A paper came out today in Ecography titled: “Historical climate-change influences modularity and nestedness of pollination networks.” The authors were interested in how past climate change (over the Quaternary) impacted the structure of plant-pollinator networks observed today. Specifically they were interested in modularity and nestedness, which destabilize and stabilize pollinator networks respectively (as shown by Thebault and Fontaine 2010).
Their method was based on four steps; (1) generate the network, (2) get network properties (in this case modularity and nestedness), (3) for each network location obtain estimates for historical climate change and contemporary climatic variables, and (4) regression analysis to determine the relationship between climate and network structure. To measure modularity the authors used the simulated annealing algorithm of Guimerà and Amaral 2005. For nestedness they used the popular NODF measure. While they were primarily interested in relative values (as opposed to significance, meaning they did not use null models) they did test for effects of sampling errors/bias. I thought their treatment in this respect was very good.
As an added twist Dalsgaard et al. separated networks depending on whether they were from mainland or island localities. As far as their estimation of historical climate change I will admit that my unfamiliarity with the literature betrays me, and I do not think I could speak well about whether it was a particularly good or poor approximation. It certainly sounds reasonable to me as I read it. They then composed a number of potential models using ordinary lest squares and spatial eigenvector mapping with various explanatory variables (climate change velocity, etc.) and selected the best models using AIC.
I will let the authors tell you their results. In brief:
We found a relatively strong negative association between Quaternary climate-change and modularity on the main- land, whereas Quaternary climate-change had less effect on modularity on islands. Nestedness was on the mainland most pronounced in areas having experienced high Quaternary climate-change, while on islands Quaternary climate- change had no effects on nestedness.
So, to reiterate, strong effect on mainland compared to islands and more historical climate change leads to less modular and more nested architectures. I find this result a fascinating, demonstration of how knowledge of stability properties of various architectures/structures may inform our understanding of community structuring.
Let me explain. Instead of climate change, let us call it simply a perturbation. We know from prior theoretical studies that some network structures are more stable than others. In the case of pollination networks, Thebault and Fontaine showed us that nestedness promotes the stability of pollination networks, while modular networks are not stable. A more stable network is more likely to resist modification following a perturbation. Thus we may expect that with a changing climate (perturbation) those structures that are more stabilizing are more likely to persist, while structures that are less stable are more likely to be wiped out. Given that logic, we would expect those pollinator networks that are more modular to be filtered out by large perturbations such as climate change. Similarly, we expect pollinator networks that are more nested to persist over longer periods of time and thus are more likely to be observed.
I posted this, but forgot to write about future implications.
I agree with Dalsgaard et al. in that this method, being as relatively simple and straightforward as it is, could be applied to more than just plant-pollinator networks. The obvious next step is food webs, since we have as much knowledge about the stability of food web structures as mutualistic ones, and possibly a much richer database (we have been studying food webs a bit longer than mutualistic networks). To take it one step futher, we could begin to analyze multiple patterns aside from the obvious modularity and nestedness. For instance, connectance and number of species explains almost all the variation in food webs (Vermaat et al. 2009), has climate change influenced these network properties as well?
Also, I believe the next logical step is prediction. We have current networks, we have a rough understanding of the relationship between climate and structure, and we have (many) predictions of how climate will change in the next 50-100 years. Can we take all this information we possess and make inference about how network structure will be impacted in the near future? I think we can. This could be important for conservation, community stability, etc.