More than just a coincidence? What the co-occurrence of species can teach us about how they interact

Freilich, M., E. Wieters, B. R. Broitman, P. A. Marquet, and S. A. Navarrete (2018).  “Species co-occurrence networks: can they reveal trophic and non-trophic interactions in ecological communities?”  Ecology, 99.


Different kinds of plants, animals, and fungi interact with each other in a myriad of ways.  Recently, researchers have been trying to infer the nature of these interactions just by looking at whether you can find these species in the same place!  In a 2018 study, Mara Freilich of the Massachusetts Institute of Technology and her co-authors examined the reliability of this co-occurrence approach.


What is a co-occurrence network?

A co-occurrence network is a tool for visualizing the relationships between different members of a community.  Each member of the community is represented by a dot, and if any two members can be found in the same place, then you draw a line in between them.  You might’ve seen similar strategies used to visualize social networks!

An example of how your friend network might appear on Facebook.  Image credit:

In their recent study, Mara Freilich and co-authors investigated how co-occurrence networks can be used in the context of ecology.  In an ecological co-occurrence network, each dot represents a certain species, and a line between any two dots indicates whether you can find the two species in the same place, at the same time.  Co-occurrence networks are relatively easy to construct, so Freilich and her co-authors wanted to see whether the co-occurrence method could provide reliable information about how species interact.


How do different species interact?

The interactions between species are often split into trophic, and non-trophic, interactions.  Trophic interactions, also known as food webs, give you information on which species eat which other species.  If you were to draw a co-occurrence network for Chesapeake Bay waterbirds (see food web below), you might draw a line between bald eagles and large piscivorous (ie, fish-eating) fish, since the eagles will probably be found in the same place as their food!

Image credit: Wikimedia Commons/Matthew C. Perry.

Species can also interact with each other in non-trophic ways, which are more subtle.  You might’ve heard of the mutualistic interaction between sharks and pilot fish, in which the pilot fish eat parasites that irritate the shark.  This is an example of a positive interaction because both species benefit (as opposed to a negative interaction, like competition, in which the presence of one species negatively impacts another).  Since you frequently find sharks and pilot fish together, you would draw a line between these species in a co-occurrence network!

Shark and pilot fish.  Image credit: Wikimedia Commons.


Why use co-occurrence networks?

So why use co-occurrence networks?  In short, because it’s a lot easier to construct a co-occurrence network than it is to figure out how all the species interact with each other.

To recap, we talked about 1) co-occurrence networks, which give you information on whether you can find two species in the same place, and 2) species interactions, which can be trophic or non-trophic.  In the previous examples, we started with species interactions and then thought about how those interactions would look in a co-occurrence network. But most of the time, we don’t know how different species interact!  It can be extremely challenging to get data on every species interaction in an ecosystem, so Freilich and her co-authors want to understand if you can go in the other direction.  That is, if you start with a co-occurrence network, can you use it to figure out how species interact?


Can you infer species interactions from a co-occurrence network?
The ecosystem examined in this study is the rocky shore of Las Cruces, Chile.  Image credit: Mara Freilich.

The researchers examined the ecosystem of the rocky shores of central Chile.  Freilich and her colleagues first used co-occurrence networks to infer species interactions, and then they compared their results with past studies on species interactions in this same ecosystem to check if the co-occurrence network approach agrees with direct data.  They found that the co-occurrence approach is not always reliable. It works best for positive non-trophic interactions (such as the Lessonia kelp and a species of grazing limpet, which lives on it), but it’s a lot less reliable if multiple species interact very strongly with a third party, such as another species or some aspect of the environment.

Lessonia kelp (right) and limpets (left).  Image credits: Seafriends and Ferrebeekeeper.

I was very lucky to have a chance to talk to Mara Freilich about her study!  Our conversation has been condensed and lightly edited for clarity.


Interview with Mara Freilich

Why is it important to know how species interact?
“It’s important to know how populations are going to change.   We need to be able to answer questions like, ‘Well, if something wipes out the snails, how will that affect the rest of the community?’

“This can be important for humans as well.  We talk in the study about a few species that are harvested by artisanal fisheries, because these fisheries are hard to regulate, but they’re very important for people’s livelihoods.  It’s important to understand not only the impacts that the artisanal fisheries have on the environment, but also the potential impacts that changes in the environment could have on people’s livelihoods.”

Crab in Concon, Chile.  Image credit: Mara Freilich.


What are some reasons that the co-occurrence method might not be reliable?
“I think there are a bunch of problems.  One is that two species might co-occur with a third species or something else that’s outside the study system entirely.  In our study, we found that a lot of species depend really strongly on this algae. From the co-occurrence networks, it looks like those species interact with each other, but they don’t; they just depend on this algae.

“More theoretically, even if there isn’t some deterministic interaction, you might still observe noise that’s correlated between species, and so they might appear to be correlated using this covariance method.  We’re actually working right now on a second study which outlines all the theoretical reasons why there are problems with the co-occurrence method!”

Concon, Chile.  Image credit: Mara Freilich.


What do you think people should take away from this study?
“We need to have better collaboration between mathematicians and ecologists.  This project came out of a study group that was trying to apply methods from quantum mechanics to ecology.  There isn’t usually a lot of overlap between these two groups! The campuses were even far apart from each other because people don’t expect these two groups to be collaborating regularly.

“But we should really make an effort to make this collaboration happen.  For example, economics is similar to ecology in terms of the types of questions that are being asked, and there’s a lot of really heavy mathematics that are used in economics.  If there were more collaboration, mutual respect, and cross-talk between these two disciplines, I think we could make a lot of advances!”

Share this:

Rohini Shivamoggi

I'm a PhD student studying atmospheric sciences at MIT. I study the formation of secondary eyewalls in hurricanes, which hopefully will help us improve our forecasts of hurricane intensity. Before I got to MIT, I grew up in Florida and studied Chemistry and Physics at Harvard University. My other interests include weather forecasting, photography, and encouraging diversity in STEM! You can find me on Twitter @RShivamoggi.

Leave a Reply