It is a tangled bank up there

Energy flows through the community from basal species to top predators. We envision this flow using food chains and webs illustrating the directionality of energy flow from basal organisms to top predators (see the humorous fish food chain below).

Shows the flow of energy from the sun (not shown) to plants, then to herbivores, and two levels of carnivores.

Energy in the cartoon system above flows from the producers through a herbivore, and two predators. This simple food chain demonstrates one very useful organizational tool in the study of ecological communities; trophic levels. Above, we clearly see 4 trophic levels, (1) the plant, (2) the herbivore, (3) primary carnivore, and (4) secondary carnivore. I think that ecology has benefited from this simplification of trophic relationships, but how well does this simplification represent reality? Do organisms really separate into recognizable trophic levels?

In the simple fish system above each trophic level consumes only the fish from the level directly beneath it. When organisms consume species from multiple trophic levels, it is called omnivory. In the past there has been a large debate over whether or not omnivory is prevalent in natural communities. Some have said that omnivory destabilizes communities, others say that while it is more likely to be destabilizing, int he instances when it is found in real communities, it tends to have a stabilizing impact. All I want to say about stability here is that yes, it does appear that omnivory reduces the probability of a system being stable (maybe more on that later) it tends to reduce variability in the return time of the system.

More important I think is the question of the prevalence of omnivorous interactions. One paper that I think is really fascinating on this topic is “Trophic levels and trophic tangles: the prevalence of omnivory in real food webs,” a paper in Ecology by Thompson et al. (2007). The basic finding is that most species cannot be assigned to an integer trophic level (1, 2, 3, 4, etc). Furthermore, those that could be assigned an integer trophic level were primarily found in the first and second levels (plants and herbivores). What this means is that as you go up the food chain higher level species are more likely to be omnivorous. This makes sense, in that the higher a species level is, the more levels it has to choose prey from. Moreover,  higher trophic level species (much as in our cartoon food chain) tend to be larger than their lower trophic level food. When you are larger, you can basically eat whatever you want to (at least morphologically, e.g. their gape size is large enough).

Now, I want to demonstrate this finding that omnivory increases in prevalence higher up the food chain. Here is the food web from Otago Harbor, NZ (which I have shown previously) plotted according to trophic level (y-axis). I should note that trophic level was calculated with the NetIndices package in R.

otagopredweb1

Predator-prey interactions in Otago Harbor, NZ

In the picture you can clearly see a defined first and second trophic level for the plants and herbivores. Once you look above the second level, it becomes MUCH more difficult to distinguish between different levels. It truly represents a ” tangled bank” of consumers.

When trophic level is calculated with the NetIndices package, it also calculates an index of omnivory as well. We can then look at the correlation between trophic level and omnivory.

predcor

Correlation between trophic level and omnivory index for five well resolved food webs.

The top left plot is for the Otago Harbor food web pictured above. Up next is 22 webs from the Interaction Web Database.

iwebcor

Correlation between trophic level and omnivory index for 22 food webs from the Interaction Web Database.

So there is a clear positive correlation between trophic level and omnivory index that appears to be a general pattern. I won’t show it here, but the correlation coefficients are all significant for the webs I have shown you (although I admit I did not do those statistics properly, without correcting for multiple comparisons). Of course, this just makes logical sense, because there are more opportunities for omnivorous interactions to occur the higher up in the food chain you are.

Nonetheless I think that this shows that when considering energy flow through communities it is important to realize that the upper trophic levels are a tangled bank rather than separate levels.

Advertisements
This entry was posted in Musings, Research and tagged , , , , , . Bookmark the permalink.

6 Responses to It is a tangled bank up there

  1. swertie says:

    Very interesting. It would be great to make something similar for parasitoids!

    • Jon Borrelli says:

      Thanks! I agree, and actually for the 5 food webs in the first set of correlation plots I have parasite data. If I am remembering correctly it showed roughly the same relationship, except with a higher number of trophic levels.

  2. Mike Fowler says:

    Hi Jon,
    This is definitely an interesting general approach, but there’s a couple of issues that might be worth discussing further:
    (a) Is it sensible to include species in the lowest (or 2 lowest) trophic level(s)? i.e., those species with a trophic index ≤2? This imposes quite a strong constraint on your analysis that almost certainly affects the result. There are more species in the lowest trophic levels, which will ‘drag’ the correlation relationship in a specific direction.
    (b) As there are more species in lower trophic levels that can’t be omnivores, they possibly shouldn’t be included in the analysis, as they (arguably) represent a different class of organism from those that can be omnivores.
    (c) Is a linear relationship the best assumption to make about the relationship between trophic level and omnivory? How is the omnivory index derived? This might have an impact on the appropriateness of that assumption.

    It’s all interesting stuff! Keep up the good work.

    • Jon Borrelli says:

      You are absolutely correct Mike. One of my first thoughts when I was looking at the correlation plots was that having levels 1 and 2 would probably bias the results for the same reason you state in (b), they by definition cannot be omnivores. So maybe the way to think about omnivory is the way it was represented in the trophic tangle paper, simply that once you go higher than level two it becomes a free for all. I do wonder though, if the correlation holds if I were to remove levels 1 and 2, I will have to do that so keep an eye out for the results.

      I have to say you effectively called me out on all the things I was sweeping under the rug in this post (hooray science!). I had actually had no thoughts on whether a linear or nonlinear relationship may be best, although now that I think about it I cannot see any reason why it couldn’t be nonlinear. The omnivory index has been a black box for me so far, I would really like to know how it works. Maybe once I figure it out I will do a follow up post.

      Thanks again for the comments!

    • Jon Borrelli says:

      A quick look at the vignette later and it seems as if the omnivory index is a measure of the variation in the trophic levels of a predator’s prey. Trophic level is calculated as the weighted average of prey trophic level +1.

      • Mike Fowler says:

        Cool – that clears things up a bit, thanks.

        Another thought about including the TL ≤ 2 data comes from a purely statistical point of view. Those values (and the associated OL = 0 value) are, and can only be, discrete (integer) values, which means Pearson’s correlation is not an appropriate tool for their analysis. Mixing discrete and continuous values (when TL > 2) in the analysis of the whole data set is probably a good enough argument against their inclusion.

        I do wonder though, if the correlation holds if I were to remove levels 1 and 2

        Doesn’t matter if it holds or not – as long as the form of your data and statistical assumptions match, you’ll get an interesting result. If removing inappropriate data shows that there is no relationship between TL and OL, that’s still an interesting (and surprising based on ‘intuition’) outcome for me. Looking forward to the new results 🙂

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