Beyond species: why ecological interaction networks vary through space and time

Poisot, Stouffer, & Gravel. bioRxiv DOI: http://dx.doi.org/10.1101/001677. Beyond species: why ecological interaction networks vary through space and time

An interaction network based around populations (left) may not look at the same as one based around species (right). Species may sometimes interact with one-another, but that doesn’t mean that they always have to. Figure 1 in Poisot et al.


Lynsey McInnes

Lynsey McInnes

Welcome back to PEGE! We’ve taken a break over the summer, but we are hopefully back fully revived for the new season. We decided to pre-pick a bunch of papers so that we are not scrabbling for a suitable paper a day before the next post is due. This also means that you can follow along with us and pre-read our choices too. From the ten papers already chosen, you will see many familar themes….if nothing else PEGE is helping Will and I home in on what we care about.

And we start with a paper that falls nicely into one of my major preoccupations…population-level processes!

The paper reads like a call to arms by Poisot et al. for network, and other, researchers to get a grip and realise where the relevant processes are occurring. Namely, it is not sufficient to study interaction networks at the species level, but rather we must explicitly acknowledge that interactions differ across the extent of any network as a function of both local trait and abundance distributions (and any feedbacks among these). In short, species will not always interact where they cooccur if, for example, there is very few of one party or a species is not expressing a particular trait value. Sounds intuitive, right? But from my very vague knowledge of network literature and Poisot et al’s larger survery, it doesn’t sound like this variability is commonly addressed.

As you can imagine, I am totally behind the authors that populations are the appropriate scale at which to quantify network interactions. Perhaps because I am already behind them, I came away a little disappointed that they did not provide more discussion of how feasible their approach might be or provide real world examples.

For instance, they outline a model of when they expect interactions to occur and this, again, made intuitive sense. However, it would have been great if they had filled the model in with some data, even with simulated data. They highlight themselves that all the factors they have lumped into their error term (landscape, climate, other species, etc. etc.) will be most difficult to parameterise or indeed characterise. It would have been nicer to see a great attempt to approach this.

For instance, when might useful properties of a network emerge at the species level after all. Do plus and minus interactions often cancel out (so that the species level does kind of become the most useful level to operate at)? Are there clear clines in interaction strength (so that, for example, populations experiencing certain climates always (or often) interact)? Do populations of more than one species replace each other in certain interactions? Can a population-level model of interaction networks be used to study how perturbations (e.g. environmental change, habitat fragmentation, etc.) might affect the network?

I also wonder how genetics might help? Are there diagnostic signals of which populations are likely to interact or not? A threshold level of genetic diversity? Fixation of certain alleles? A certain age of population? A certain isolation of population? Are there useful ways to put genetics to use in characterising interaction networks? My feeling is this is already a field and I don’t know much about it….

Finally, I would have really liked the authors to provide a practical cheatsheet on what types of data to collect. How many populations to study, what traits to measure, how to measure them, how much uncertainty can we cope with, and so on. At least for one or two study systems with which they are familiar.

I recognise that the authors have probably thought about or even have answers to most of the above and not everything can be covered in one paper. Perhaps I was extra harsh because I already agreed with them and did not need convincing, so I was ready for their inevitable followup paper already. Hopefully the paper convinces waverers of the merits of population-level study, both because this is where the action takes place, but perhaps more importantly (at least if you are of the monitoring networks for their continued existence bent) because understanding at the level at which interactions occur will enable us to piece together how diversity pattern look or will probably look in the future.

As ever, exciting times!


Will Pearse

Will Pearse

Aaaaand we’re back! I’m glad Lynsey picked this paper, because I have a secret soft-spot for interaction networks. This is both an interesting suggestion for future work and a good (opinionated! hooray!) review of the state of the field. The general idea seems to be that we shouldn’t assume interaction networks are static: abundances and environmental factors determine what interacts with what, but (most interestingly to me), dependent on what other species are interacting with.

I like the idea, and the authors definitely sold me on it in their review. My only criticism (and I’m saying this to try and be helpful, since this is a pre-print) is that they really didn’t go into enough depth on the ‘what other species are interacting with’ (indirect interactions). The manuscript leads wonderfully up to the point where I’m really looking forward to the magical solution to indirect interactions and then… it stops. It’s a hard problem to solve, so I sympathise! Personally, since a lot of the interactions the authors describe as completely shifting the whole system are so major and difficult to a priori predict, I think just having the presence of certain keystone species as a factor that alters the probabilities of everything else is a fine way to handle it. Treat all the interction strengths as a statistical model (as they suggest, and as others have too), and then look for particular species that change almost everything. Maybe that’s not so mathematically pretty or flexible for smaller changes, and you’d either need some fancy machine learning or (shock! horror!) a priori hypotheses to find those keystone species, but I’d be happy with that.

The authors talk a lot about how interaction networks could ‘improve’ neutral models. A lot of (all?) neutral models are inherently about single trophic levels, so I’m not surprised that interaction networks would make big changes to them, but this did get me thinking a lot about some of the cool work that’s been done on neutral evolutionary models and their ‘non-neutral’ ecological outcomes. We absolutely need more models of interaction networks that take into account evolution; and I don’t just mean measures of phylogenetic dispersion (which are good), but integrated models where we simulate evolution under interactions. We can even fit models to phylogenies now, which is awesome, but maybe it’s time we went a little further.

Advertisements

About will.pearse
Ecology / evolutionary biologist

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s

%d bloggers like this: