Paper discussion: Recognition and completeness: two key metrics for judging the utility of citizen science data

I just published a paper (with fellow iNat users coreytcallaghan, wcornwell, and sgorta) on the usability of iNat data for Australian invertebrates.

In a nutshell, we analysed over 1 million Australian observations of terrestrial invertebrates (observations up to December 2021) and defined and calculated two metrics: recognition and completeness. Recognition is the % of records identified to species (not directly equivalent to ‘identifiability’ as we define it), and completeness is the % of known species in a taxon for a given area that have been uploaded/recorded. We calculated these metrics for 39 different ‘iconic’ taxa (butterflies, ants, spiders etc), and then categorised them into four groups, which then informed a framework that can advise which taxa have ‘usable’/robust data for conservation and research either now or in the future, and which taxa will likely never have robust data due to eg difficulties identifying species from photos. We also did the same analysis for Taiwan and the Netherlands as a comparison.

Paper is open access and available here:

Our main figure looks like this:

I’m keen to hear what others think about this framework we developed, and especially which taxa you think would have a ‘universal’ position on the axes no matter where you go in the world (eg most likely butterflies and Odonata), and also which may differ compared to where you’re from (eg well-known and recorded groups that perform poorly for Australia).


Completely fascinating. Thank you for sharing that! From my quick skimming of the paper, I’m taking away two conclusions for my own iNaturalizing (especially if I ever get to Australia!): make more observations of the less obvious invertebrates; and help develop observe and identifier expertise in those organisms.

I’ll add one other point, based on my decade or more of using nymphs and exuviae to assess the status of state-listed and uncommon Odonata in Massachusetts in the US. While adult Odonata are indeed fairly easy to locate and photograph sufficiently for identification at the species level, it was only when we here in Massachusetts started focusing on nymphs and exuviae of Anisoptera (dragonflies proper, not damselflies) that a more complete picture of the status of some species become obvious. In short, adults of some Anisoptera can be difficult to catch (they are flying in the middle of big rivers, for example), but their exuviae are much easier to find and ID (the exuviae don’t fly away, if nothing else!). I’ve posted a few exuviae on iNat, but rarely has anyone IDed them, so one goal I might set myself is to disseminate information on how to ID exuviae (and maybe advocate for an annotation for them). Note that this side ramble of mine says very little, if anything, about the strength of your paper’s main points, by the way.


When you say “identified to species level” do you mean “identified correctly to species level”? Or was a species-level ID sufficient?


Amazing! I’ve been wanting to see an analysis like this. Thank you for posting (and doing) it!
If you want another test case, I would love to see result for California (which has over 11,000,000 observations).


Very interesting, and a helpful visualisation. Of course there are further levels of granularity that could be attempted. I speak for flies which barely creep into category C. I expect if they were broken down into families some groups (notably hoverflies) might even make it into category B, whereas a great many will definitely be in D.

EDIT: a 2020 paper suggests 160 Australian hoverfly species, from 2022 iNat has 60 species, there are 4320 observations of which 3044 are at species level, which puts them at (0.375, 0.704) which must be in the top right of C. not far left of Moths 2001


Outstanding, excellent paper! I hope I contributed to the Isopoda observations reaching species level :slightly_smiling_face:


A species-level ID was sufficient. As you can appreciate, it wasn’t feasible for us to vet 1 million records given a) the quantity and b) our authorial team didn’t have the taxonomic expertise across the breadth of all terrestrial inverts to do so (although I’ve added IDs to >50,000 observations within the dataset).

We did make an explicit statement about this in the paper: “We note that in some cases observations may be misidentified as an incorrect species, which can impact recognition; however, such misidentifications are not typically of “new” species for iNaturalist, and therefore have little effect on completion.”


at the time hoverflies were indeed the best performing fly family (of the most observed families)

good to see they’ve improved in the time since as well!


most up to date stats has 193 Australian hoverfly species:


Thinking out loud here…

If a species is determined not to be ID’able down to species through photo, would confirming IDs at the genus level then make that observation research grade?

Yes, for iNaturalist. I don’t know if the authors of this paper would treat it that way.

1 Like

Only if someone check marks the “It’s as good as it can be” box. But checking that same box for something at broader IDs will cause it to become Casual (I forget where the cutoff is).

it has to be below family

we used both needs ID and RG records, so cases like this didn’t matter for us

1 Like

This topic was automatically closed 60 days after the last reply. New replies are no longer allowed.