Read on Twitter
(1/ ) Thread: New work by @gndaskalova @AlbertPhillimo2 @IslaHMS is _super_ important for the #InsectApocalypse conversation: a simple statistical change to account for pseudoreplication applied to the 2019 German study switches 4/5 of the reported declines to non-significant. https://twitter.com/gndaskalova/status/1350090413110157317
(2/ ) The German study (which inspired some big headlines warning of the coming #Insectageddon), suffered from year-based pseudoreplication – i.e. many of their sampling events took place within the same years (a limitation with many biodiversity studies).
(3/ ) Another issue with the German paper was their first sampling year had high numbers of insects in comparison to later years, so when Daskalova et al. removed this first year, the results contained much more uncertainty.
(4/ ) The authors of the German study defended their work and claim the statistical changes are inappropriate for time-series insect biodiversity data: https://doi.org/10.1111/icad.12467 Credit to them for responding in good faith, and especially for making their data public.
(5/ ) I asked the lead author Gergana Daskalova about the Seibold et al. response and she acknowledged that most insect time series data will be short in duration, but still useful for drawing inferences...
(6/ ) ...and that temporal autocorrelation (observations taken closely together in time) would ideally be accounted for in models (as neither paper included this in their scope) but that ultimately this will further increase uncertainty in estimated trends.
(7/ ) However, Daskalova et al. disagree with the Seibold et al. interpretation of one of the central statistical concepts used in both of their papers. This seems to be the heart of the issue. https://twitter.com/gndaskalova/status/1350384707918307333
(8/ ) And Daskalova et al. disagree with Seibold et al. on how these kinds of models should be interpreted. https://twitter.com/gndaskalova/status/1350384935627059202
(9/ ) This new paper probably won't attract the same media attention as the German one, because statistical methods just aren't as shareable or clickable for readers. Its also unlikely the major media outlets will update their original stories to show the results are disputed.
(10/ ) Its sad b/c this new paper adds a lot of perspective to counter some of the (ludicrous) downstream claims in the media (imminent extinction of all insects etc). It shows the importance of how data is collected, analysed and framed. Especially when big claims are being made
(11/ ) In a wider analysis of published invertebrate biodiversity time series data, Daskalova et al. found no evidence for insect declines on average, and that the most extreme reports of decline tend to be associated with studies with a shorter duration.
(12/ ) "It is striking that some of the prominent studies in the insect decline literature have been of short duration, potentially revealing a bias toward high impact journals publishing more extreme and “surprising” results and/or their subsequent amplification by the media"
(13/ ) Lead author originally noted on twitter that diversity/abundance in one year will influence diversity/abundance in the next (autocorrelation between consecutive years). This wasn't tested by the German study or Daskalova et al, but would increase uncertainty even more.
(14/ ) Long term, robust, thorough insect monitoring studies tend to show a mixture of trends, and some even show overall increases. E.g. the 50 year moth biomass study by MacGregor et al: https://doi.org/10.1038/s41559-019-1028-6 (paywalled)
(15/ ) One of the things this kind of long-term research shows us is that when it comes to insect biodiversity declines, it depends on your date range and how many samples you take. This is because of the high variability between years and even decades.
(16/ ) None of this is to say there are no declines happening. There is good evidence for some location-specific declines in certain groups of insects, and we should care about this. But this is not the same thing as evidence for panic-inducing global insect declines.
(17/ ) Daskalova et al. note they can't rule out an average trend of 1-6%/year decline for various metrics in the time series data, they note this would be serious, but the take away is that the type of data currently available do not support the 'insect apocalypse' narrative.
(18/ ) Everyone seems to agree on the importance of long-term, standardised insect monitoring studies, and on the importance of discovering and describing Earths biodiversity as quickly as possible. IMO #Insectageddon media discourse is a distraction from what is really important
(19/ ) One of these is the field of taxonomy - discovering, describing, and classifying species, and it deserves more recognition and funding: http://bit.ly/2ZxdWWc
(20/ ) As Daskalova et al state: "media attention that overstates the problem runs the risk of later undermining wider public confidence in biodiversity research." I agree, and think in the long term, the science needs to be robust and honest about its limitations.
(21/ ) In case anyone is interested, I tried to summarise the main problems with the literature on insect declines back in 2019: https://tomsaunders.co.nz/insect-apocalypse/
and recorded a #Scigest podcast with some colleagues about the issues too: http://bit.ly/2N4LruC
and recorded a #Scigest podcast with some colleagues about the issues too: http://bit.ly/2N4LruC
