6/6 When the dataset you are using to assess diversity change does not consider the dominant driver of diversity loss, and relies on studies where biodiversity can only go up (e.g., recovery, succession), then it not representative of how humans are altering biodiversity on Earth.

5/6 BioTime does not account for the single largest contributor to biodiversity decline -- habitat loss. Not one study in that dataset shows how 43% of Earth's land surface that used to be forested, but has been converted to agriculture, has influenced biodiversity.

4/6 BioTime has many studies of colonization of new habitats (e.g., dragonflies colonizating new pools). Studies of recovery, succession, and colonization are designed to monitor unidirectional increases in diversity, and do not tell us how humans are altering biodiversity.

3/6 But one must recognize several things about these time-series datasets (e.g., BioTime) that have been used to conclude +'s and -'s are equal: (1) BioTime includes many studies of recovery/succession after a disturbance (plants recovering from the Mt. St. Helens eruption).

2/6 No, it's not a surprise there were many examples of + trends. We've known this since Sax's pubs. There are, in fact, many locations on Earth where invasions have outpaced extinctions. And no one ever expected all locations on Earth to be in diversity decline.

But there is no substitute for reading someone's paper(s) in full and using expert opinions (individual or group) to determine if someone's work is high-quality and original. I wish there was an entirely objective way to evaluate, but a combination of quantitative/qualitative criteria works better.

Great question, but I'm struggling with this like everyone else. Although they get a bad rap, I think metrics like impact factors/citation rates should remain a part of evaluation b/c they have valuable information and allow for comparison. 1/2

It's far easier to say we should ignore current metrics of research quality than it is to propose ones that are better.

Conclusions of papers that use datasets like these are questionable at best, wrong at worst (garbage in = garbage out). Ecology needs more efforts like Gaume and Desquilbet. If researchers who put together datasets won’t do their proper QA/QC, then others should do it for them.

Gaume & Desquilbet show the #InsectChange database is riddled with errors, as are its parent databases (e.g., #BioTime). Errors include misinterpretation of data from original studies, incorrect data entries, duplicate entries, incorrect units, incorrect organismal groups & more.

L. Gaume and M. Desquilbet (10.24072/pcjournal.469) dive into the #InsectChange database, which has been used to show the abundance of terrestrial insects is declining less than previously thought, and that freshwater insect diversity is increasing (e.g., van Klink et al. 2021).