trait list: hierarchical levels

[fdschneider]

Do we want to have more hierarchical levels than the two levels "class and group"?

[fdschneider]

For practical reasons, I would keep the two-level hierarchy. I think the tree-hierarchy used by TOP and SITA is useful, because it is not limited in depth. But that makes it less feasible managing the trait lists on BExIS. The two-level hierarchy proposed by Moretti et al is quite simple and fairly complete, although they do not as clearly differentiate response traits and effect traits.

For our trait-datasets, it is almost irrelevant which hierarchy is behind the trait names and IDs.

I am inclined to use a system of traitIDs similar to the referential taxonID that was used on the BExIS species list: A character string combining the reference list with the numerical ID, e.g. "bexis:traits:invertebrates:42". With this, we can map our traits on a more complex trait ontology later, if one becomes available.

[nadjasimons]

I also agree that a hierarchy with a fixed number of levels would be nice. I managed to summarize the SITA hierarchy to 6 levels: Category Group Type Trait Strategy Attribute But the first two levels are not really needed. In order to make the definition of traits clearer, more than one level is often good. E.g. Voltinism can refer to both the phenology (i.e. in which months are adults dispersing) as well as development (i.e. which stage is the dispersing stage). I can create a new column in the trait list with a suggestion for a two-level trait name and then we can see if all traits are clear.

[fdschneider]

18 is the thread that takes this discussion further.

We provide a starting point by providing a terminology for trait lists (see the traitdata standard draft) . The 'linkability' to corresponding, higher or lower terms would enable researchers to build a decentralised semantic network of trait definitions.

I think this is more sustainable than providing a fixed hierarchy of terms.