iNaturalist's earthworm problem and how to fix it

iNaturalist’s earthworms desperately need identification and curation. I’m sure there are other iNat users who have also dedicated themselves to identifying some group that is frequently misidentified and I would like to know what you’ve done to improve these situations. Trying to review ~150k observations essentially on my own is a lot of work and I do wonder if I am wasting my time on this task.

If you aren’t familiar with what’s going on with oligochaetes, here is an explanation of the issue: Earthworms (Crassiclitellata + Moniligastrida) are usually very difficult or impossible to identify precisely from photos of live animals, and few iNaturalist observers or identifiers are aware of how to identify earthworms. Regardless, many earthworm observations have been wrongly identified to species, and when untrained identifiers allow misidentified or unidentifiable observations to reach Research Grade, these incorrect IDs train the Computer Vision to suggest more misidentifications. The most egregious example is probably *Lumbricus terrestris*, which the CV suggests for nearly all oligochaetes, even those that do not resemble that very distinctive species; this is probably also due to observers being misled by the ridiculous common name “Common Earthworm.” There are also so few worm taxa in the CV that it just doesn’t have the training to provide good IDs to even the family level. CV identifications of earthworms that I have corrected in the past several months have been wrong about nearly 70% of all IDs and wrong about nearly 90% of species IDs. Although there have been some attempts from earthworm specialists and amateur identifiers to correct misidentified earthworms by either providing correct IDs or moving observations to higher taxa, none of their admirable efforts have managed to undo the CV issues due to the sheer volume of a decade of misidentified observations.

After realizing these issues last year, I started identifying earthworm observations to the best of my ability and tried to provide information about earthworm ID through comments and links to my journal posts. I also applied for curator status so that I could address long-ignored flags on various earthworm taxa; porting the earthworm taxon framework from WoRMS means iNat inherited some bizarre, antiquated, or simply incorrect taxonomy, and many of these issues still stand since I’m not fully sure how to deal with them or haven’t noticed them yet. Few earthworm specialists regularly use iNaturalist (which is understandable since it does not seem very useful for their work) and that means I do not get taxon-specific advice on my own curation flags and have nobody familiar enough with earthworms to provide needed IDs for global earthworm observations.

So far, I’ve focused on correcting and/or supporting Research Grade observations of various common species to try and rein in the CV. Although nearly all observation photos still lack sufficient details to key worms to species (and in some cases, the CV may never be able to offer correct IDs on photos even at family/genus level because dissections are needed), adding more taxa to the CV might at the very least give it more options and make it more “cautious.” Thanks to a few observers and identifiers who are helping me out, several species and genera of earthworms have been included in recent CV models while others have been lost. Recently, I finished correcting all Research Grade Lumbricus terrestris (only a little over half of the ~4,000 observations were correct–no wonder it was causing issues!). Hopefully this is reflected in the next CV update, but it seems unlikely to me that the model will ever perform at even half the level it should, or, more importantly, that enough observers will know to avoid choosing its incorrect suggestions.

What more can I do? I would like to train others to help out with corrections but do not know if there are people willing to spend their time on this. I plan on making more photographic guides to earthworms, but even though I get encouraging and thankful comments on those posts, only a few people who say they will take worm ID photos actually follow through with that. How can I get more people to provide good earthworm photos? Am I identifying things the right way? Are my copy-paste explanations of why users’ blurry earthworm photos aren’t identifiable actually helpful or just a waste of time? What should I do to curate the errors in iNat earthworm taxonomy and common names that both lead to many incorrect IDs?

I know nobody but myself is making me attempt this task, but I feel bad letting this problem persist. If I can’t review recent observations for even a few days, the backlog takes a long time to work through and more wrong RG observations crop up. Even though they are a group that really should not be identified from photos, it still seems a shame for iNaturalist’s earthworms to remain in this situation. I believe that there is still potential for iNat to be a useful tool for recording difficult organisms like these and want to help get it there, but there is a lot of work left to do.

Skimming thru your journal posts - now to the comments. That is your potential pool of interested identifiers willing to help.

From your link https://www.inaturalist.org/observations/identify?reviewed=any&quality_grade=research&taxon_id=81545&disagreements=true&place_id=any

in https://www.inaturalist.org/journal/thirty_legs/116702-state-of-computer-vision-earthworm-id-02-09-2025#new_comment

my African slice is cleared. There are only 10 obs left …

I don’t have much input into the matter but I will say thank you for having the concern and communicating this. I wouldn’t of been aware otherwise so I feel making a post here was a good step.

Remove it then, there has been a flag open about it for months with no objections.

Given the sheer volume of the task, I would put a priority on working with others to break it down into smaller portions. It will be less discouraging if you do not have to feel like you are trying to tackle this entire mountain alone.

I think you have done an excellent job laying the groundwork for this by writing journal entries; maybe the next step now is coordinating efforts. Since it sounds like there is interest, maybe it would make sense to get in touch with some of the users who have attempted to tackle this in the past, along with people who have responded to your journal post, and come up with a plan to divide up the labor – maybe each person can take a country, or you all agree to start by focusing on a species like Lumbricus terrestris until it has been cleaned up and then moving on to another one.

Another thing you might consider doing is removing the name “common earthworm” if this is contributing to people selecting one particular species.

You might look at some of @matthewvosper 's posts in his syrphid project for ideas about how to use project journal entries to coordinate a team of willing IDers.

Thank you for your service! It can be a lonely battle while you’re waiting for backup. Keep doing what you’re doing. Help will eventually come!

The very fact that you have so many earthworm observations to deal with, shows that many people are interested in earthworms.

Maybe 1% of them might be interested in training up as fellow soldiers in your battle to fix the CV. It’s just a matter of persistence until you find those special people! Hold the line!

Similar problem with fungi: you might find this discussion interesting.

https://forum.inaturalist.org/t/ai-suggestions-in-fungal-observations/69297

I started out IDing barnacles and ran into a lot of the same issues. They’re a weird and confusing group with very few people interested or actively IDing. Very general common names, lots of look-alikes, so many blurry or far away pictures. It took a lot of research to get a basic level of understanding, and I still felt mostly confused and overwhelmed by the sheer quantity of observations globally. Lately, I’ve been more into figuring out keys for North American crab groups where there are more active IDers. One day I’ll get back to tackling the barnacles like you are with earthworms!

Ultimately, there is limited things you can do. The CV is just not well equipped for a group like this.

I would say join others and advocate for CV changes / improvements. There are many threads.

Or double down and try to get more earthworm taxa in the CV to help it become more cautious. There are a few earthworms that could probably get to eligibility if misidentifications that have not been withdrawn were made maverick. But this would require other people.

https://www.inaturalist.org/taxa/484184-Dendrobaena-veneta
Has up to 48 stuck observations

https://www.inaturalist.org/taxa/484185-Aporrectodea-longa
Up to 15

https://www.inaturalist.org/taxa/364093-Amynthas-agrestis
Has up to 47, actually more than its current number of observations

Its possible these could get in the CV from observations already identifed with an extra push. I’m sure there’s a few more. It is unfortunately common for some people not to withdraw IDs. So a pile can form.

Firstly thanks a ton for tackling this task. I think you’re doing everything well, it’s just one of the biggest misidentification issues on the platform so it’s going to be frustrating. I’ve been aware that earthworms are a mess for years but other than pushing back an L. terrestris observation or two I haven’t made any attempts to help because I had no idea what to do.

Basic questions like:

  • what rank do I push L. terrestris back to?
  • are any earthworms identifiable from photos or are they all hopeless?
  • maybe some of them can actually be ID’d as L. terrestris and I shouldn’t push all of them back?
  • if any are identifiable, what photos are needed and how much work does it take to analyze the photos?

Last time I tried checking I couldn’t find anyone on iNat to answer those questions. I only have a couple observations of earthworms in part because it didn’t feel worth taking photos of them if I didn’t know what angles to get, and I didn’t expect them to ever get identified even if I got the correct photos.

What is the easiest way we can help you? Like I’m happy to help by going through earthworms in my region, e.g. there are 2400 verifiable Crassiclitellata observations ranked at genus or lower in Ontario which isn’t too overwhelming. But thinking about trying to count body segments on all of those is overwhelming. Maybe a flowchart would help - if no underside photo → ID to order/family(?), if clear underside photo can I tag you? Or add it to a project/observation field for review later?

Yeah, there’s been a couple people who have started getting diagnostic pics of the US invasive pheretimoids like A. agrestis. I had hoped for that species, A. tokioensis, and M. hilgendorfi to all get in the CV together since they are always together and nearly indistinguishable, but looks like hilgendorfi is going to get there first since it’s easier to ID. But I am looking forward to that, since it might fix the “Metaphire” over-ID situation, and maybe give more “Megascolecidae” IDs in the US (CV seems to have forgotten the family there, which is a big problem!)

I do agree something needs to change about how the CV works and what taxa it can be applied to. Birds and butterflies you can get away with photo recognition, but there must be far more taxa than anything identifiable by sight that you need to dissect or examine spores or genitalia for. I’ll put my two cents in on that suggestion thread eventually

I’ve been meaning to make identifier guides that pretty much do that, but I’ve not got around to it yet! Thanks for the suggestions though, they are very helpful and I appreciate any feedback on my ID resources.

To answer those questions:

  • rank pushback: Depends on area and what’s present there. In Ontario, for example, introduced Lumbricidae dominate the earthworm community and you’d be right more often than not if you took everything to that family. If features visible, genus Lumbricus usually is recognizeable. But if you’re IDing in India, the family Moniligastridae (which falls outside of the order Crassiclitellata that contains most earthworms) is prevalent and so it is often not safe to ID past Oligochaeta. Basically everything tropical is in such a poor state taxonomy-wise that family IDs are a pleasant surprise. Lumbricus terrestris is only present in Northern Europe, temperate North America, Tasmania, New Zealand, and southern South America; most other observations are incorrect.
  • ID by photos: Also depends on area/taxon. Certain things like Allolobophora chlorotica and Amynthas hupeiensis in the USA/Canada can be identified from even blurry photos because they are the only greenish worms there and they are readily distinguishable from one another. But in the 1,000-spp strong pheretimoid clade in East Asia, genera are separated artificially by genital structures and so you get several paraphyletic genera that cannot be identified past family Megascolecidae; Metaphire, Amynthas etc. are phylogenetically meaningless.
  • terrestris: Lumbricus terrestris is actually very distinctive and a good number of observations are actually correct, if only by chance! There is the matter of L. herculeus, a cryptic species identical in every way to terrestris and only separable by DNA. No input from lumbricid experts on the “species group” flag, so I have nothing further to say about that unpleasant situation until someone who knows more than I do tells me what I should do.

Segment counting is only for Lumbricidae for the most part and that’s only to determine clitellum position. If male pore (usually 15) is visible or first segment visible, counting can be possible from poor photos not showing underside. But it is definitely time-consuming even if you don’t have to count tip to tail.

I’d definitely appreciate suggestions for a identifier flowchart. It wouldn’t even take training for anyone interested in just taking dried-up decayed worm jerky or photos from 20 feet away back to Oligochaeta.

Okay looks like in Ontario most are Lumbricidae, a few are Megascolecidae and based on looking at observations and your journal here those are distinctive based on the narrow bright pale clitellum?

That sounds like a pretty strong case for having a complex on iNat to me. Ideally you can find a reference in literature for formal use of a complex with those species. Normally there’s a fairly specific definition of a “species complex”, but because we don’t have the flexibility of more informal ranks like species groups I’ve used complexes in the past without that using this logic.

Is L. herculeus known to occupy a similar range to terrestris? And other than that, what is distinctive about terrestris compared to other species in the genus/family?

Counting segments and recognizing the perichaetine setae is more reliable than color (clitellum can darken with age/reproductive status, and some lumbricids have a pale clitellum), but yes, pheretimoids are pretty different in appearance to the lumbricids once you know what each tend to look like.

This paper is probably perfect if you want to read up on local worms; just note that some of Reynolds’ species/genera are not reflected in current taxonomy or on iNaturalist: https://www.researchgate.net/publication/364122906_The_Earthworms_Lumbricidae_Megascolecidae_and_Sparganophilidae_in_Canada

L. terrestris: prostomium tanylobous (most common lumbricids epilobous), clitellum on 32-37, tumescences on 25,26, dark reddish brown head, pale flat tail, probably the largest lumbricid in most of its range. Coloration is variable with age/diet/moisture. L. herculeus is apparently not present in the US samples taken so far, but seems fairly widely distributed in Europe; it does form a monophyletic clade with terrestris. Here’s the paper reinstating the species: https://www.researchgate.net/publication/233970366_DNA_Barcoding_Reveals_Cryptic_Diversity_in_Lumbricus_terrestris_L_1758_Clitellata_Resurrection_of_L_herculeus_Savigny_1826

I mainly have been hesitant to make a complex on my own since it will impact so many observations (most of which aren’t even Lumbricus) and wasn’t sure what bounds it should cover. Just Europe? US too to be safe? What about NZ “terrestris” which haven’t been examined for cryptic species? There is too little information even on this well-known species for me to be confident without expert advice. Same can be said for nearly all common European species, which have their own horrid cryptic lineages that can be even more complex and confusing than that of terrestris/herculeus.

A good read, and probably an even worse situation than earthworms, to be honest. The outcome of general frustration and “something about the CV needs to change” is what I would expect the same discussion about earthworms to end in, although at least fungi have specialists already working on them!

I can not help you a lot with your specific questions and I also feel I can not really help you on your ID quest, as I am rather busy with my own already. But I really would love to let you know how much I do appreciate your work! You left a comment on one of my earthworm observations about how I could provide better photos and I since keep that in mind when I observe earthworms. It does not always work out, but I for sure try now to get the right shots. And I was really amazed that now I have some observations IDed to genus or even species, incredible! Thanks for taking on such a tough taxon in so many regards. I hope you get a strong task force togehter to help you out! You surely do a lot to make it happen, I love your journal posts! So inspiring. I always wanted to start something like this, but I just never come around to it.

I wonder if you might be able to attempt something of what @zdanko has done in the past and host zoom meetings where you go over the ID of a particular earthworm taxon and then get people to help out on cleaning up that taxon on an every-few-weeks basis, once you have some kind of identification guides put together? Hopefully Zachary doesn’t mind being tagged here, but I think his FlyGuide ID courses were a pretty cool approach to tackling some of the commonly mis-IDed fly taxa on iNat. It also generated a little bit of a community of people who were interested in working on IDing them, which would be a huge benefit here too. You might also be able to take some cues from the FlyGuide site’s style of photo ID guides as a framework for your own guides.

Of course, the earthworm problem is definitely larger in breadth than most of the problems with fly taxa here. But that might be a start! If I had infinite time I’d happily offer to help take on that guide-making, even though I’d be starting from scratch on worm ID (but i DO love learning how to identify things in general…). That said, I really admire that you’re working on tackling this, and I’ll have to take a look at the resources you have put together too. If you do keep putting together guides & ever want to host an event to talk about worm ID or something, count me in as interested =}

I want to echo others in saying huge thanks!
Sounds like just getting others on board to help will make a big difference.
I´m sure there are many of us who would be glad to lend a hand.

I was one of the people who said they would take better photos in the Spring but then got distracted / left my country of residence to travel. This is a good reminder!

I´m a little bit rammed right now with things the coming weeks, but in general I´d love to help out

  • with basic identification support in pulling things back to a higher level
  • working through Icelandic/UK obs…
  • with documenting more species and doing a better job of it….
  • even with field guides could be cool if you need help with layout / design.

So if you make a list of people to garner help from in general, please include me. Otherwise I will aim to come back to this more soon! I imagine others like me, with good intentions… may also just need a prod to be reminded but are genuine in their desire to take better obs, etc if there is support with ID.

Thanks, I’d appreciate the help!

There is a UK national earthworms recorder on here, and he tends to remove all observations that aren’t extra lucky or stellar quality (including some that I would be confident about species/genus ID, but it’s understandable) to family Lumbricidae. Even just going after super blurry or dried up specimens that you don’t need to key out would be a great help!

Given the variation in guidance by region just in this one reply, I would definitely try to divvy up your guidance into region-specific guides that go over all of the topics in this post (what can you ID from a typical iNat observation, what can you not ID, what’s a safe taxon to push things back to) - per continent might be good for a first stab, and then if a particular continent’s guide gets too complex, split it into sub-regions of the continent.

Then publicize the guides however you can - on the forum, on your profile, having a short blurb and link to your guides as something you can just quickly copy-paste into your IDs, etc.