This blog post is provided by Paulina A. Arancibia and Peter J. Morin and tells the #StoryBehindThePaper for the paper “Network topology and patch connectivity affect dynamics in experimental and model metapopulations”, which was recently published in the Journal of Animal Ecology.

Global change has increased the rate at which habitats are fragmented, increasing the creation of spatially discontinuous populations linked by migration (metapopulations). The location of a favorable habitat patch within the surrounding unsuitable habitat matrix can influence the quality of its resources. However, its connections to other patches within the network can dictate how “accessible” or “isolated” a patch can be simply by the number of its direct or indirect neighbors. These patterns of connectivity reflect the topology of those networks. This “architecture” of a network describes the spatial arrangement of its patches, dictates the level of movement in and out of them, and describes the distribution of the proportion of patches with different levels of connectivity in the network.

Given the recent advances in the fields of ecological design and restoration, ecologists now have the ability to create, modify, or protect particular habitats. Understanding how these very abstract notions can be applied to spatial networks therefore becomes relevant for the conservation and management of fragmented habitats. However, the large scale of natural patchy or fragmented systems makes their study very expensive and often impractical.

To circumvent these constraints, we created our own experimentally tractable micro-landscapes that differed in network topology. We used different network layouts to compare two types of network configurations representing the two ends of the biologically plausible spectrum of spatial networks: random and scale free. On one end, a random network represents a system where the connections between patches are random, leading to a structure where most patches have a similar/average number of connections and only a few are either sparsely or highly connected. On the other end of the topological continuum, a scale free network represents a system where most patches are very isolated (with few connections) and only a handful are very highly connected. We used populations of the aquatic ciliate Paramecium tetraurelia in our experimental landscapes, which we constructed from two 24-well plates for each replicate. Each well (filled with bacterized protist media) corresponded to one isolated habitat patch that we connected to a subset of others using capillary tubes through which protists could swim freely and move to occupy other patches.

Three times a week, protists in each patch were counted with the aid of a microscope. The large amount of time required to process the number of replicates necessitated splitting the experiment into two temporally non-overlapping blocks. Abundance and occupation patterns were followed for roughly 16 protist generations (3 weeks). Even at this “small” scale, experiments replicating more than two network configurations for each of the random and scale-free topologies were not feasible, so we also developed a simple patch occupancy model to simulate dynamics over at least 200 different network layouts per topology and compared those results to our experimental data.

After many hours spent counting protists at the microscope, our data showed that the pattern of connectivity of the spatial network of habitats can affect both occupancy (presence/absence) and abundance in these metapopulations. In scenarios where colonization and extinction rates are relatively low, randomly connected metapopulations performed better as determined by higher abundance in patches as well as a higher proportion of occupied patches. In contrast, in scenarios where extinction and/or colonization rates were higher, the effects of topology were indistinguishable and both networks performed similarly. The logistics involved in both lab-based experiments and in field studies greatly limit the size and scope of empirical research. However, experiments such as this can provide a reality-check for the extensive literature based on mathematical models that so far has addressed the issue of the role of network structure in metapopulation dynamics.