Thread by @zachportman

There’s a new paper reporting on major bee declines that has been picked up by the media: cell.com/one-earth/full…

I don't think the results of that paper are robust.

Feel free to follow along my meandering, largely anecdotal exploration of some of the data. 1/

The paper reports declines after 1990 and big declines after 2005.

My first thought on reading is that the observed pattern is likely due to the lack of taxonomists and taxonomic bottleneck. I examined the data a little deeper, and I think what I found supports my hypothesis. 2/

I’m not a big stats person, so I wanted to just drill into the data for a group I know pretty well, the genus Perdita. With 634 species in the genus, it contains about 12% of the ~5200 bee species in North America. 3/

First, I followed the link in the publication to get the raw data (gbif.org/occurrence/dow…). At 750 MB of data, this dataset is very large. I extracted all the bees in genus Perdita, which came out to 59,722 records. 4/

In total, the dataset has 453 Perdita species (ignoring subspecies). Of these, 420 are recorded after 1950. 5/

The publication compared 1950-1990, post 1990, and 2006-2015. How do the Perdita species numbers break down in those timeperiods?
1950-1990: 404 species
Post 1990: 228 species
2006-2015: 103 species
6/

Looking at those numbers it may seem like there has been a precipitous drop. But examining the 103 species from 2006-2015, the majority of them are from the eastern US (which are easier to ID) or I personally identified them or helped with their identification. 7/

Looking at the numbers, out of the ~4700 records in 2006-2015, about 3900 of those are identified to species. I’m listed as the identifier for ~1600 of those, though it looks like I likely identified more but my name didn’t end up in GBIF. 8/

So there aren’t a lot of Perdita being identified recently. I’ve identified a bunch, but if I had the time I could have identified many many more. It also looks like many that I did identify for recent bee lab projects surveying national parks did not make it into GBIF. 9/

At least in the case of Perdita, there are really only a few people who can identify them. And even for the people who can identify them, it takes a lot of time/effort and Perdita really aren’t a priority for anyone. 10/

Philip Timberlake was the last person to work on Perdita in a major way. He was a prolific taxonomist who described about 460 of them. He died in 1981, and there are big swaths of Perdita species that no one has looked at in-depth in the last 40 years. 11/

There is one subgroup that has been worked on recently, that is relevant for the present dataset. For my PhD, I revised the subgenus Heteroperdita, describing 9 new species and treating 22 species in total. 12/

That work, published in 2016 (researchgate.net/publication/31…), involved getting specimens from many different sources, identifying lots of specimens, as well as going into the field and looking for those species. 13/

Of those 22 Heteroperdita species, all of them have been found since 1990, and 18/22 have been found in 2005-2016. Of the 4 species not found in 2005-2016, I looked for but could not find 2, I didn’t look for 1, and I did find 1 but it’s not in the GBIF dataset. 14/

This also touches on a related point that these species often aren’t found unless you go out to the desert and look for them, which not many people do in the first place. 15/

Anyways, I think it’s telling that the one group that has been worked on recently is disproportionately well-represented in the GBIF dataset. If I had not worked on this group, maybe a handful of the species would have made it into the post-1990 GBIF dataset. 16/

Can you extrapolate from Perdita up to the rest of the dataset? Hard to say for sure, since Perdita are probably not the most representative group, but there are many other genera that are restricted to arid regions, rarely collected, and have had no recent taxonomic work. 17/

The authors did discuss the taxonomic bottleneck as a potential confounding factor, but they cited the low proportion of species only determined to genus as an argument against that. 18/

However, at least when I was in the Logan Bee Lab, we prioritized databasing things that had species determinations or were relevant to current projects. Specimens that were only identified to genus simply stayed in the unsorted drawers. 19/

Overall I think that the results of the paper are likely more a reflection of the lack of taxonomic work being done than anything else. But that’s scary in its own way, because in terms of how bees are doing overall, for most species it’s essentially just a data black hole. 20/

I think it's certainly possible, or even probable, that Perdita species have declined, but I don't think there is any way to tell from this data. Any underlying decline is masked by the incompleteness of the data. 21/

If I had taken the Perdita data from GBIF and then wrote a paper claiming that 200 or 300 species of Perdita had experienced recent declines, people would rightfully call it absurd. 22/

It’s frustrating because we know many bee species are definitely declining. We’ve even found a Perdita that has been extirpated from parts of its former range (Perdita meconis - link.springer.com/article/10.100…). 23/

But for the most part, few people can identify Perdita, and no one really cares about them, or at least people don’t care enough to support basic taxonomic research. 🙁
24/24

25/24

Here is a continued discussion with a response from one of the authors:

https://twitter.com/mmlopezu/status/1353826208207794178