rduerr
commented
3 years ago OK, now this is confusing. The parent term "ice field" has the definition "A field of ice, formed in regions of perennial frost." with a cross-reference to ADL:FTT - a resource that is totally Greek to me (no link so hard to figure out where to look). However, I note the term "perennial" which in geography means continuously over the year (or longer even). That leads me to think that the term really only applies to the glacier ice field definition... since sea ice fields may not last year round.
I suggest eliminating that parent object, since the use in glaciology is so different from the use in sea ice research. In that case the definition of glacier ice field could become "A cryoform that 1) has a roughly level surface, 2) consists of glacier ice with possibly snow, 3) fills a mountain basin or covers a low-relief mountain plateau, and 4) with flow at least partially controlled by terrain."
Yes, that definition is much more complicated than the original but gets across all of the qualities of a glacial ice sheet that I am aware of (synthesized from the 8 definitions that the GCW had). The only concern I have is about the terms snow and glacier ice since snow currently is restricted to water ice as is glacier ice. This works fine for glaciers on Earth but I am sure it won't work for glaciers on Mars or Pluto and I note that other planets have other kinds of snow (e.g., some weird planet I heard about this week has lead snow due to high temperature and pressure - can you imagine flakes of lead falling out of the sky?).
About the 4th clause above - the point is that ice fields are distinguished from ice caps by size (ice caps are bigger usually) and by the factor that ice caps tend to flow radially down from the top of a plateau or mountain whereas ice fields may not flow at all if they are in a basin of some type.
the subclass axioms could then be something like:
( 'composed primarily of' some ('glacial ice' or snow)) and (participates_in some glacial transport process)
or you could do the equivalence class thing and make it a cryoform and .... stuff from above
the text def says AND but this axiom says OR
During summer there may be no snow (all melted or turned to ice) and the original definition was talking about the accumulation process which would involve snow but the composition doesn't require it.
is snow part of a glacier or is there some other relation? see discussion below
Snow isn't part of a glacier - it lies on top of the glacier sometimes....
(partially_surrounded_by some ('structural basin' and ('part of' some mountain))) or ('adjacent to' some (plateau and ('part of' some mountain))) this is quite a complicated disjunction that ultimately doesn't buy us much.
I don't know how to add the roughly level surface bit (first clause) or for that matter the filling or covering bits (there doesn't seem to be "on top of", a "covers", or "fills" relationships anywhere).
As for the other definition:
http://purl.obolibrary.org/obo/ENVO_01001531 sea ice field
An ice field which is primarily composoed [typo] of sea ice floes greater than 10 kilometers across.
SubClassOf:
'ice field' 'composed primarily of' some 'sea ice' -- this could become the differentia in a logical definition
yes it could if you kept the current parent class and changed its definition slightly to get rid of the perennial bit
'has member' min 1 'ice floe' -- why introduce a new relation here? -- maybe this is because you need a cardinality constraint and this can't be done with transitive properties; but "min 1" is an anti-pattern, you can just say some, it means exactly the same thing, and it falls within the EL profile, is simpler, and can be used with transitive properties
OK, I am fine with 'has member' some 'giant ice floe' - A giant ice floe is by definition greater than 10 km across whereas an ice floe could be as small as 20 m across... but ENVO doesn't have numeric size limits as part of its axioms, even when they are part of the definition of terms, so no one has added all the terms for different sized of ice floe AND as noted in #1034 none of the subclasses of the current sea ice floe term are actually subclasses!
cmungall
pbuttigieg
looking at the two children of 'ice field', differentia glacier vs sea. My comments in bold.
http://purl.obolibrary.org/obo/ENVO_01001532 glacier ice field
An ice field which is constituted by a continuous accumulation of snow and glacier ice that completely fills a mountain basin or covers a low-relief mountain plateau.
SubClassOf:
http://purl.obolibrary.org/obo/ENVO_01001531 sea ice field
An ice field which is primarily composoed [typo] of sea ice floes greater than 10 kilometers across.
SubClassOf:
general comments
I always recommend trying to simplify, avoid constructs like OR, QCRs, nesting.
In other ontologies we often stick to a simple EL profile, with just AND, SubClassOf, EquivalentTo, and SOME, with a maximum nesting level of 1. This is easier for humans to reason over as well as machines, easier to visualize, and can make the ontology more maintainable by having more consistent predictable patterns here.
Some of the axioms here seem like 'rococo' axioms (see my talk: https://douroucouli.wordpress.com/2020/11/02/aligning-design-patterns-across-multiple-ontologies-in-the-life-sciences/). They don't buy us much. We don't get much inference from them, and if they were to go, would a user care?
Not everything in a textual definition needs to be captured logically. Sometimes less is more, and it is better to capture less in a simpler fashion than try and capture everything.
I also advocate for sticking to simple consistent design patterns. I think the 'composed primarily of' differentia pattern is very useful in ENVO.
E.g. "X ice field" = "ice field and composed-primarily-of some X ice"
Adding equivalence axioms is very powerful and buys you a lot in reasoning, both for classification, and finding inconsistencies and incoherencies.
use of part-of between ice/snow and glaciers
The axioms above include a. clause that introces models (in the OWL sense) where mountains can have snow as part. I am copying a discussion from slack that is relevant:
schild 11:08 PM simple question I hope: what is the best pattern fordescribing some feature of some place. Like “glacier on the Matterhorn”. I’d expect using a GAZ ID or similar to refer to the Matterhorn, and EnvO has a term for “glacier’. But describing the combo: “Matterhorn glacier”? Thanks! (edited)
schild 12:19 PM @channel wondering if anyone can answer my question from last night…(just above)
Damion Dooley 1:08 PM It would be a structure, like “glacier and ‘located in’ some [Matterhorn mountain GAZ ID]” (edited)
cmungall 1:08 PM part of?
Damion Dooley 1:09 PM heh! 1:11 yes… that takes my mind into question of whether glacier is intrinsically part of the mountain, or is extrinsic to it. Calls for a position on mutable glaciers.
cmungall 1:17 PM I'm not sure concepts like intrinsic or extrinisic help so much. I think we should just look to the definition of mountain, which will hopefully tell us about composition. Is it just the rock component, or does it also include water bodies, etc. Ideally we have explicit has-part axioms, and partonomic disjointness.
cmungall 1:20 PM Note that if the definition of mountain is to restrict the composition to be the rock parts, then located_in is likely wrong
cmungall Today at 1:20 PM Note that if the definition of mountain is to restrict the composition to be the rock parts, then located_in is likely wrong
11 replies
cmungall 32 minutes ago the definition of SABP suggests that the relationship would be neither part-of nor located-in
Damion Dooley 31 minutes ago yeah. more like “adjacent to”
cmungall 30 minutes ago I think we'd want to be stronger than that, and have something that is less alien to domain scientists
cmungall 27 minutes ago Note that we have some classes right now such as glacier ice field that have complex logical axioms with unions of adjacent_to and partially_surrounded_by
Damion Dooley 27 minutes ago Where is John Muir when you need him! A poetic “cradled by” would do the trick! :smile: 1
cmungall 21 minutes ago I think a simple on-top-of relation between features would be best here. But I would defer to geographers use cases here.
Damion Dooley 21 minutes ago Just a glance at glacier geography. Speaks to a relation that packs result of process in it: accumulation - “accumulated on”
cmungall 18 minutes ago There are already too many relations used in ENVO SABPs. I would simplify and go for fewer relations, not try and pack everything into a relation. A spatial relation does not need to encompass causation. We can make separate processes such as 'glacier formation' which will clearly articulate the processual/historic/causal aspects.
cmungall 17 minutes ago When I look at the SABP hierarchy I see many complex rococo axioms that look impressive but don't do much work, and make the ontology harder to maintain
Damion Dooley 15 minutes ago yes, I agree. I like “on top of” or “on surface of” as more general solution.
Damion Dooley 13 minutes ago Always interested to be challenged on a friday afternoon (distracted!) by basic grammatical equivalency needs in ontology. We will catch up to grammar soon!