Estimating local biodiversity change: a critique of papers claiming no net loss of local diversity
Authors: Andrew Gonzalez, Bradley J. Cardinale, Ginger R. H. Allington, Jarrett Byrnes, K. Arthur Endsley, Daniel G. Brown, David U. Hooper, Forest Isbell, Mary I. O’Connor, and Michel Loreau. 2016. Ecology97(8): 1949 – 1960.
Blog author: Maria Goller
First author bio: Andrew Gonzalez
Professor at McGill University
His lab focuses on causes and consequences of changes in biodiversity. They translate their research into conservation policy and are particularly interested in retaining ecological connectivity in cities.
Twitter handle is @bio_diverse
Second author bio: Bradley Cardinale
Professor at University of Michigan
Director of the Cooperative Institute for Great Lakes Research, CIGLR
He uses modeling, meta-analyses, and observational research to look at how human behavior impacts biodiversity.
Overview
Earth is losing many species incredibly quickly.
Biodiversity is decreasing but it is unclear what local trends are.
Locally, species richness varies with human impact.
Locally, humans can increase diversity (introducing new species, etc...).
This paper is a direct critique of Velland et al. (2013) and Dornelas et al. (2014), who argued that local biodiversity is remaining the same (species turnover may occur, but total richness is constant).
These authors wanted to know whether the findings of these two papers are representative of global trends in biodiversity.
They quantified three potential areas of bias in the two meta-analyses.
1) Quantified spatial bias
Created maps of human impact across globe using various impact databases
Randomly sampled same number of locations as used in each paper's dataset, did this 1,000 times, then looked at spatial bias of their studies.
Caveat: human impact maps weren't created to show impact on biodiversity specifically
Spatial bias of both studies
V. et al. study focused on US and EU => underrepresentation of tropics, boreal forest, tundra, desert
Overrepresentation of forests now recovering from logging
D. et al. study focused on only one of the oceans (North Atlantic), but did include the most heavily impacted areas based on their maps
RESULTS: studies limited in scope to specific areas, ignoring areas of major changes in human impact, and also focused on recovering habitats
2) Quantified temporal bias
Analyses based on short time scales do not accurately reflect changes to diversity
Used the datasets and did a linear regression looking at the effect of study duration on biodiversity decrease => excluding long studies underestimates diversity loss
They argue this is because "extinction debt" may be "repaid" only after many years (local extinctions take time)
3) Quantified post-disturbance bias
Studies often include species richness measures AFTER the ecosystem has begun to recover, so overall it seems that fewer species were lost
When they removed post-disturbance categories, significant negative effect on richness
Take-away:
Richness is decreasing locally
Scientists should be careful about applying their findings on different scales
They mention in the introduction that diversity is increasing in some areas I would think abundances of a few species would increase and cause a decrease in diversity, could they be attracting species that benefit from humans? But this would increase local scale diversity and lower regional. I understand the issue they have with the previous paper and bias but based on what they are saying there just isn't available data to create an accurate representation of global diversity so as long as the authors are aware of the bias, producing a paper like that may still provide some type of benefit.
ReplyDeleteI agree with Alex. I understand that the authors disagree with the claims of the authors of the two papers that stated that their results were a global and truthful representation of the changes in diversity. But if you look at their data, of course you could find biases, as the data available does not come from studies specially design to address this topic at a large scale. I think this should be a wake up call to more studies on urban ecology and urban surveys of local - and large scale - diversity that could be used in this kind of analyses. And, as the authors say, to promote surveys and studies in less studied regions. I think this is something that could be improved by engaging citizens in surveys of local diversity.
ReplyDeleteI do think it's important to consider all potential biases in research, especially in research that has high reaching implications like decrease of biodiversity. However, I think the scale and question asked in each research needs to fully considered in such a review paper like this one. Lumping different research results into one study can mask more specific factors such as scale and resolution.
ReplyDeleteI think this was a very good companion paper. It was interesting to read about what other approaches have found on local biodiversity trends. I agree with the authors that there are serious problems with generalizing Dornelas et al. (2014) results as global biodiversity trends.
ReplyDeleteOf course no global study is truly perfectly global! Even the maps they used in fig 1 to represent global distributions are still made with interpolations and algorithms that estimate what they expect to be there. No one can go sample every square kilometer of the entire earth to produce those graphs. Of course the Dornelas paper was biased, every dataset is biased in some way. I felt they were being nitpicky and every one of their papers better not have any bias or lofty hypothesis at all if they’re gonna write about their colleagues like this. While they may have some valid points about the other papers, there are surely less aggressive ways to write about them.
ReplyDeleteI agree with Willow. This paper is kind of aggressive towards the other papers. Of course no set of data can be truly unbiased. I enjoyed reading the section "A need" where they talk about monitoring programs extending to underrepresented biomes.
ReplyDeleteScience is a process that relies on observation and data analysis. Often, researches use all the available data that suitable for the analysis considering the quality and quantity because it is almost impossible to have complete consensus count on every individual. Therefore, I agree with lots of points in this critique. Bias is everywhere and the quality depends on the question. A lot of statistical analysis considers the bias and errors unless the bias has a major structural deficiency. But same time, I understand those authors who did the assessment. Critique paper is great.
ReplyDeleteSomebody said it in class when they said its the broad papers that always spark further investigation and this is definitely proof since the author keeps referring to other papers. In one of the paragraphs the author mentions that one of the papers were wrong because their findings "confuses variables and spatial scales". Do the two not go hand and hand? Meaning variables are interchangeable and spatial scales are length or distance... so can't the spatial scales be interchangeable too?
ReplyDeleteEchoing what's been said in these comments as well as our discussion the other day- that a complete lack of bias seems pretty difficult to achieve. It was interesting to read up on biases that I hadn't heard of, and I think the take-home of "be more careful" is a good reminder.
ReplyDelete