Contextual Analysis and Community Ecology

OK, last, or maybe next to last essay on the use of contextual analysis.  The reason I say maybe next to last is that I strongly suspect that a lot of people have data sets that could be used for contextual analysis.  Thus, I would suggest that without collecting any more data we could greatly increase our examples of group selection in nature.  If I can find the right data set I will show you what I mean.   This week, however, I want to look at an entirely different use for contextual analysis, and in the process perhaps point out the surprisingly tight linkage between population genetics and community ecology.

In a study published in 1999 Jane Molofsky and her graduate student, Shannon Morrison and I published a paper on the invasability of Reed Canary Grass (Phalaris arundinaceae) (Molofsky, Morrison, and Goodnight1999 Biological Invasions 1:181).  Reed Canary Grass (RC grass) is an invasive grass species that was originally introduced as a forage crop, and to stabilize areas prone to erosion (we never will learn, will we).  Since then it has become a problem, and is now classified as a pest species.  The interesting thing is that this pest does not always invade an area, so the question becomes why  does it invade some times and not others?  We figured there were two possibilities to consider.  First it might well be that there are some strains that are more invasive than others, and second it may be that some aspects of the environment or the community may affect the ability of the plant to invade.  It would be nice to separate these apart, and because RC grass can reproduce clonally it is something that is experimentally feasible.  That is, we can plant out clonal replicates into the field.  Of course the problem with this is that a natural field cannot be standardized, and these differences in community and environment will be confounded with any genetic effects we see.

Reed Canary Grass

Reed Canary Grass, Phalaris arundinaceae (photo from

 To test this we (OK, I will be honest, Shannon) chose three clones of RC grass that Shannon knew were distinct, and did two things.  First she took 50 ramets from each of the three clones and planted them out in 150 randomly chosen locations in a flat section of pasture in Jericho Vermont.  She allowed one week for them to establish, then randomly selected 30 individuals from each surviving clone.  Around each clone she measured the community, both for the area covered by each species, and measures for diversity and species richness around each plant.  This gave us the community measures.   At the end of the season the plants were all harvested, whether or not they were alive, tiller number, and above ground and below ground dry weight were measured.

For the genetic measures she took 10 ramets of each clone and grew the in a greenhouse individually in large pots under uniform conditions.  The large pots were chosen because they were large enough to allow the plants unrestricted growth.  After one month the plants were measured for the number of tillers they had produced and the above and below ground biomass (dry weight).

Note the important point here.  We have two very disparate types of data.  From the field we have “fitness” measures, growth, and survivorship in a natural competitive environment, and measures of the species and diversity in the community with which they are interacting.  From the greenhouse we have the growth the plants in the absence of competition.  To combine these we did a contextual analysis suing the “fitness” traits (really measures of their ability to invade), and used the community and greenhouse data as independent variables.  We fully recognized that growth in the greenhouse was not a measure their performance in nature, however, our thought was that growth rate in the greenhouse would be correlated with characteristics that were important in a competitive environment.  If that was the case then the greenhouse measurements should be predictive of growth in nature.

There were a lot of parameters measured, percent cover  of 11 different species, species richness and community diversity, soil moisture, and the green house measures (tiller number, above and below ground biomass), so we chose to use a stepwise regression to select a subset for the analysis.  This is always a problem:  We did not have an apriori idea of what would be important and therefore measured as much as possible.  This will inevitably be a bit “post hoc”, but then that is often the nature of field work.

In any case, on to the results.  First we found that the survival of the different clones was entirely do to clonal differences, and apparently unaffected by the community or environmental factors we measured.

Molofsky figure 1

(Molofsky, Morrison, and Goodnight1999 Biological Invasions 1:181)

This difference in field survivorship was significant in a logistic regression, with tillering rate being the best predictor of survivorship.

Molofsky CA table 2


(Molofsky, Morrison, and Goodnight1999 Biological Invasions 1:181)


The other analyses become more complex.  For your interest I will put in the analysis:

Molofsky CA table 3

(Molofsky, Morrison, and Goodnight1999 Biological Invasions 1:181)

but I think a verbal description is more useful.  Basically we found that above ground biomass is primarily a function of the community, and particularly the percent cover of Anthozanthum odoratum.  The greenhouse measured traits having almost no effect on above ground biomass.  In contrast the below ground biomass was a function of both clonal characteristics (tillering rate) and community characteristics (cover of A odoratum and species richness).  Finally the above ground to below ground biomass was a function primarily of the community charactistics and the soil moisture, with a possible, but non-significant effect of tiller number.

In short, the ability to survive appeared to be primarily a function of the genetic characteristics of the plant, and particularly its tillering rate, whereas the competitive ability, and a particular plants ability to establish and spread in the community appeared to be a function of both the plants intrinsic characteristics and the specifics of the community and environment in which it was placed.

This study is rather far afield from the emphasis of this blog on evolutionary theory, and yet it really isn’t.  It shows that the basic concepts of contextual analysis can be applied in a wide range of situations.  In this study the greenhouse and field data are really incompatible and cannot be combined using traditional methods, and yet with contextual analysis they can be combined.  The important caveat, of course, is that while the significance levels are useful, the actual values the regression coefficients take on are perhaps less useful because of the very different nature of the greenhouse data.

The second point that this shows is that community ecology and multilevel selection are really not that different.  Because of the experimental design, in this situation we know, that the  “heritability” of the community structure is zero, so it makes sense to interpret this using an community ecology framework.  However, it does not seem that far fetched that a similar study could be interpreted in a community selection framework instead.

As a final note, I will also mention that this similarity is not unique to contextual analysis.  For example, the community diversity measure used in this study is Hurlburt’s PIE.  It turns out that this is mathematically identical to the formula used to calculate F, the inbreeding coefficient.

Leave a Reply