suggested axiom changes to ice hierarchy

[cmungall]

Continuing ice discussions from #900 #782

Current ice hierarchy:

remove water-body-derived ice

This has a single child, sea ice.

Single children are a bad smell. If they exist, there should at least be counter-examples provided. What are non-water-body-drived ices, and what are non-sea-ice water-body-derived ices

water-body-derived ice has a comment:

In the United States, NOAA considers ice formed from the freezing of the waters of the Great Lakes as being the same as sea ice. This class attempts to provide a distinction for clarity.

I'm not sure how this grouping class helps?

If the class is to be retained I would suggest making a logical def as it looks compositional. But I don't understand why it's here and what use cases it supports

add logical definitions following a compositional pattern

in general X ice = ice and composed-primarily-of some X

remove equivalence axiom on blue ice

Text defs and logical defs should match

text: Water ice containing little to no air bubbles resulting in a reduction of internal light scattering and, given sufficient volume, a blue appearance. [https://en.wikipedia.org/wiki/Blue_ice_(glacial)]

logical: ice and ('has quality' some blue)

These don't match, and there are materials that fit the logical def that do not fit the text def

E.g.

I suggest removing the logical def. I also suggest changing the primary label or a synonym to reflect the meaning

Minor: the text def veers quote a bit from the wikipedia def even though the wikipedia page is given as a source.

clarify ice vs water ice

I understand the need to separate these to support some astro-geo use cases. However, many users will be expecting these to be equivalent. We should at least add an rdfs:comment to ice suggesting that users may want to look to the subclass 'water ice'

We may also want to look where ice is used and use a subclass and see if we in fact mean water ice:

I assume this is intended to be water ice?

or are there glaciers on other planets made from methane, ammonia, etc? If so we really have to consider the complexity of using envo for multi-planet use cases

ice surface

we have both

and

It's confusing to have these shadow classes. The 2D shadows were presumably added to correspond to sweet: http://cor.esipfed.org/ont?iri=http://sweetontology.net/realmCryo/IceSurface --- but it's not clear this is truly the intended semantics of sweet. sweet isn't built on bfo, and it's not clear why the sweet class wasn't mapped to the more intuitive ice surface in envo rather than the 2D immaterial entity

where we do populate shadows, and these are compositional, we should have equivalence axioms to make the correspondence clear. currently these are just subclassof

note: currently overlaps is being used. it's not clear this is correct, and a 2D immaterial surface overlaps with the 3D material entity it is the boundary of. In any case, it seems a boundary relation is more useful here

[cmungall]

Quick protege tip - subClassOfs can be grouped and turned into equivalents:

[kaiiam]

remove water-body-derived ice

@pbuttigieg is this class intended to differentiate sea-ice from freshwater ice? Or water-body-derive ice from ice formed on land?

add logical definitions following a compositional pattern

👍

remove equivalence axiom on blue ice

I love the blue ice-cream counter example! +1 for removing the logical def. As for the label/synonym reflecting the meaning, I think based on conversations with my professor, Dr. Hong Cui from the UA information school that it's better to use the labels as they exist in the community rather then something new or varying to much from what the domain scientists expect to see or already use (a mistake library scientists made for a long time according to her). But we can ask @rduerr if there are other blue ice synonyms or if she has other suggestions.

clarify ice vs water ice

Totally agree on all points. Depending on @pbuttigieg's feedback I can make another PR and try to fix the incorrect uses of ice intended to be water ice. Regarding the multi-planetary use case I think it's good to leave that door open, but not let it get in the way of representing things on here on earth (our primary use case).

ice surface

I agree that the duplication is confusing, especially since these fiat ice surface classes are the only thing we have under continuant fiat boundary. Since we don't know the extent to which SWEET actually maps to BFO, perhaps we change the mapping of http://sweetontology.net/realmCryo/IceSurface from two-dimensional fiat water ice surface to ice surface layer?

Thanks @cmungall for the tip on grouping subclasses to equivalence, we have many classes with multiple of subclass axioms e.g.

If I understand correctly merging these into a single class expression with an equivalence axiom will give us a lot more from reasoning/inference. Should we make this a standard practice?

[rduerr]

OK weighing in here; I wasn't involved in all of these definitions; but did notice based on this ticket that yes, there are some screwy things in the current hierarchy, though they aren't quite the same things as @cmungall is talking about, though I will try to address at least some of those comments as well.

My first complaint is that the terms under sea ice floe (nilas, pancake ice, young ice) are not necessarily sea ice floes... Two of of those terms (frazil & shuga) plus all of the New ice terms and several of the terms under sea ice (first year ice, multiyear ice, second year ice) are actually terms describing the stages in the life of sea ice which starts off as either frazil or shuga which then progresses to new ice and eventually first-year, etc. ice. Pancake ice is also not a floe it is an ice form. In general the Earth sea ice community has standardized terminology along 3 dimensions - concentration, form and stage (which is a proxy for age). Yes, there are other terms in use; but those are the major characteristics of Earth's sea water ice (even if NOAA uses some of the same terminology for fresh water ice) and which show up everywhere in the literature and on sea ice maps used by shipping companies, etc. I suggest that concentration, form, and stage are qualities that sea ice has?

Second there are glaciers on both Mars and Pluto. I would say that glacier needs to come out from under water ice since Pluto has glaciers made of nitrogen.

Continuing ice discussions from #900 #782

This has a single child, sea ice.

Single children are a bad smell. If they exist, there should at least be counter-examples provided. What are non-water-body-drived ices, and what are non-sea-ice water-body-derived ices

water-body-derived ice has a comment:

In the United States, NOAA considers ice formed from the freezing of the waters of the Great Lakes as being the same as sea ice. This class attempts to provide a distinction for clarity.

I'm not sure how this grouping class helps?

I tend to agree - the issue is that NOAA uses the same terms for young and new ice for Great Lakes ice as they do for sea ice. However, the ice on the great lakes melts out completely each summer so the terms for late stages of the life of ice (e.g., first year, second year, multiyear) don't apply. Personally, I think we should just make the Great Lakes a special subset that parallels the stages of sea ice but doesn't include terms that don't apply. As far as I know other countries do not do this.

remove equivalence axiom on blue ice

Text defs and logical defs should match

text: Water ice containing little to no air bubbles resulting in a reduction of internal light scattering and, given sufficient volume, a blue appearance. [https://en.wikipedia.org/wiki/Blue_ice_(glacial)]

logical: ice and ('has quality' some blue)

These don't match, and there are materials that fit the logical def that do not fit the text def

Agreed - the distinguishing feature is the lack of air bubbles which results in a blue color; so a better logical definition would be something like:

water ice and ('has quality' some 'absence of air bubbles') or something like that...

I don't know how one would derive the color blue ontologically without adding a whole bunch of physics about light scattering and volume of ice which is the reason for the blue color. As for why it is important, it is because when that color is observed (in data or by seeing it) you can then automatically derive other characteristics (i.e., pure - no salt, drinkable, old, no bubbles, etc.). These qualities are meaningful not just to science but also to people (especially up north) who live there. Any suggestions for how to fix this?

Minor: the text def veers quote a bit from the wikipedia def even though the wikipedia page is given as a source.

clarify ice vs water ice

I understand the need to separate these to support some astro-geo use cases. However, many users will be expecting these to be equivalent. We should at least add an rdfs:comment to ice suggesting that users may want to look to the subclass 'water ice'

We may also want to look where ice is used and use a subclass and see if we in fact mean water ice:

Definitely! As I think there are a lot of places where that distinction needs to be made...

I assume this is intended to be water ice?

For Earth (and Mars) glaciers yes it is water ice.

or are there glaciers on other planets made from methane, ammonia, etc? If so we really have to consider the complexity

for Pluto's glaciers - no, the glaciers are Nitrogen ice

remove water-body-derived ice This has a single child, sea ice.

Single children are a bad smell. If they exist, there should at least be counter-examples provided. What are non-water-body-drived ices, and what are non-sea-ice water-body-derived ices

non-water-body-derived ices:

• Methane clathrate ices here on Earth
• Nitrogen ice on Pluto (water ice on Pluto acts like rock here on Earth)
• CO2 ice on Mars (as well as water ice)

non-sea-ice water-body-derived ices:

• river ice
• lake ice
• ice cubes from municipal water supply (?)

water-body-derived ice has a comment:

ice surface

we have both

and

It's confusing to have these shadow classes. The 2D shadows were presumably added to correspond to sweet: http://cor.esipfed.org/ont?iri=http://sweetontology.net/realmCryo/IceSurface --- but it's not clear this is truly the intended semantics of sweet. sweet isn't built on bfo, and it's not clear why the sweet class wasn't mapped to the more intuitive ice surface in envo rather than the 2D immaterial entity

where we do populate shadows, and these are compositional, we should have equivalence axioms to make the correspondence clear. currently these are just subclassof

note: currently overlaps is being used. it's not clear this is correct, and a 2D immaterial surface overlaps with the 3D material entity it is the boundary of. In any case, it seems a boundary relation is more useful here

makes sense to me....

[cmungall]

@kaiiam :

If I understand correctly merging these into a single class expression with an equivalence axiom will give us a lot more from reasoning/inference. Should we make this a standard practice?

No! My tip is only intended to be used when you truly want to make an equivalence axiom! It's certainly true you will get a lot more from reasoning if you blindly convert all subclass axioms to equivalence axioms, but most will be wrong! See David's comment in https://github.com/EnvironmentOntology/envo/issues/900#issuecomment-723885445

equivalence axioms should only be added if you are completely sure the expression on the left side is equivalent to the expression on the right side. If you can think of an example of the expression on one side that is not an example of the expression on the other side, it should not be an equivalence axioms!

I think we need a lot more docs about this in the ENVO docs as I have seen repeated issues here.

If in doubt, only make EC axioms if they conform to a validated design pattern template, and even then if you are sure!

[cmungall]

@kaiiam - other than the EC issue, I agree with all your suggestions - I think a PR would be great if you can do that!

[cmungall]

blue ice:

Thanks for your comments @rduerr - I think based on what you and @kaiiam we should simply remove the logical definition

re:

a better logical definition would be something like:

remember, it's not so much whether it's better, it's not a matter of degree - it's either true or false. Better to say nothing than something false!

I don't know how one would derive the color blue ontologically without adding a whole bunch of physics about light scattering and volume of ice which is the reason for the blue color. As for why it is important, it is because when that color is observed (in data or by seeing it) you can then automatically derive other characteristics (i.e., pure - no salt, drinkable, old, no bubbles, etc.). These qualities are meaningful not just to science but also to people (especially up north) who live there. Any suggestions for how to fix this?

Yep - just drop the logical definition! Some things are fine with just a clear text operational definition that is intended for humans!

[cmungall]

glacier composition:

Thank you for your comments @rduerr, these are really useful! I did not know about nitrogen glaciers on pluto! I worry about the complexity introduced to ENVO for the multi-planet use case. Many users will choose the term glacier thinking it means water ice glacier.

But for now it seems that (generic) "ice" is the appropriate level of granularity for (generic) "glacier"

We may want to also think about simple compositional patterns such as "X glacier = glacier and composed-primarily of X" which will help make it more explicit (at the expense of some complexity via latticeyness).

What about the axiom: glacier subClassOf has-part some snow. Is this true of all glaciers? Do glaciers on pluto have snow? I suspect not.

In fact I suspect even earth water ice glaciers do not have snow as a part, as you say in a separate ticket the snow lies on top.

[kaiiam]

Action items for a PR: (comment edited Nov 18th):

1. Change equivalence axiom on blue ice, to be a subclass axiom and use 'water ice'.

2. Identify and fix possibly incorrect uses of ice intended to be water ice (perhaps glacier is ok to use ice).

3. Move dbxref of http://sweetontology.net/realmCryo/IceSurface from two-dimensional fiat water ice surface to ice surface layer @cmungall and @kaiiam +1, @pbuttigieg -1?, also possibly obsolete two-dimensional fiat water ice surface?

4. Move water-body-derived ice comment: In the United States, NOAA considers ... to sea ice.

Questions to be sorted out:

1. Remove water-body-derived ice? -> probably not there are counterexamples like terrestrial ice, lake ice these just haven't been added yet.

2. Sort out sea ice floe subclasses?

3. Discuss classes for ice concentration, form, and stage.

4. Discuss moving glacier away from water ice, having part snow, and or having a compositional pattern. -> looking like yes.

[rduerr]

We may want to also think about simple compositional patterns such as "X glacier = glacier and composed-primarily of X" which will help make it more explicit (at the expense of some complexity via latticeyness).

I agree

What about the axiom: glacier subClassOf has-part some snow. Is this true of all glaciers? Do glaciers on pluto have snow? I suspect not.

I suspect all glaciers can have snow of type X (e.g., Nitrogen snow on Pluto) but it isn't a part, it is something that would lie on top.

In fact I suspect even earth water ice glaciers do not have snow as a part, as you say in a separate ticket the snow lies on top.

Agreed

[rduerr]

@kaiiam maybe all of those questions can be discussed at the next semantic harmonization telecon?

[cmungall]

What about the axiom: glacier subClassOf has-part some snow. Is this true of all glaciers? Do glaciers on pluto have snow? I suspect not.

I suspect all glaciers can have snow of type X (e.g., Nitrogen snow on Pluto) but it isn't a part, it is something that would lie on top

In fact I suspect even earth water ice glaciers do not have snow as a part, as you say in a separate ticket the snow lies on top.

So it looks like the current axiom is wrong twice: in the relation (parthood vs some kind of 'on top') relation, and existentially (A subClassOf R some Y: all instances of A stand in relation R to some Y)

[kaiiam]

@rduerr yes lets bring this to the floor in our next meeting. @cmungall if you have time to join us, it this Wednesday, November 18 2:00 – 4:00pm ET at this meeting link, if not no problem I will try to raise the issues you brought up.

So it looks like the current axiom is wrong twice: in the relation (parthood vs some kind of 'on top') relation, and existentially (A subClassOf R some Y: all instances of A stand in relation R to some Y)

Recap in English (correct me if I'm wrong) the relation of glacier having part snow is wrong 1) because the snow is on top of, not a part of the glacier itself, 2) We can't say that all glaciers have part some snow, (because some don't).

[cmungall]

Recap in English (correct me if I'm wrong) the relation of glacier having part snow is wrong 1) because the snow is on top of, not a part of the glacier itself, 2) We can't say that all glaciers have part some snow, (because some don't).

2: We can't say that all glaciers have part some snow OR have some snow on top, (because some don't).

[pbuttigieg]

Thanks @cmungall

remove water-body-derived ice

This has a single child, sea ice.

Single children are a bad smell. If they exist, there should at least be counter-examples provided. What are non-water-body-drived ices, and what are non-sea-ice water-body-derived ices

Sometimes, these are there as the working group didn't have time / scope to add more. We can have, for example, 1) lake ice, 2) pond ice, etc

There are also terrestrial ices which @mpsaloha can elaborate on.

water-body-derived ice has a comment:

In the United States, NOAA considers ice formed from the freezing of the waters of the Great Lakes as being the same as sea ice. This class attempts to provide a distinction for clarity.

I'm not sure how this grouping class helps?

Quite right, that comment belongs on "sea ice"

add logical definitions following a compositional pattern

in general X ice = ice and composed-primarily-of some X

Yes, sounds reasonable for things like "methane ice" etc (not for things like sea ice, which would use "formed as a result of" patterns, I think).

Text defs and logical defs should match

text: Water ice containing little to no air bubbles resulting in a reduction of internal light scattering and, given sufficient volume, a blue appearance. [https://en.wikipedia.org/wiki/Blue_ice_(glacial)]

logical: ice and ('has quality' some blue)

These don't match, and there are materials that fit the logical def that do not fit the text def

Indeed - will move it to a subclass axiom and use 'water ice'

.. I also suggest changing the primary label or a synonym to reflect the meaning

A synonym could be added, but "blue ice" is a standard term.

clarify ice vs water ice

I understand the need to separate these to support some astro-geo use cases. However, many users will be expecting these to be equivalent. We should at least add an rdfs:comment to ice suggesting that users may want to look to the subclass 'water ice'

We may also want to look where ice is used and use a subclass and see if we in fact mean water ice:

I assume this is intended to be water ice?

or are there glaciers on other planets made from methane, ammonia, etc? If so we really have to consider the complexity of using envo for multi-planet use cases

This is a broader issue that we should examine in a dedicated issue, I think - ENVO is going interplanetary, but most users will want the Earth-bound meanings attached to these labels. Modularisation may be the way to resolve this (envoEarth.owl, envoTitan.owl...), but that's a bit to instancey. We'd need a way to handle a likely explosion of terms ([methane,carbon,...]-ice glacier) that would make things pretty tricky to use without subsets or interfaces that do the subsetting for specific communities. I think we need those anyway, as the ontology needs to satisfy many user groups with different needs.

[pbuttigieg]

On the fiat boundaries...

It's confusing to have these shadow classes. The 2D shadows were presumably added to correspond to sweet: http://cor.esipfed.org/ont?iri=http://sweetontology.net/realmCryo/IceSurface --- but it's not clear this is truly the intended semantics of sweet. sweet isn't built on bfo, and it's not clear why the sweet class wasn't mapped to the more intuitive ice surface in envo rather than the 2D immaterial entity

The SWEET class traces back to a geometry class, so it's more in line with 2D fiat entities.

However, this definition needs correction, as it references material composition of an immaterial entity.

note: currently overlaps is being used. it's not clear this is correct, and a 2D immaterial surface overlaps with the 3D material entity it is the boundary of. In any case, it seems a boundary relation is more useful here

Noted.

[cmungall]

The SWEET class traces back to a geometry class, so it's more in line with 2D fiat entities.

I think this is reading too much into SWEET. It's a thesaurus that was built with loose use of subclass and no ontological commitment to BFO.

according to SWEET an ice surface is both a planetary realm and a numerical object:

This doesn't make any sense from a BFO perspective, but it's totally fine if you take the view that the hierarchy is just there to help people find relevant concepts, like MESH, or a organization of items in a supermarket.

In SWEET, hurricane is (1) a phenomena (2) a mathematical process and (3) a property

A marine protected area is (1) a chemical substance (2) a numerical entity and (3) a human activity

(note: all bold lines in these diagrams are SubClassOf)

This is incoherent ontologically, but that's fine, SWEET isn't intended as an ontology, just a loose hierarchical organization of concepts.

Which is to say that I'm not sure being able to trace back to specific upper levels in SWEET is justification for a particular BFO interpretation, or justification for introducing complexity to ENVO

Going back to @kaiiam's question:

Since we don't know the extent to which SWEET actually maps to BFO, perhaps we change the mapping of http://sweetontology.net/realmCryo/IceSurface from two-dimensional fiat water ice surface to ice surface layer?

I think we do know the extent to which SWEET maps to BFO: zero. But yes, the suggestion still stands, map to the more useful and intuitive " ce surface layer" in ENVO and obsolete the confusing 2D classes.