Observed species total counting error

[pecosvalleydiamond]

Platform:Online

App version number, if a mobile app issue:

Browser, if a website issue: Chrome

URLs (aka web addresses) of any relevant observations or pages:

https://www.inaturalist.org/observations?subview=map&user_id=pecosvalleydiamond

https://www.inaturalist.org/projects/unm-biology

https://forum.inaturalist.org/t/observed-species-total-going-down-over-past-few-days/51143/8

Screenshots of what you are seeing:

Description of problem:

For the last week each time I add new observations I increase the number of observations and species in my iNat main pages (i.e. first link). Following a short delay, the observation numbers go up by an equivalent amount in groups totals (i.e. second link), but my species count decreases by an equivalent number to the increase (e.g., if my species “lifer” count increases by 5, it decreases by 5 in the group page). This error began occurring somewhere around 12,253 spp. and has continued for a week or so (now an ~800 spp. discrepancy). I searched the forums and there seemed to be an eerily similar bug reported a few years back (roughly same spp. counts and problem - third link): https://forum.inaturalist.org/t/observed-species-total-going-down-over-past-few-days/51143/8

I have checked all the same issues as the poster in May 2024 (captive/cultivated, casual observations, etc.) and this does not account for the discrepancy. I have tracked this for several days as I upload records from the summer, and each time the increase/decrease are equivalent (after a slight refresh delay).

Any help correcting this would be appreciated. Thanks

Step 1:

Step 2:

Step 3:

[t_0x16c]

Why are the step numbers listed but kept blank?
Regardless, I guess it won’t matter since this seems like something that’s hard to reproduce.

[pecosvalleydiamond]

@t_0x16c

ep numbers listed but kept blank?
Regardless, I guess it won’t matter since this seems like something t

I wasn’t sure what to list as steps since the effect takes place when I add new records.

[johndreynolds]

I had a similar problem last year. Here is a report that I sent to Tony Iwane @tiwane, who kindly consulted an engineer to look into the problem. I really appreciate their efforts with this mystery. During our correspondence I learned that iNat uses an algorithm to estimate species totals, rather than using exact counts, presumably to reduce computing time. After an initial email exchange, I started tracking my species totals over time. Eventually, the issue stopped happening, and I haven’t noticed it happening since. I’m not sure whether iNat staff fixed the problem, or if for some reason it just stopped. Here is my report to them from last year, in case this helps.

The “Elastic” Algorithm Subtracts Species on the Explore Page Instead of Adding Them

@JohnDReynolds. 18 Sept 2024

Since 10 July 2024 I have added 17,120 observations but “lost” 351 species from the estimation.

The Explore URL I have been using is https://www.inaturalist.org/observations?view=observers.

With the following URL suggested by Tony Iwane… https://www.inaturalist.org/observations?hrank=species&lrank=subspecies&place_id=any&subview=map&user_id=johndreynolds&verifiable=any&view=species

…the number of species is going up while the total from the Explore page is declining.

Clearly, the “Elastic” estimator has been subtracting species instead of adding them.

The slope of the relationship is not quite 1:1. I have been losing about 0.8 species for every new species I add.

Does this happen to other people?

Yes. Since 5 July 2024 I have been occasionally tracking species totals from the Explore page of observers with a lot of observations. I noticed that @damontighe dropped steadily from 12,059 species on 5 July to 11,919 on 29 August despite adding 725 observations. Today (18 Sept 2024), the first time I looked since then, he has gone up for the first time since 5 July, to 12,108 after adding 254 observations.

Not due to taxonomic changes or IDs

Clearly, taxonomic changes (and IDs) cannot explain these declines, since the URL suggested by Tony shows my totals are indeed increasing. Note too that the problem only started on 10 July 2024, and community IDs usually add more species to my total than they remove.

This has happened previously, and been solved!

There is a Forum thread in which people noticed the species total on iNaturalist was decreasing steadily. A fix was reported.

https://forum.inaturalist.org/t/total-number-of-inat-species-is-decreasing/31811

It also happened to @cypselurus for his species totals. He told me it eventually stopped happening:

https://forum.inaturalist.org/t/observed-species-total-going-down-over-past-few-days/51143

The negative trend over 9 weeks lasts year suggests there is an issue with the way the cardinality aggregation is working. Maybe it was over-estimating, and started correcting itself, triggered by my addition of new species, leading to subtraction of species? I don’t know; I’m grasping.

Postscript: As noted at the start of this post, the above is what I wrote to Tony a year ago. I am no longer losing species from the reported total on the Explore / Observers page when I add species. But I noticed @dbeadle has lost about 150 species over the past 2 months. Again, kudos to the iNat team for looking into it, even though the cause remains a mystery, to me at least. I know how busy the iNat team are, and it’s amazing to me that they are able to keep up with so much growth on the site, while maintaining its speed. This may not be a very important bug, but the Explore / Observers page is kind of fun, so I’ll be curious to see if this leads anywhere.

John Reynolds

[pisum]

@pecosvalleydiamond

now an ~800 spp. discrepancy)

there typically is a discrepancy between the species count and the observer’s species count because the former is actually a “leaf” taxon count, and the latter is supposed to be a count of species. see: https://www.inaturalist.org/pages/how_inaturalist_counts_taxa.

however, using a variant of your dynamic life list data from https://jumear.github.io/stirfry/iNat_observations_taxonomy?user_id=pecosvalleydiamond&verifiable=true, i can tie to your species (“leaf” taxon) count but not your observer’s species count.

so it appears there’s something wrong with your observer’s species count.

here are your verifiable=true counts (https://www.inaturalist.org/observations?user_id=pecosvalleydiamond&verifiable=true&view=observers):

here are your verifiable=any counts (https://www.inaturalist.org/observations?user_id=pecosvalleydiamond&verifiable=any&view=observers):

note that verifiable=true should be a subset of verifiable=any. so the numbers for verifiable=true should be lower than the numbers for verifiable=any. this is generally true, except for your observer’s species count (12049 > 11850).

based on my own attempts to calculate your observer’s species count, it seems like the verifiable=true observer’s species count is too high, and the verifiable=any observer’s species count is too low.

my own counts look reasonable (although i didn’t try to tie exactly):

• https://www.inaturalist.org/observations?user_id=pisum&verifiable=true&view=observers
• https://www.inaturalist.org/observations?user_id=pisum&verifiable=any&view=observers

…

@pecosvalleydiamond

https://forum.inaturalist.org/t/observed-species-total-going-down-over-past-few-days/51143/8

technically, i think this describes a different problem, because this is talking about a problem in the species (“leaf” taxon) count, which does not seem to be the problem in your case. however, whatever is the underlying cause of your problem could be similar to the problem in the other thread, which is actually better described in https://forum.inaturalist.org/t/total-number-of-inat-species-is-decreasing/31811.

[pecosvalleydiamond]

Thanks all - sounds like just best not think about it and continue on (and it might autocorrect in the future). I took a series of screenshots tonight while adding observations and I’d agree that it always seems to lag a bit (not a 1:1 ratio, although it seemed to catch up over time). Interesting that it seems to be a phenomenon that people keep noticing at about 12k spp.

Cheers,

-Moses

[pisum]

@pecosvalleydiamond

sounds like just best not think about it and continue on

that doesn’t sound like the right path to me. it’s worth at least tracking down why the observer’s species count is strange in your case, especially while the problem clearly exists. if nobody looks at it until later, and the problem has mysteriously resolved itself then, then it’ll be hard to troubleshoot the problem.

[pecosvalleydiamond]

If anyone is interested in pursuing - here are some screen captures from the past couple days. The left is the rankings in the UNM Bio group and the right is my verifiable observations page. The UNM Bio group should be including only verifiable records, so these numbers should match (the mismatches in observations tend to indicate lag in the system, which usually catches up within half a day).

Thanks all

[pisum]

@pecosvalleydiamond

The UNM Bio group should be including only verifiable records, so these numbers should match

they will not match. i’ve already explained why:

@pisum

there typically is a discrepancy between the species count and the observer’s species count because the former is actually a “leaf” taxon count, and the latter is supposed to be a count of species. see: https://www.inaturalist.org/pages/how_inaturalist_counts_taxa.

that said, it looks like your observer’s species count now looks reasonable. whatever was the problem i noted earlier seems to no longer be a problem. so it’s possible the iNat staff have already addressed the issue, or else the issue has resolved itself, and it might be difficult now to troubleshoot whatever the problem was.

[pecosvalleydiamond]

I think I need to be more specific in my language:

The UNM Bio group should be including only verifiable records, so these numbers should match

In this case I was referring to the number of total verifiable observations I have made in iNat. I should have been more specific in my language. At 7:28AM on 9/5/25 the lag had caught up, thus the number of observations was equivalent both on my main page and in the group (95,886)

See above - observations (in red) are equivalent (indicating that iNat lag has caught up and both my main page and the groups page are including the same number of observations). Species numbers are not equivalent (in blue) which I understood as the discrepancy between leaf counts and the estimator as you described above. This is a quirk, but not the problem I was reporting.

In your previous response you linked my verifiable=true counts and my verifiable=any counts

Since those screenshots were taken I have uploaded additional observations to iNat, many of which were new species for me. This is reflected in an increase in my species count on my main page (seems to work as it should). This is however not reflected on the groups page, where each additional species results in a roughly equivalent decrease. The screenshot in my previous post shows the following:

Time: Group species count, main page species count

5:52 PM on 9/3/25: 12053, 12872

8:39 PM on 9/3/25: 12049, 12883

9:48 PM on 9/2/25: 12037, 12907

7:28 AM on 9/5/25: 11984, 12943

The main page species count is increasing correctly as I add additional observations of new species, while the group species count decreases at the same time (e.g., between 9:48 PM on 9/2/25 and 7:28 AM on 9/5/25 my main page species count increased by 36 while my Group species count decreased by 53, with matching numbers of observations exhibiting that the group page had caught up to the lag and was considering the same set of observations)

I have taken a screenshot from each of the two links you posted in your initial response at the current time:

Note that since your response my verifiable=true count in dark gray has increased from 12881 to 12946 (+ 65), while my count in white has decreased from 12049 to 11985 (-64).

This is the problem I am reporting - increasing my observed species on the main page (uploading additional unseen taxa) is decreasing my observed species in the group page. Even if one is an estimator and one is a leaf count (thus they are not equivalent to each other), they should both increase as I add new species records.

I do not know how iNat is coded, but the fact that one increases and the other decreases by a roughly equivalent amount over time (hard to tell exactly because changes in identification are taking place simultaneously in the background) makes me think this bug might be something like the group species count code has a polarity reversal in its counting iterator - e.g.,

Group_spp_count -= new_taxon_observation

rather than

Group_spp_count += new_taxon_observation

This seems similar to the previous reports by other iNat users around 12k species, so perhaps the code switches estimators above some value to save time in calculations? Again, I’ve never looked at the background code, so this is me speculatively troubleshooting code.

I hope this clears up some of the confusion I introduced. I appreciate the time you have taken on this issue.

Thanks,

-Moses

[pisum]

read the page that i referenced. it talks in detail about the difference between counts based on leaves and counts based on species.

@pisum

there typically is a discrepancy between the species count and the observer’s species count because the former is actually a “leaf” taxon count, and the latter is supposed to be a count of species. see: https://www.inaturalist.org/pages/how_inaturalist_counts_taxa.

…

UPDATE:

@pisum

using a variant of your dynamic life list data from https://jumear.github.io/stirfry/iNat_observations_taxonomy?user_id=pecosvalleydiamond&verifiable=true, i can tie to your species (“leaf” taxon) count but not your observer’s species count.

i modified this page slightly to include optional columns that indicate “leaf” and “species”. this simplifies the process of independently determining “leaf” and “species” counts because now you can simply search for the icon representing “leaf” or “species” and use the number of occurrences found as the count. it also helps to visualize the difference between “leaves” and “species”. for example, the screenshot below shows that you have 3 observations of genus Haemopis, which adds 1 to your “leaf” count" but does not add to your “species” count.

https://jumear.github.io/stirfry/iNat_observations_taxonomy?user_id=pecosvalleydiamond&verifiable=true&options=leaf,species

@pisum

that said, it looks like your observer’s species count now looks reasonable.

using the updated page to independently determine the observer’s species count (based on “species”), it looks like the latest count shown for verifiable=true is still off. currently, it is displayed as 11975, but it should be 12055.

the species count (based on “leaf”) is correct. it currently shows as 12958.

for verifiable=any, the species count (based on “leaf”) also correct (13129), and the observer’s species count (based on “species” is off by just 1. that latter count is displayed as 12214 (but it should be 12213).

[pecosvalleydiamond]

I did an experiment trying to help debug, so I spent the last hour uploading batches of photos and checking how it affected a few different counts.

Each time I uploaded a batch of ~30 observations, I then went to go check the numbers at each of the following pages:

1 - Verifiable Leaf - restricted to all verifiable records

2 - Verifiable Species - restricted to all verifiable records + high-limit species/hybrid (based on example on the page you linked)

3 - UNM Biology - the example group * note that these numbers often lagged a bit (I assume iNat only updates group rankings every few minutes rather than continuously)

4 - Verifiable=TRUE - link from your first response

5 - Verifiable=ANY - link from your first response

Here are the results over the hour, along with the difference between each time point and the previous time point:

I will use the numbers above with the subscripts o for observations and s for species counts (leaf or species, whichever is displayed).

1o = 4o makes sense, just shows that all observations are being tallied

1s > 2s makes sense, there should always be more leaf observations than species observations

3o & 3s is always a bit laggy, but eventually seems to match 4o & 4s, which makes sense according to your link (species count used in comparisons, not leaf counts)

Δ1o = Δ4o = Δ5o makes sense, in this case because all uploads during this time period were wild (verifiable) images. This also shows that link 5 includes the same observations as link 1.

4s - this is the column I believe is causing all the problems (3s is just a lagged copy of it). Δ4s is always negative and almost always with the same magnitude as Δ2s.

Comparing column Δ2s and Δ4s, one can see that as I upload new observations, some are ID’d as new leaves (Δ1s) and a subset are ID’d as new species (Δ2s). For some reason, simultaneously, each additional new species (2s) subtracts from the total species I have observed that are Verifiable=TRUE (4s).

Basically, something is wrong with the species counter in 4s, which should be following the counter in 2s (both calculated based on species not leaf species), but subtracts by the amount it should be adding. Groups seem to check the value in 4s (which initially alerted me to a problem), it just does so with some lag.

Cheers,

-Moses

[pisum]

@pecosvalleydiamond

1s > 2s makes sense

in my mind, 2s is relevant only to the extent that 2s should be equal to 3s and 4s. so to the extent they are not all equal, then there is a problem.

1s should be greater than or equal to any of 2s, 3s, and 4s (since 2s, 3s, and 4s should all be equal).

@pecosvalleydiamond

4s - this is the column I believe is causing all the problems

while i agree that 4s(=3s) currently is not correct, 5s also is not correct (although 5s is less incorrect than 4s). fundamentally, the observer’s species count (based on “species” from the API’s /observations/observers endpoint) is not correct for you.

the problem is probably related to the the settings on the system’s indexes and the fact that you have so many observations – probably similar but not the same as the issue noted above:

@pisum

https://forum.inaturalist.org/t/total-number-of-inat-species-is-decreasing/31811.

…

UPDATE:
i dug around, and it looks like iNat staff have identified another similar issue before and have chosen not to address it: https://forum.inaturalist.org/t/number-of-species-greater-than-number-of-observation/59789/4. that other issue deals with the observer’s observation counts, but it’s possible that the same analysis applies to the observer’s species counts.

[astra_the_dragon]

@pecosvalleydiamond considering that this is a known feature of the idiosyncrasies of iNat’s counting systems and not a novel bug, I suggest we close out this bug report.

[pecosvalleydiamond]

Update - either someone fixed this, or I powered through the error (as mentioned by @johndreynolds and the post by @cypselurus).

In case someone is still working on this, or it comes up again in the future, I added a couple more updates to my spreadsheet from earlier in the day. Note that at some point this evening the column 4s suddenly began going positive again (for the first time in ~ a week).

Thanks to @pisum for all your help on this. I’ll go ahead and close out the report.

-Moses

[pisum]

@pecosvalleydiamond

either someone fixed this, or I powered through the error

your observer species counts are still off a bit, although they currently are less off than they were. maybe it’s close enough that nobody should care, but i just wanted to make it clear that there is still a problem, in case iNat staff still were planning to look at it.