I’ve been meaning to write about this for a while – last autumn, Nathan Kraft and colleagues published a very interesting article in Science, in which they use two dataset on woody plant richness across an elevational and a latitudinal gradient to assess whether there is an effect of latitude and elevation on beta diversity. The paper was recently complemented by two replies (1,2), together with the corresponding response.
Before dealing with the ecological conclusions, some notes on the method: while the traditional “reflex” would probably have been to use a regression approach, Kraft et al. use a null model, shuffling individuals around subplots while preserving alpha and gamma diversity as well as total local abundance to see whether the observed beta-diversity deviates from the null expectation (which seems to yield different results than a regression, as the replies and responses show). People have done similar things before and are increasingly doing so, but still, this underlines the potential of going away from the more traditional regression methods towards null model / simulation based methods when it comes to complex statistical questions (see also our review here), and I think that there are still a lot of questions left where a clever analysis like this could produce quite interesting and new results, even with old data.
In terms of results, the main conclusion is that there is no systematic effect of latitude and elevation on beta diversity when the effect of gamma diversity (which is correlated with these variables) is accounted for by the null model (however, responding to the reply by Qian et al., the authors conceit that there is one when the data is confined to South American plots south of the equator, but also note that the significance of such an effect in a non-random selection of the data should be treated carefully). I don’t want to go in all the details of the replies and the counterarguments here (they’re interesting though, and they make a lot of good points). I agree that the data, although impressive in size, is probably not general enough to be sure that there is not such relationship in general (other taxa, and particularly other scales), nor to be sure that you couldn’t find effects in particular subsets of the data. However, my personal impression was that there is still convincing evidence that the effect of latitude and elevation on beta-diversity is at least not generally present at the scale that was examined here, and, as a side remark, that one can come to very different conclusions when different statistical analyses are applied.
So, what do we learn from that about community assembly? What was also shown in the study is that there is clearly some nonrandom effect on community assembly, as beta diversity was generally higher than expected under the null model. The exact reason for this, however, is uncertain; the authors invoke the usual suspects such as habitat filtering, dispersal limitation, etc. Yet, it would probably be premature to conclude from the fact that this non-random effect was more or less constant that also those mechanisms of community assempbly are constant across the examined gradients – processes could go in different directions and cancel each other out. The authors conclude very carefully that
the net outcome of local community assembly processes is consistent (in terms of their effect on b diversity) across these gradients (33) at the scale of our study.
Whether this is because the processes of community assembly processes are constant or not seems to be open – at least, there is no direct proof of a systematic dependency of local community assembly on latitude and elevation from this study, which is also noted by the authors when they state little later that
Therefore, there may be no need to invoke different local assembly processes when trying to explain latitudinal or elevational differences in b diversity.
Yet, I have to say that I’m still a bit skeptical about whether this will hold in general, I find it actually hard to imagine that there is no effect on the mechanism of local community assembly whatsoever across a latitudinal gradient, given the different ecological strategies and evolutionary history that is found in the tropical tree species vs. their temperate relatives. Also, it should be noted that the spatial scale at which beta diversity was defined in this study (heterogeneity between ten 0.01 ha subplots on a plot) was very small, and there may be effects at larger scales. The authors remark that this small scale is sensitive to fine-scaled environmental heterogeneity and local interactions, but might miss coarser-grained environmental effects. An analysis of beta diversity at larger scales would certainly be interesting. In fact, it would probably be more precise to drop the term beta diversity altogether in favor of an expression that includes the scale that was examined such “compositional turnover at the 10 m scale” to discussing results like this, but that has been said in vain by many before, beta-diversity simply seems to catchy a word to give it up.