This recent paper makes for interesting reading on spatial and taxonomic biases in biodiversity knowledge (it is based on GBIF data, so includes iNaturalist data, although it doesn’t mention iNaturalist specifically).
Less than 7% of the globe is represented by biological records in GBIF*, when analysed at a cell size of 5 km (see map from the article, below). “Birds are hugely overrepresented; the 85 most-sampled bird species each have more records individually than all reptiles.” A relatively small number of species dominate the available data: 11.4% of all animal records in GBIF come from just ten bird species.
There will always be biases in these sorts of datasets, but based on this study, we can identify some clear pointers for those who want to help compensate for them:
make the most of opportunities to observe in undersampled areas, especially those distant from roads and cities, and in the tropics, deep sea, etc. (of course, these areas are not accessible to many of us - hence the undersampling! - but perhaps there is an undersampled park or coastline not far from where you live, or where at some point you can travel to)
focus your observing on less charismatic groups, such as plants, invertebrates, fungi and even bats, rather than on birds and large mammals (perhaps invest in a cheap clip-on macro lens for your phone, to make it easier to photograph small organisms and plant structures)
for identifiers, find places to add value, such as countries and taxa where identifiers are in short supply. There are likely already many species that have been photographed and added to iNaturalist that have not been identified yet on the site
if you know any specialists, encourage them to join iNaturalist and contribute to identifying some of the undersampled or under-identified groups
I would argue community science can fill many gaps, we just haven’t yet found those people who do spend their time in those not represented remote places.
And yes, no plants at all, while they clearly fall under same biases, e.g. road sampling.
True! And no insects either. Although I guess the general patterns would have been similar. I edited my post to mention that.
I agree… I think we have found some of them, and hope we can find more!
In remote places I usually switch to airplane mode to save battery, make sure I have a good GPS fix, and take photos with the cellphone camera, to upload via the website later. I wouldn’t recommend trying to upload from the field.
Submit the observations you want to submit. You don’t know what future researcher will want, or what future users of the photos will want. Enjoy. And when you can, get further from home to take more pictures, even if they are of the same species.
That’s simply a function of Gbif importing all records from eBird regardless of if it has media. It is more a bias in what records Gbif prioritizes to import.
That’s exactly what I said when I read about those mallards, you go to a place, easily see 700 mallards, no surprise there’re so many records of them if town populations generate more data!
@dlevitis Yeah, it’s because fish are rather hard to photograph normally and clearly enough to get an ID. Especially small fish and fish that hide.
The only reason I have a lot of those rare fish observations is because I catch them first before tasking pictures
Add under-representation during school, especially universities? Because professors and interns, teach class and cannot go survey an/or over-representation during spring and summer breaks, maybe December holidays too?
For bees - over-representation of larger, more charismatic species like Bombus, while smaller species are noticed far less - may also work for other species.
Plants - under-representation of pods and seeds, which are not exactly the same as fruits, right?
Botanically speaking, they actually can be fruits. However, any plant that isn’t showy is likely to be underrepresented. This goes for flowers as well (e.g. grasses in bloom). Overall, I suspect plants that have evolved to attract animals for pollination and/or seed dispersal are much more likely to be noticed and photographed than plants that utilize wind dispersal, for example. Add to it that grasses and similar are hard to key out and identify, especially from pictures, and they are probably heavily underrepresented on iNat.
How do I access the trips feature? How do I get to the form that allows creating a new trip? I see the form described here: https://www.inaturalist.org/pages/trips but how does one get to that form?
I’ll note that iNat’s social media posts of “icky” or “weird” organisms are almost always the ones that get the most views. Pretty pictures of lions or birds are among the worst performers if you go by that metric. Our audience seems to be into the non-mainstream stuff.
Oh, thanks! I didn’t notice the big “New Trip” button on the right side of that page. I think I have a good use for this but will explain that further in the trips that I’ll create.
As for biases, I do have a few thoughts to throw in here.
It seems to me that the iNat data en masse can actually be used to develop a measure of the biases involved, by first computing a “denominator” for each taxa that sums all “expected - actual” population numbers (based on theoretical or “best-known” estimates) in order to “normalize” our actual observations. These ratios could then be factored into “root causes” such as the factors already suggested in this thread, and ratios developed for each of those factors. Thus one could develop an iNat Bias model.
