I’m the one who started the “placeholder” convention, and it’s based on how BugGuide does it. The intent was for “placeholder” to name ambiguous groups whose ambiguities might be resolved later. This species isn’t ambiguous, but I do see how undescribed species could be rolled into the placeholder mechanism.
1 Like
Hi Jeremy. Here’s my attempt at a super-succinct explanation:
Arachnologists almost universally identify spiders by genitalia. Post photos of spider genitalia from the proper angles, and we can probably take them to species.
And we preserve spiders in alcohol, which changes their color and makes them floppy. So we can’t go through a museum taking pictures of spiders and expect the pictures to look as they do live. And sometimes multiple species and even multiple genera look identical until you examine tiny features under a microscope, so those could never be IDd from in situ photo.
We have to collect spiders live, photograph them live, and then examine their genitalia. Field arachnologists who have done this a lot have a sense of a number of spiders, particularly in their areas of specialty, and that sense isn’t documented well anywhere. There aren’t many field arachnologists on iNat, and I myself come and go in spurts as I get frustrated.
So we don’t have a reliable way to ID spiders by photo. But a bunch of us are trying to work all this out on BugGuide by posting live photos and then photos of genitalia so we can all come to agreement on the species. Once we have agreement on the species, we have a sense of just one color form that spiders of that species can take. Sometimes they take multiple color forms.
And until we have reliably identified photos of all the spiders of a genus in a particular area, we can’t know that some other spider of the genus won’t look exactly like the one we just identified. So we still can’t call spiders that look like this newly identified one the species we found it to be.
It’s mostly the highly visual animals that can be reliably identified by photo. Those include birds, butterflies, and dragonflies. It also includes jumping spiders – at least the males, which tend to be colorful for the same reason that male birds and butterflies are colorful.
6 Likes
Also Jeremy, for an example of how this works, you might explore how four arachnologists recently worked to identify a beautiful species of spider posted to twitter: https://twitter.com/derekhennen/status/1112470447378059269
for what it’s worth, we also identify a large number of plants via examining and sometimes dissecting their genitalia! But it seems less ‘odd’ when you consider that those genitalia are things like flowers and pine cones and such. And they are a lot more visible than a spider’s genitalia
4 Likes
Sometimes I feel like it wouldn’t be a huge jump from the requirements of IDing some of these sedges to IDing spiders by their genitalia…
5 Likes
Thanks, I appreciate you taking the time to explain all that.
But, I’m also already aware all of that. My point was that since - due to iNaturalist’s design - only the groups which are well documented are accurately identified, you might want to focus less on the Sisyphean task of fixing individual observations and more on creating better documentation aimed at a non-expert audience. Then the voting system will work, and a lot of these misidentified observations will get fixed much more easily. Sort of a “many hands makes light work” approach.
I think I may also be more optimistic than you about the identifiability of live spiders to species from photographs. There are a lot of unknowns and uncertainty about how to do it right now, but new knowledge is being created, collected, and distributed at an amazing pace. I am reminded that until Peterson published his first field guide in 1934, bird identification was in a state much like spider identification now: a focus on characters measurable from specimens in a collection, an assumed familiarity with taxonomy and anatomical vocabulary, and restricted to experts. That changed completely, and I think the same thing will happen to spider identification in time. It happened faster for the highly visually attractive species like birds, butterflies, and dragonflies, but I have no doubt it can be done with spiders as well.
Anyway, we’ve veered well off topic, I think. iNaturalist isn’t a great source for range maps for many species thanks to the number of misidentifications, the bias towards populated places, and so on. (Though occasional species have a surprisingly useful map.) I tend to prefer range maps which show the known breeding range of the species, plus regular and expected migratory corridors and wintering ranges for those species which have them. I think range maps should not include places vagrants are occasionally spotted. Although, a lot of range maps would be much improved by some indication of density, so if you did that then areas of sparse and/or rare occurrence could easily be added.
A lot of existing range maps are pretty arbitrary along the edges, honestly. While doing some field work in Iqaluit (Baffin Island, Canadian Arctic Archipelago) I was able to photograph 3 species of birds breeding there which had never been recorded breeding there before. (Junco, Savannah Sparrow, and White-crowned Sparrow. Got them published too. The Junco was particularly far north of its published breeding range.) We also failed to get photographic evidence of several other species breeding there. Before you ask, this wasn’t due to climate change, just because nobody had noticed yet. (Nunavut has roughly the same land area as Indonesia or Mexico but the population of a small town.) Another example of arbitrariness is the range map for Rattus norvegicus. There is (Edit: was; check the rather entertaining comments in the file history and the discussion tab) a hole there for Alberta, because Alberta is known to be rat-free, but rats generally can’t survive in the boreal forest and tundra to the north. The range map shows them there because they can and do overwinter in heated buildings, and they aren’t known to not be there, but they aren’t nearly as widespread as the map implies. They probably aren’t present in most mountains, deserts, and other high-altitude and arid areas. But a global map like this can’t reasonably display that level of detail.
For iNaturalist specifically, the computer vision system probably shouldn’t suggest a species unless there are 10+ research-grade observations within a certain distance, with the exact number and distance depending on the density of research-grade observations of the species.
6 Likes
I can get behind that!
2 Likes
Rather than trying to add all these different layers onto a map, especially given they are not relevant for many families of life, is it not better to use the mapping capabilities of the observation list. By filtering for instance to certain months, not only do you see where for example the bird is in the summer, but also potentially pick up either interesting 'out-of-range- sightings and/or records that should maybe get a 2nd look to see if they are accurate ?
Density is an intersting questionm because I’m never sure if what you see is a function of species density or observer density. Comparing northernb to southern Ontario, I know what the data will say about densities of Black-capped Chickadees, but I am less than convinced it truly says anything about density or commonality of one location versus the other.
3 Likes
the density of sampling issue for iNat is real, but it’s true for almost all science. No one is doing random-point based inventory of chickadees in northern Canada. The location bias for scientific studies is different (known ecologically significant areas, nature preserves, public land, area near universities, etc) but it’s just as real. In fact, on a global basis, academic research is probably MORE skewed than iNat - towards either wealthier countries or high profile areas such as parts of the African savanna or the cloud forest of Costa Rica, etc.
Realistically, non-biased inventory sampling is pretty onerous unless it’s over a small area or very coarse.
3 Likes
What do you think we do up here in July (or as we call it ‘summer’)
4 Likes
ok, well i don’t even. Birders are ridiculous. (good ridiculous) I guess it’s grid based rather than random based but in a way, that is better.
1 Like
There’s a place for both automatically generated maps of all the observations of a species on iNaturalist (or some subset of them), and for carefully compiled maps which attempt to de-bias and show the true probability of observation. I was mostly thinking of the latter, above. The B.C. Breeding Bird Atlas has a good explanation of how you might create the second type.
1 Like
I see iNat as being “rough picture” data… range maps would be biased and approximate given the likelihood for mis-identifications, but that in itself is still useful. Anyone wanting to develop more accurate range maps can use the range boundaries in iNat as places to start investigating and validating the data. If they find particular areas that appear to be affected by bias or mis-identification, then they can organise more thorough and accurate surveys to establish exact range boundaries.
We get so hung up on “it’s right or it’s wrong”, and forget it’s a spectrum! The polarity comes when we are evaluating/deciding whether the available data will answer the questions we are wanting to ask. We just need to make the observations as accurate as is practicable given the methods and skills of the observer
5 Likes
Also I’d just like to add, some places don’t have id’s of a lot of times because they’re new. So nothing shows up on the range map for certain taxa, even though there can be specimens of those things living in the place. And not very species I see mentioned in the tropics in papers is listed under the corresponding genus on iNat.
I saw that happen once, with Ranatra, which I don’t know how to id based on features myself. So I searched to see what Ranatra has been recorded in any inventory of the island. But Ranatra was not observed on the island ever so it’s not on the range map, and the same with the encompassed species. Likewise, because the species they found is not under Ranatra’s taxonomy yet on iNat, it was not possible to add it as an id (e.g. if I wanted to add the id on some confidence the researcher/s found all of them, which is not the case necessarily).
Also, hi. I’m new to the forum, and maybe new to iNat and id-ing (relatively?..).
6 Likes
Hey all,
I’ve been looking at some of the CNC results, and I noticed one of the options to select when downloading data was “out of range”. When I downloaded the observations from my area, I noticed many species that I have seen frequently in the area for 15 years. This got me thinking about the ranges that iNat gives for species, which I had never really looked at before. Many of the ones I looked at seemed super arbitrary, with no consistency of how ranges are actually identified. For instance, the graylag goose range does not include the United States at all even though there are clearly a lot of observations, perhaps because it is not native. The european starling, however, is also introduced and has a range listed across North America.
Does anyone have thoughts on this issue? To me, these ranges don’t seem like a very useful metric for analyzing actual changes in species distribution as of right now, since there is so much inconsistency between taxa. (and some, like the ruff, are just plain screwed up.) I don’t know if this would be an easy thing to improve, but I feel like it would at least help to start out with a general protocol on how ranges are classified.
The range maps are often old and not recently updated or perhaps not including every country. I usually ignore them
My understanding is that Graylag geese don’t have an established breeding population in North America, whereas European Starling definitely does. Most of those geese will be either captive or escaped from captivity.
1 Like
Where I live currently
It’s a pain when I’ve posted over 45 plus gray-footed chipmunks
Only found in this mountain range (Sacramento Mtns NM) going to to davis mountains of Texas
And the AI still won’t suggest it when I’m loading one up
The other is southwestern white pine, it’s the only one in New Mexico
The AI thinks it Scots pine or western white pine
And my favorite limber
One would think after posting 100s of observations in this region
It would get it
Before I forget
The gray-footed chipmunk
At one time the AI
Nailed the ID at about 80 percent
But the last few months it’s at about 5 percent
Instead of improving it went south with accuracy
1 Like