' some 'marine environment')'`: this is weaker than the text definition "An environmental zone which is bounded by material parts of a marine environment.".
• [ ] apply consistent DPs. E.g. ice accumulation zone is defined according to overlaps, its child glacial ice accumulation zone is defined according to a process. We should stick to one DP. (but note 'contains process' is too weak)
• [ ] Tackle low hanging fruit
  • e.g. the two subclasses of littoral zone can easily be axiomatized by part-of to marine/freshwater
  • two subclasses of coast
• [ ] Add DPs for vegetated area. I suggest a new shortcut relation 'primarily covered by'

Note on modeling zones as immaterial

One of the consequences of modeling zones as sites (immaterial in BFO) is that no material entity can be part of a zone. This can be desirable in some cases, if we truly want zones to be abstract areas defined entirely by relationships to physical areas, and we want to prohibit material parts. This can be useful for spaces in an animal body, e.g. gut lumen (but even here it causes confusion). But it's very unintuitive for concepts that area in the domain of discourse for scientists like rhizosphere.

I would strongly argue that rhizosphere is a region of soil, a material physical entity with material physical parts, it has a physical presence that has a material influence on its inhabitants, and is physically influenced by its parts and surroundings. This is in contrast to an immaterial entity which is an abstraction defined entirely by spatial relations to physical entities.

Immaterial entities are also harder to define in OWL, and are harder for humans to properly understand the OWL axiomatization. See https://github.com/EnvironmentOntology/envo/issues/684#issuecomment-468852208

Would there be any downsize to treating environmental zone as a region (i.e. material entity) with defined boundaries?

[pbuttigieg]

Approaching the broader issue first:

One of the consequences of modeling zones as sites (immaterial in BFO) is that no material entity can be part of a zone. This can be desirable in some cases, if we truly want zones to be abstract areas defined entirely by relationships to physical areas, and we want to prohibit material parts. This can be useful for spaces in an animal body, e.g. gut lumen (but even here it causes confusion). But it's very unintuitive for concepts that area in the domain of discourse for scientists like rhizosphere.

Yes, this is how we want zones to be represented here. I wouldn't really call them abstract, as they have boundary conditions, just immaterial.

I would strongly argue that rhizosphere is a region of soil, a material physical entity with material physical parts, it has a physical presence that has a material influence on its inhabitants, and is physically influenced by its parts and surroundings. This is in contrast to an immaterial entity which is an abstraction defined entirely by spatial relations to physical entities.

I tend to agree, but the defs that are used often refer to areas and regions, which are immaterial. It's likely these terms are used colloquially, so this is more a question of whether we want to take an ontologist's prerogative and say "What you really mean is...". This can be dangerous at times, but in the case of rhizosphere I think it's okay. Indeed, biosphere, cyrosphere, etc are all material systems.

Immaterial entities are also harder to define in OWL, and are harder for humans to properly understand the OWL axiomatization. See #684 (comment)

Would there be any downsize to treating environmental zone as a region (i.e. material entity) with defined boundaries?

I'm thinking of biogeographic areas and the areas used in remote sensing - from what I've parsed through, these really do refer to spatial entities that coincide with the ecosystems and features that define them, with a little error included. There are also some cases where the zone in question is defined by an arbitrary collection of things that aren't very closely related causally or otherwise (just a space bag that has stuff in it). There is always potential error is this call - the area defined may not actually line up with the defining entities themselves, so my intuition is that we want a little semantic distance between the spatial entity and the material one.

Do you think this is over-engineering? I could imagine that a link to a material entity can be made, but then we wouldn't have a 1:1 semantic descriptor for many geospatial classifications, inhibiting ENVO's role as a broker between them.

[pbuttigieg]

On to the list:

create a DP for designated areas. I propose a shortcut relation and a simple pattern environmental-zone and designated-for some

Happy with that. We should also include a 'designated-by' relation to then link to the type of organisation or authority that's in play. This would provide a good link to policy-aligned processes and actors in SDGIO (although we would want to avoid import madness).

classify the zone hierarchy; currently lacks logical defs so cannot be done automatically

Sure, but I think we need a bit more content to do this comprehensively. We can give it a first pass with what we have.

Come up with a strategy for ORs. We need to be able to classify with Elk. We can use the el-shunt pattern

Not familiar with the el-shunt pattern or what the strategy issue is - can you elaborate or link out?

consider candidates for moving to material entity branch: coast

Thought about that, this is very much treated as an area. We could re-label to "coastal zone"

iron-reducing zone of petroleum contaminated sediment

That's an area that is delimited by the activity of a process. If we want to push it to the ME branch, we could relabel to "(portion of) petroleum contaminated sediment undergoing iron-reduction", which is more accurate anyway. The existing label can be pushed to a rel. syn.

marine park

This is one of those potentially arbitrary space bags - would definitely keep this as a site.

rhizosphere (I feel strongly about this one; microbiome people will be confused if they cannot say microbes part of rhizosphere)

As a microbiome person, I can see the thinking here. My take on the issue was described above - we will then have to be bold and clean up the existing defs that are floating around. I don't mind this, as long as we add comments describing the commitment. We may have to do that with terms like "microbiome" (a semantic mess) anyway.

tectonic plate boundary (these have

Not sure about that one, this is again a site where plates meet. It's limits are not clearly defined by any difference in the plates themselves. I think this and others like it are more accurately left in the zone hierarchy.

anything under zone with "geographic feature" in the genus of the definition or anything that has an ENVO material entity as the genus of the definition. E.g. Ramsar site: "A wetland protected by..."

Keep in mind that not all textual defs have been aligned as we're still thinking about this. But a Ramsar site (and things like it) can be a space bag that's not anchored in material entities (as I'm learning now). Those should definitely remain in site as this can be a cross-check against what they actually overlap vs what their designations claim they overlap.

Some of the more physical geography stuff can likely be moved to ME branches.

text definitions should be aligned with logical definitions [see my blog preview] 'marine environmental zone' EquivalentTo ''environmental zone' and ('part of' some 'marine environment')'`: this is weaker than the text definition "An environmental zone which is bounded by material parts of a marine environment.".

Yes, as noted above, the textual defs need work after we stabilise our approach in general. What would be the right (stronger) axioms in the case above?

apply consistent DPs. E.g. ice accumulation zone is defined according to overlaps, its child glacial ice accumulation zone is defined according to a process. We should stick to one DP. (but note 'contains process' is too weak)

Agreed - Is there a systematic treatment of these relations or should we start a wiki page describing them? Are new ones needed? Overlaps was my go-to generalisation to fix the reasoning errors some time back, but I agree it's not terribly helpful as it doesn't give us bounding semantics.

Tackle low hanging fruit e.g. the two subclasses of littoral zone can easily be axiomatized by part-of to marine/freshwater two subclasses of coast

Sure.

Add DPs for vegetated area. I suggest a new shortcut relation 'primarily covered by'

What would the shortcut expand to? Coverage is an interesting thing - it suggests that the area is underneath the covering entity, which doesn't sound right (although that's probably how the satellites see it).

I think the relation we're aiming at between the site and the ME is something along the lines of:

• "site contains some concentration/density of ME" which can have related relations like:
• "site primarily overlaps/contains ME" (that is, most of the space in the site overlaps/contains the ME(s) in question)
• "site contains some minimal/threshold concentration/density of ME"

[rduerr]

@pbuttigieg coverage is indeed an interesting thing; because sometimes it means exactly what it says. For example, sea ice coverage is what fraction of the area discussed is covered by sea ice (as in sea ice on top of an area of sea water); ditto for snow coverage (i.e., snow on top of either bare ground or ice)....

[cmungall]

I tend to agree, but the defs that are used often refer to areas and regions, which are immaterial. It's likely these terms are used colloquially, so this is more a question of whether we want to take an ontologist's prerogative and say "What you really mean is...". This can be dangerous at times, but in the case of rhizosphere I think it's okay. Indeed, biosphere, cyrosphere, etc are all material systems

I think it is good to make changes to definitions to allow them to follow a genus-differentia form where the genus is a parent in the ontology - provided of course it does not change the intent of the domain expert. If we truly believe that these are zones then the genus should say zone. If the domain expert finds that funny then it suggests our placement under zone may be problematic.

[cmungall]

Apologies if any of the suggestions below are redundant with my first comment, I forgot I had already written a ticket about this:

Proposed naming conventions

• Every class under EZ should have either "area" or "zone" or "point" or "boundary" in the primary label to clearly signify to the user (who may be using a system that does not show all superclasses to root) that this is an IM
  • I would favor "zone" as consistently the primary label

Changes to comply:

• savanna - change to savanna area. Add "savanna" as exact synonym to cognate ecosystem/biome
• iron-reducing zone of petroleum contaminated sediment - should be under 'sediment'
• marine oligotrophic desert - Text def is "an expanse of marine water which...", which places it in under ME. If need be, make a "MOD zone" term that shadows MOD.

Axiomatization

General principle: all EZ classes should have logical defs, the logical defs should be trivial simple ones that typically leverage a cognate ME or P. There should be few or no "primary" EZs.

I propose the following patterns

1. area of X = EZ and occupies some composed-primarily-of some X (where X is a ME)
2. area designated for X = EZ and designated-from some X (where X is a P, and designated-from is a new shortcut relation that can bundle the semantics of human intention and law)
3. Others: X zone/area = EZ and occupies some X

(I need to check on the precise relation for occupies, perhaps it is site-of. The important characteristic is that it is functional and inverse functional (1:1), and maps an IE to a ME)

Antipattern: use of overlaps in a logical definition, as this is usually more inclusive than intended.

Examples of antipattern, the problem, and how to fix:

• ice accumulation zone = 'material accumulation zone' and (overlaps some 'ice mass'). Counterexample: a water accumulation zone that includes melting ice would be incorrectly classified here. Note also deviation from text def. Fix: axiomatize using a P
• terrestrial EZ = 'environmental zone' and ('part of' some land). Note deviation from TD An environmental zone which is bounded by material parts of a land mass or the atmosphere or space adjacent to it. Cryptic unsatisfiability: sea grass bed is classified here, but this is inconsistent with part of some land.
• saline wetland = overlaps some (land and ('composed primarily of' some (soil and ('has part' some 'saline water')))). Note the def has no genus, see my blog post on definitions. Note that according to this definition, the universe is a saline wetland.
• forested area ='woodland area' and (overlaps some 'forest ecosystem')`. Note the deviation from the TD, itself an anti-pattern. Note that by this definition, forested areas are viral. It is impossible to have any non-forested woodland area adjacent to a forested area, as the whole thing becomes forested.
• bush area = 'vegetated area' and (overlaps some ('forested area' or 'scrubland area' or 'area of scrub' or 'area of developed open space' or 'area of sedge- and forb-dominated herbaceous vegetation' or 'area of gramanoid or herbaceous vegetation')). Again, bush areas are viral. Also, use of long chains of OR are an antipattern in my book.

TBD: use of has-quality, as in desert area = 'area of barren land' and ('has quality' some arid). Note in BFO, IEs cannot have qualities. I think this is too extreme as clearly spaces have qualities such as volume. But I partially agree in that it is odd for an IE to have a quality such as arid. I think this is symptomatic of over-use of IMs, and the problem is fixed by making desert primary, and having desert area shadow it with a 1:1 relation.

Text defs

Should be the basis for logical defs, and in general should shadow ME/P in a predictable way

• crater zone = The site which is located in a crater. Use of "the" is a bit of an anti-pattern, there is no one site in a crater, which one? Is this the maximal one?

Review

• Move these out from EZ into the ME hierarchy:
  • farm and its only child ranch

Wetlands: the EZ hierarchy has some potentially useful classes like "intermittent wetland", "saline wetland", "peatland" (and subclasses), if only they were under ME. The textual definition of these classes strongly suggest ME. I suggest either moving en-masse, or making cognate classes linked 1:1 (but strongly prefer the move option, keep it simple).