Simplifying Description of Species

Finding Simpler Descriptions of Species

In order to understand the controls on ecosystem functioning and manage our biosphere, we must improve our understanding of how the detailed interactions among species in a community emerge in measures of community structure and ecosystem function (Levin 1992; Bazzaz 1993; Caldwell et al. 1993; Clark 1993; Jarvis 1993; Nicolis 1994). The enormous number of species in any ecosystem prevents any serious attempt to describe the fundamental actions and interactions of all species. Indeed, if we must understand the role of each species and all possible species interactions in order to understand ecosystem function, we will be unable to answer the pressing ecological challenges facing our science and our society (Nicolis 1994; Lubchenco 1995; Turner and O'Neill 1995).

One strategy to solve this problem is to group species interactions on the basis of broad functional responses, or guilds (Dawson and Chapin 1993; Körner 1993). This is a sensible approach because it attempts to maintain some of the information about species without losing the ability to make predictions with limited resources and time. Predicting ecosystem function on the basis of a functional or guild-based description of species is difficult for several reasons. First, the sensitivity of nonlinear systems to small differences suggests caution in applying these broader descriptions of species function and interactions to ecosystem prediction (Guckenheimer 1987; Roberts 1987; Hastings et al. 1993; Nicolis 1994). In addition, there are different criteria by which aggregation can be carried out, and hence, different ways to create functional groups (Roberts 1987; Dawson and Chapin 1993; Körner 1993; Solbrig 1993). Furthermore, keystone species can play unique roles in system function and thus cannot be lumped with other species (Bond 1993; Caldwell et al. 1993; Turner and O'Neill 1995). Even groups of species that perform similar functions may nonetheless have individualistic responses to environmental conditions that prevent treating them as a unit (Clark 1993; Rastetter and Shaver 1995). Here we explore how the development of a simplified description of species’ strategies influence community dynamics in this individual-based simulation model of forest dynamics.

In the SORTIE model, the functioning of each of the nine species is based on 10 parameters estimated from field data. For this work, principal components analysis (PCA) was used to search for simpler ways to describe the relation among the species (Harris 1985). The PCA method uses the multivariate correlations among species in parameter space to build a simplified description of the relationships. Ten principal components, or factors, can be estimated because SORTIE has a 10-parameter description of tree species. However, the motivation behind PCA is to find a lower dimensional description that explains most of the variance among species positions in parameter space. From the PCA, we found that two factors (components) could explain 69% of the total variance in species position. Both factors have strong, biologically meaningful interpretations that strengthen their utility in understanding species functional responses.

PCA factor loading graph

Species plotted in factor space defined by the first two factors. Tentative guild labels are meant to aid interpretation (detailed definitions of the factors below).

The first factor can be interpreted on the basis of the four strong correlations with the original parameters controlling mortality, light, extinction, and canopy depth. High values for Factor 1 are associated with trees having shallow, relatively transparent crowns. Further, these trees exhibit high mortality when suppressed. Low values for Factor 1 are associated with trees having deep, opaque crowns and low mortality when suppressed. Thus, this axis can be interpreted as "Shade Tolerance." Species with strongly negative scores on this axis are shade tolerant, and species with large positive values are shade intolerant.

The second factor is correlated with the two growth parameters as well as allometric variables that describe height as a function of stem diameter. The lower end of Factor 2 contains species that are ultimately very tall but show slow growth in height during establishment. These trees also exhibit fast maximal growth rates in full sun but slow growth in low light. Conversely, high values of Factor 2 are associated with shorter trees that extend quickly in height during establishment. These trees also have the low maximal growth rate but the highest relative growth rates at low light. Factor 2 is the tree’s "Growth Strategy." Low values on growth strategy are characteristic of light-demanding, early-successional species, whereas high values indicate trees that specialize on growing quickly in low to moderate light.

The two-factor PCA seems to capture the biologically meaningful differences among species. It seems a plausible step in creating guilds or other simplified measures of species function. We present simulations of the PCA-defined species in order to assess whether the two-factor representation of the species retains the qualitative and quantitative behaviors of the forest landscape in SORTIE.