Identification Matrix

#1

The latest Audubon app has an amazingly functional bird identification module. If iNat devs haven’t examined it, they should right away.

It basically asks for a set of diagnostics, but is laid out in a way that is easy to choose the relevant ones, rather than work through a linear or dichotomous workflow. It uses location and time to automatically narrow the band, and then you can select size, color, type, activity, habitat, voice, wing shape, and tail shape. Each selection narrows the search until you get pretty close. I’ve used it a few times to get to species with good success, which is more than I can say for my success rate on plants with dichotomous keys, which often dead end at questions I can’t answer.

I hesitate to suggest this, because it entails an enormous effort. They have the advantage of being limited to a small branch of the taxonomic tree, but I think we could crowdsource a tool for a broader range of taxa by assigning characteristic categories to help people narrow in on finer taxa, and creating GUI to filter by those categories.

I wanted to at least propose it and see if it sticks.

0 Likes

Photo taking recommendations by taxon?
#2

There is a similar resource for plants in new England here:

https://gobotany.newenglandwild.org/full/

my guess is this is far beyond the scope of iNat, but it is a neat idea.

1 Like

#3

It is a bit different with birds. There are finite species, they are easily identified (in that no microscope or key work is required). You couldn’t do this with plants very easily as there are several thousands of species per area, and finding “exceptions” (species that are not listed, or aberrant that don’t match all the key features) is actually quite common. This is even worse for insects, with exception of some prominent groups (e.g. butterflies), where there isn’t even enough distribution data to know what occurs where and when, even in very common species.

Right now I can’t imagine it being feasible. The iNaturalist “auto-ID” is as close as we can get for the time being.

4 Likes

#4

While I suspect that silversea_starsong is right for plants, I am totally with you! I had a bit of a rant about it the other day - bottom line is I think it’s possible for comparatively small subgroups but yes, it would require a lot of work.

0 Likes

#5

The technical term is “multi-access key”. LSU has one for Carex (sedges) here: http://www.herbarium.lsu.edu/keys/carex/carex.html which is pretty impressive.

You can generate them with software packages like Lucid (the full version of which is unfortunately out of reach of my amateur-taxonomist budget).

Keys in iNat would be amazing, but that would circle back to some of the sticky issues we’ve discussed in the past, including how we delimit taxa, who has the final say, and then who maintains the keys on an ongoing basis.

3 Likes

#6

There is also Sliks/SAIKS which has been used by many, but not maintained by any, that I can tell. I do think it could help to narrow down choices to reasonable ones, but it would require that the people who own/produce the underlying data also release it to the public domain. Even then, though, in places like S. California, one can run across nearly any species of plant or insect, so things like location as a filter need to be handled carefully.
http://stingersplace.com/SLIKS/

2 Likes

#7

I tried gobotany several times and had about the same luck as I’ve had with dichotomous keys in books. Part of the disconnect, I think, is in the way botanists (or any other specialist) looks at their subject, versus how an amateur looks at it. Most keys seem to think, what is the platonic ideal of the plant and how can we describe it, rather than what does this actual specimen give you to work with. That is one of the beautiful things in the Audubon categories: It’s all in plain english, and works with attributes that are observable on the fly.

I think it does fall within the scope of iNat, though I agree that it would ideally be a separate open-source database that any service could use. Often, image recognition can get you so fine a taxon suggestion, and even then it is best to learn about the distinctive attributes of the species before you agree. A multi-access key (thanks @choess!) would be a great educational tool and identification guide.

Maybe this already exists. I couldn’t make much of the fairly technical SLIKS/SAIKS or LSU Herbarium links, and Lucid also looks pretty user-unfriendly.

Here is the vision:
Database for everything on the tree of life, with useful attribute tags to describe the distinctive features of every species.

Woah. Every species?

I’m sure this could be worked on gradually, and rolled out incrementally. Entries for every species could be generated the same way that they are for iNat. The work would be volunteer-based and crowdsourced online. Specialists of different stripes could contribute to the development of their branches of interests. As branches are filled out, they could be marked as ready for use, and suddenly if you get to that taxon, you have a set of attribute questions you can answer to narrow down to species. This is how we all use books, but we scan for attributes in the narrative-form species descriptions, rather than categorizing by attribute tags.

I’m proposing it rather than just doing it because:

  1. I don’t know it there are any other projects working on this kind of thing.
  2. I’m not a developer, and don’t know how to get the database started, or best practices on how to keep it organized.
  3. I’d like to see if there is any interest on the forum I know about on the internet most likely to geek out on this kind of project.
  4. I think there would be a great integration between this kind of database and iNat.

There are several genera that, if I had the technical know-how, I’d love to work up this kind of key.

4 Likes

#8

Multi-access keys have been around for a long time, I use them often to identify my plants. To make something truly great it should be done mostly automated while showing the steps it takes to achieve the ID, such as viewing the thousands of observations of fruit from a Eucalyptus species to eliminate species in the list, then to bark etc. If the automated system become wrong, then we can still have control over what not to do.
I’ve always uploaded multiple photos in an observation regarding Flora, it would have been appropriate to add metadata such as tags to those ‘individual photos’ for use of multi-access key or future automated ID tools. How can software recognize a Eucalyptus if it has no reference to the fruit? In fact, no one can search for Eucalyptus fruit on iNat. I still must use Google to do this.

“with useful attribute tags to describe the distinctive features of every species.” Exactly!!!

0 Likes