Species ID errors in difficult taxa?

General
1

I noticed that certain taxa have species IDs but are only identifiable by DNA according to the literature. For example, Genus Cryptocercus (a less than charismatic cockroach). In the Appalachian mountains 4 species occur, separated by geography, but not morphology. Nevertheless, we see species-level IDs (often out of correct geographic area) and Research Grade confirmation when no evidence is provided.

So my question: It is useful to researchers to have this kind of squishy data or does it really matter here?
https://www.inaturalist.org/taxa/337963-Cryptocercus

3 Likes
2

I think it’s up to data users to have an understanding of how identifiable particular species are and to evaluate & interpret IDs appropriately. I think the best general approach for researchers is to identify all observations themselves unless working with species that are very easy to ID, or working on a project where ID errors won’t be very problematic.

10 Likes
3

I also think it’s fair in cases where it is well demonstrated in the scientific literature that it isn’t possible to conclusively ID certain species based on photo evidence to ID observations like these to genus, and tick the “cannot be improved” box in the DQA. Some people go through observations and clean up situations similar to the one you’re describing. If you do this, I’d encourage you to make some generic (pun-intended) copy-paste to explain why you’re IDing to genus so other users understand the situation and aren’t unduly frustrated. This might be worth doing on observations where the IDs are obviously mismatched with the ranges.

Otherwise, as @aspidoscelis noted, it’s on researchers to clean and check the data that they use before analyzing it. If they aren’t familiar enough with the system that they’re using to know that there’s a potential ID issue with the data for the taxa that they’re studying, they probably shouldn’t be doing that study!

9 Likes
4

I work with complex plants groups and my answer is for such taxa you choose the most common designated for an area. But experts knows that it could be another (often rarer) species aswel but in the absence of evidence for this rarer taxon we stick to the most commonly accepted taxon.
It’s not necessarily a problem, that’s something common with some groups.

2 Likes
5

It seems like that approach introduces false precision into the data, and can lead to circular reasoning (e.g., it’s mostly like species X because that’s what most of the iNat records in this area are).

I always encourage the opposite - identify to the finest level you can and leave it at that. There is no harm in an observation remaining at genus or any other rank higher than species, and those data still have value.

17 Likes
6

I totally agree! One should be very careful with the “It´s the most common of similar species and thus I ID it as such”-approach. Why not just leave it at genus-level.

2 Likes
7

Usually, once these observations are correctly pulled back by expertise, the rest of the community gleans this information and uses it to limit future IDs themselves as well.

5 Likes
8

As I understand it, iNaturalist discourages identifying based on range. On the other hand, I have had one of my IDs questioned based on range, even though I explained my morphological reason.

9

In the European Adelidae (Microlepidoptera) there are currently two groups whose species can only be separated by DNA barcoding: Nemophora degeerella / scopolii / deceptoriella and Nematopogon adansoniella / prolai / garganellus (from Italy). iNat offers a good solution here: in the list of proposed species, "Complex Nemophora degeerella” or “Complex Nematopogon adansoniella” appear as options. This identification (ID) is more accurate than the indication of only the genus and makes it clear that determination of the species by photograph is not possible in these cases.

I am happy with this solution. In this way, false reports are reduced. Since the distribution of the species is still largely unexplored, the actual areas of the respective species would ideally not be blurred if used consistently. However, this is not guaranteed even with the current solution, since far too often Nemophora degeerella or Nematopogon adansoniella is proposed without in-depth knowledge and then research degree is achieved by negligent confirmation.

I therefore wish that the automatic ID would already draw attention to the problem. Perhaps with these species complexes, an ID at species level can only be made possible after a second precise query?

5 Likes
10

While I agree that morphological evidence should definitely take precedence over range, in many taxa (such as many insects) it is physically impossible to identify without range from regular photos. There are just too many.

3 Likes
11

I did not know that and changed my observations of these moths to the complex now… Two were even at research grade.

Thanks for making me aware

5 Likes
12

This question is important to me. I’ve gone to a lot of effort to participate, but wonder if the data (mine is mostly photos of birds) go anywhere. I find myself getting really frustrated when people who are much less experienced make ID errors, and I have to delete data that are not research grade. But does research grade really mean anything? And are these photos ever used? Also, where are the rules for not using range to separate similar species? Should we upload poor photos? Do numbers matter? Should we repeat images of individuals in flocks? I just deleted a number of IDs with spoilers in them today, and I’m wondering about the system and who it works for. I guess I’m going a bit beyond. But I’ve been wondering the same thing about a couple of Empidonax flycatchers that should really only be IDable by call. I want to help nature, and it’s not transparent how we should best do that here. Thanks for listening!

1 Like
13

If so, is it worth uploading? Not a rhetorical question. Just deleted a bunch of observations where incorrect IDs bumped to genus or above.

14

You should not delete observations just because you don’t agree with the identifications that other users have posted. In many/most cases, these will work themselves out in time as other users chime in. You can also leave comments asking users to explain their ID/explaining your identification to start a discussion. You can tag other expert users asking them to chime in (though this should be done with discretion, and not spamming other users).

I would suggest not getting hung up on Research Grade. Even non-RG observations are accessible to others (just not via GBIF). In some cases, observations don’t reach RG for years.

As far as any specific observation being used, that’s hard to know, but iNat data have certainly been used in publications. You can check out non-exhaustive lists:
https://forum.inaturalist.org/t/published-papers-that-use-inaturalist-data-wiki-1-up-to-2019/2859
https://forum.inaturalist.org/t/published-papers-that-use-inaturalist-data-wiki-2-2020-onwards/20913
https://www.gbif.org/resource/search?limit=50&contentType=literature&literatureType=journal&gbifDatasetKey=50c9509d-22c7-4a22-a47d-8c48425ef4a7

10 Likes
15

You can read the help page https://www.inaturalist.org/pages/help that has many answers to your questions.

2 Likes
16

I suspect empids, especially, need multiple photos and still may be inadequate. Migration and post-breeding dispersal create special problems for birds. Recorded calls are great but rarely occur outside of expected territories. I use Merlin when I can because the data can still be used, even if not on iNat.

2 Likes
17

People often don’t like observations of very common species that cannot be taken beyond genus. I know I would love my Halysidota tessellaris to be just that but there is another species (H. harrisii) that is indistinguishable as an adult. So 99% of images will have no collection/dissection and should be kept at genus. However some species pairs are now under “species complex” ID’s that can narrow the genus down to a pair of species, and this at least finer than genus level only.

I think bouncing any H. tessellaris that isn’t a dissected adult or a caterpillar to genus is necessary but there are a ton of observations of this moth and it’s hard to keep up on.

2 Likes
18

Researchers need identifiable photos with accurate locations or dates. That’s all. We all like good quality photos better than poor ones, but as long as it’s identifiable, it’s useful. So post what you have.

What’s identifiable? Often we don’t know! So we do our best and let the IDers for a particular group work on it.

Research Grade means something, but less than it sounds like. If the group is hard to work with, even photos correctly identified by experts may languish as “needs ID” for years (e.g. some of my Festuca photos). And some things get to RG without good ID’s. Researchers really do use iNaturalist data. If they’re smart, they review the data they download before analyzing it!

I figure that identifying things from photos is hard enough that we should use all the clues we have, importantly including geography. I’m happier – more confident – when I can use morphology AND geography, but if geography is all I have, I use it. Some people get all huffy about that. Oh, well.

Technically, iNaturalist allows you to upload a photo of, say, a flock of birds many times, once for each individual bird. However, please don’t! It wastes the identifier’s time and wastes yours, too.

You do not need to remove an observation that has what you consider an incorrect ID added. You can just leave it, for more identifications to accumulate. Even better, explain why you consider your ID correct. (I have a file of comments I use frequently, so adding them can be fast.) Additional people who come along will read that, perhaps learn something, perhaps agree with your ID, eventually getting your observation to RG.

iNaturalist is so big that it can be frustrating to work with, but I hope you find it satisfying despite a certain amount of sloppiness, which is inevitable in citizen science projects.

5 Likes
19

I thought of an experiment. Photograph a dandelion, including all the possible traits that some people think define microspecies, and see if anyone actually ventures to identify the microspecies, or if they all just leave it as “Section Taraxacum,” with a didactic comment that there are a lot of microspecies. It is very easy to look smart by disputing a common ID; less so to offer something of substance.

2 Likes
20

I really don’t think it’s about “looking smart”, it’s about people who want to create accurate data. Personally, I get very annoyed by people IDing things to species level when there definitely isn’t enough info to support it - a single top-down view of a blurry dandelion, for example. I don’t even pretend to know how to ID dandelion microspecies, but I can definitely say there is not enough info in that observation to support the ID.

Now, if that photo is accompanied by a note detailing the various features the observer checked, and referencing how they support that particular ID, that would be a different matter. Or if there are a lot of photos showing minute details. But let’s be honest, that’s virtually never the case.

9 Likes