Text by Lauri Laanisto
First of all the main references:
The original study – Fraser et al 2015 Science
My reanalysis of this study (co-authored by Mike Hutchings) – Laanisto & Hutchings 2015 Science
Rebuttal to our comment by the original authors (plus Meelis Pärtel) – Fraser et al 2015b Science
Accompanying blogpost in Meelis Pärtel´s work group´s blog – Is there a global unimodal relationship between species richness and productivity?
So, I got my first paper in a “big” journal. It´s another battle in the ongoing “hump!-no hump!” war in the already much blasted battleground of productivity-diversity relationship. Well, actually all these war-related comparisons are overstatements – so far it has been a quite civilized, yet intense scientific debate between members of a rather small group of plant ecologist. It´s just the venue that makes this thing look bigger than it is. This debate might suit much better on the pages of some plant or ecology journal (but that might just being me unaccustomed to read about species pool from big journals…).
First a little background. About the long-forgotten world that is the scientific literature published more than five years ago. I´m familiar with the productivity-diversity relationship debate since the beginning of my PhD studies (back in 2004), when we carried out a meta-study on this relationship (see Pärtel et al 2007 in Ecology; and also Laanisto et al 2008 Global Ecology and Biogeography). Back then we showed that the shape of the relationship depends on large-scale evolutionary and historical processes, like the shaping of the species pool in geological time-scale: tropical ecosystems have been rather stable for millions of years and that is why the productivity-diversity relationship there tends to be positive, not humped; while the temperate zone has been tampered by glaciation just 10 000 years ago and as it scraped away most of the soils and species, species pools on the high end of productivity have not yet recovered.
But then came the Adler paper (Adler et al 2011 Science) that reported, based on global meta-experiment, that there is no hump on global scale. Criticism was heavy. Grime himself was the co-author of one comment, although he stated in IAVS 2011 conference in Lyon that he now thinks that diversity and productivity are not really related to each other. Anyway. Adler paper had indeed many significant flaws. Especially for a study that claimed to be global – there were basically no datapoints from other climate zones than temperate zone, and about 75% of all the datapoints were in the USA. The other aspect that really bothered me in that study was the fact that richness of all the studied sites was compared in absolute values, and there is plenty of adequate literature suggesting that this is not the right thing to do (e.g. Gotelli & Colwell 2001 Ecol Letters). Plus, of course the old problem of invariable plot size – all the data from all the sites was collected using the same quadrat size, although for example in low productivity alvar grasslands 1x1m quadrat will host hundreds and hundreds of individuals, while quadrats of the same size in high-productive sites might just have couple of dozens of individuals (all that was already pointed out in Oksanen 1996 J of Ecology).
Couple of months after the IAVS 2013 in Tartu I received an email that asked if I would be interested in reviewing a manuscript that proposes new meta-experiment to study the productivity-diversity relationship in grassland habitats. It was more or less a reaction to Adler´s study – contra-meta-experiment if you will. Of course I said yes. (I sign my reviews so it´s not revealing a secret or anything.) Unfortunately the manuscript described a methodological framework that was only a tiny bit better than Adler´s study. I suggested the authors to take species pool into account and think about the plot size as well. And there was some other stuff as well. The editor agreed with my suggestion and the decision was major revision. I never heard about that manuscript again until I discovered it was published basically without any changes. So that became the Fraser et al 2013 JVS paper.
Based on the points covered in the review I wrote a comment to Fraser´s study, titled: “On the importance of ecological context when studying the productivity-diversity relationship on a global scale”. To prove my point, I also did some re-analysis on the Al-Mufti et al (including Grime) 1977 paper in Journal of Ecology which is kind of the first empirical proof of Grime´s hump, showing that the species pool of different habitats, which were sampled in this study, described species richness much better than the productivity. Here´s the graph with outtake from that manuscript:
By using the list of flowering plant species present in each sampling site I extracted the actual species pool for each site (see the Appendix in Al-Mufti et al. 1977). From the HBM graph (see Figure 13 in Al-Mufti et al. 1977) I registered the number of species found in the quadrats of each site. As expected, there was a strong positive correlation between the size of the species pool and the species richness in the sites (Fig. 1). Then I calculated the percentage of the species pool recorded in each site´s quadrats and tested the relationship between species pool adjusted diversity versus productivity of the sites. There was still a humped-backed relationship present (Fig. 2), but the data points in the graph were rearranged in comparison with their original placement (see Fig. 13 in Al-Mufti et al. 1977 – dot numbers in Fig. 1 and Fig. 2 are the same as in Al-Mufti et al.).
This simple analysis shows that the role of ecological context – in this case the size of actual species pool – is crucial when testing the relationship between diversity and an environmental variable. Although in this case it did not eliminate the HBM, it might change the productivity-diversity relationship shape in other cases. But without considering pertinent theories and alternative mechanisms the results of proposed global coordinated experiment might end up strongly biased, for example towards geographical location (Pärtel et al. 2007) or type of habitat (Cornwall & Grubb 2003) or community composition (Laanisto et al. 2008).
Figure 1. Relationship between the number of species found in sample quadrats (Species-density) and the total number of species found in each sampling site (actual species pool). Data for x-axis is extracted from Appendix and y-axis from Figure 13 from Al-Mufti et al. 1977.
Figure 2. Relationship (quartic polygon function) between species pool adjusted diversity (percentage of species pool represented in quadrats) versus productivity of the data points from Figure 13 from Al-Mufti et al. 1977.
Well, that manuscript was rejected as completely pointless and unconvincing. Both reviewers were highly negative.
Then the Fraser et al 2015 Science paper was published, and some of the data in was made public a month later in Dryad, when I was just doing some fieldwork in Svalbard. I had a day off before the departure and in the comfortable couch of Czech Polar Station in Longyearbyen I did similar reanalysis on the Fraser data. Then I convinced Mike to be co-author, and we more or less rewrote my original JVS manuscript to be suited for Science. And this time it was accepted.
Now, together with our comment, also the response to it was published (Fraser et al 2015b Science). This rebuttal claims that “that by calculating a more appropriate measure of species pool, community completeness, both regional and local processes shape local richness.” Couple of days later also a blogpost from Meelis Pärtel´s work group was published, which covers the same counterarguments. So here, in turn, are my quick counter-comments (to counter-comments of Fraser et al. 2015b Science to comments in Laanisto & Hutchings 2015 Science to study by Fraser et al 2015 Science (yeah!)).
1) “In their figure 1, Laanisto and Hutchings’s regression of local richness versus species pool is problematic because these two variables are inherently related, not independent, and local richness cannot exceed species pool size (10).” says Fraser et al rebuttal and this is also the main message in the related blogpost.
Well, our main point is that INDEED, these two variables are inherently related in the sense that local richness cannot exceed species pool size. This is why you cannot compare the absolute richness values from habitats with different species pools. Apples and oranges! And even if instead of this simple regression, log-ratios are used, which should make the variables independent, the result is basically the same positive relationship between local richness and species pool size (Fig 2B in Fraser et al 2015b Science).
2) I said “basically the same”. Actually what Fraser and others claim, is that because this relationship has a tiny curvature, it changes everything. The blogpost concludes that: “When the relationship between plot richness and local species pool was calculated using this method, a curved relationship was revealed, suggesting that plot richness increases less as local species pool increases. This in turn leads to the conclusion that species richness does decline at ever increasing levels of productivity.”
These conclusions I really don´t understand. How come the fact that the relationship between local species richness and species pool size has a slight curvature results in global hump? According to this Fig 2B the species number in 8x8m plot should be at least several thousand in order for that curve to become hump. But this is not really happening in our planet… not from the species density nor the productivity viewpoint. (And I don´t think we should start formulating ecological generalizations on a galactic level, when we cannot even get anything done on mere Earth level.)
Not to mention the issue with invariable plot size that I pointed out earlier (Oksanen 1998 J of Ecology paper titled “Is the humped relation between species richness and biomass an artefact due to plot size?”)
3) Moreover, both the rebuttal analysis by Fraser et al 2015b Science and accompanying blogpost omitted the two sites where only one single species was found from the 8x8m plot (which was the basis of calculating the species pool in this case). Rebuttal says: “…two monoculture sites were excluded, but these two points were outliers in any event.” I agree that these two sites are outliers. But then, why were they included in the initial analysis? Let me hypothesize why. These two sites were had the highest biomass levels. This means that the original analysis by Fraser et al 2015 Science (Figure 2A) had 128 data points (2 sites with 64 1x1m quadrat each) on the right end of the x-axis. And all of these 128 data points had the lowest species richness possible – 1! Considering that the r2 of this humped-backed curve is about 0.07, then I can only assume how persistent this relationship would be, if these “outliers” would have been excluded…
Yet, the accompanying blogpost says: “Laanisto and Hutchings provide an r2 value of 0.74 to describe the strength of the relationship, but how much confidence can we have in this strength? And is it appropriate to compare this r2 value with those generated by other regression models?” I could ask with similar rhetoric: How much confidence can we have in this original analysis (with r2 of 0.07), especially considering that it did not include species pool in any ways, but included very powerful outliers that clearly helped the curve to decline in the end of the x-axis?
I would really like to see that same original analysis (Figure 2A in Fraser et al 2015 Science) that omits all of these 128 outliers…
But. In conclusion I want to say that I do think that the humped-backed curve might be the most common shape between plant diversity and productivity/biomass. But in temperate zone. Also in polar regions. But probably not in tropics. If we really want to make a difference, we should maybe carry out a meta-experiment that is more radical and ecologically smarter. One really nice diversity-productivity paper that has not been getting the attention it deserves is by Liira and Zobel (2000 Oikos). They did three things that, when incorporated into such research framework, could really have a huge and in both ways positive impact: 1) they took into account the species pool of every site sampled; 2) they used dynamic spatial scale and sampled richness not per fixed area, but per fixed number of ramets; 3) they looked also below the ground and measured root biomass in addition to aboveground biomass.
And here is the full citation of my paper: Laanisto, Lauri, and Michael J. Hutchings. “Comment on “Worldwide evidence of a unimodal relationship between productivity and plant species richness”.” Science 350.6265 (2015): 1177-1177. (link to full text in ResearchGate)