• Clint Leach

    Clint Leach
    Harvey Mudd College
    Investigated the structural properties of the Gulf of Alaska, Aleutian Islands, and Bering Sea food webs.




    The ecosystems of the North Pacific are incredibly complex, yet much can be learned by studying the topological properties of their food webs. In my internship, funded by the CAMEO (Comparative Analysis of Marine Ecosystems Organization) project, I investigated the structural properties of the Gulf of Alaska, Aleutian Islands, and Bering Sea food webs. Specifically, I tried to determine whether or not these food webs could be broken apart into densely connected subunits, referred to as communities.

    I began my investigation by first assembling the food webs themselves using data on predator stomach contents collected over almost thirty years. Each species in the database was represented as a node, and two nodes were linked if one species had ever been recorded in the stomach of the other. However, because species-level identification was not always possible for prey found within predator stomachs, many of the prey nodes were general categories (i.e. "Euphausiids"). These nodes tended to be highly connected, and as such, they tended to wash out any structure that would have been observed in the food web network. Because of this, these general nodes were removed, and the food webs were constructed using only prey that had been identified to at least the family level.

  • Figure 1

    Figure 1
    Aleutian Islands Food Web. Circles represent different communities within the food web. Colored links are within community links, gray links are between communities.

    Figure 2

    Figure 2
    Gulf of Alaska Food Web. Circles represent different communities within the food web. Colored links are within community links, gray links are between communities.



    Once the food webs were constructed, I implemented a community search algorithm using the "igraph" package in R. This showed that the Aleutian Islands food web broke into 14 different communities, most of which were organized around a central predator (Pacific cod, pollock, halibut, arrowtooth flounder, etc., Figure 1). The Gulf of Alaska broke into four communities, but with most species falling into two dominant communities (Figure 2). The largest of the three food webs, the Bering Sea would not break apart into smaller communities at all (Figure 3).



  • Figure 3

    Figure 3
    Bering Sea Food Web.



    While this seems to suggest that the Bering Sea is the most tightly knit ecosystem and the Aleutian Islands is the most fragmented, it is important to note that there is a much greater sampling effort for the Bering Sea than for the other regions. To test for the effects of sampling effort, a random subsample of the Bering Sea stomachs was chosen such that the sample size per predator was equal to the Gulf of Alaska. Assembling the network from this data, we saw that the Bering Sea food web now broke into many different communities, resembling the structure of the Aleutian Islands. Thus, it is likely that much of the observed differences in community structure are a result of differences in sampling effort.

    Despite this result, there is still much to be done. The next step will be to slowly increase the sample size used to subsample the Bering Sea data to determine at what sample sizes the food web begins to knit together. In addition, I plan to quantify link strengths and add this information to the community search algorithm. Weighting the links by the frequency of their occurrence will diminish the influence of rare prey species that may be transgressing and linking separate communities; the new linkage technique will allow a truer underlying community structure to become transparent.