Pattern and process—Challenges in ecological data analysis

Multivariate statistical (or data analytic) techniques have long been descriptively used, especially in vegetation ecology, to cluster ecological sites into distinct communities (classification) or position them along continuous axes (ordination).  The patterns exposed have also been used to generate hypotheses about causal factors or underlying environmental gradients.  The results of pattern analyses, however, are sensitive to the models underlying the technique used and the sampling sites from the space of environmental possibilities (Faith, Minchin, and Belbin 1987; Minchin 1987).  Popular techniques, such as principal components analysis and detrended correspondence analysis, when tested on simulated data, do not recover well the simulated environmental gradients.  Techniques that reduce this model-dependence also tend to produce degenerate patterns (Faith, Minchin, and Belbin 1987).  The Catch-22 is that one needs to know a lot about the causal factors behind the data in order to design efficient and distortion-free multivariate techniques that would expose those factors (Austin 1980, 1987).

To some extent such problems can be overcome through the use of analysis of variance and related statistical techniques on data from replicated, multi-factorial field experiments (Underwood 1997).  Strictly speaking, however, such results are local, that is, contingent on the configuration of other factors held experimentally or statistically constant for the experiment (Lewontin 1974).  Localization poses few problems when ecological engineering affords control over conditions and isolates the system from any surrounding dynamics.  But these are special cases; in naturally variable situations, observations constructed for testing of specific, single-factor hypotheses may not be useful for thinking about anything beyond the local configuration observed.  Similarly, lack of generality (Kelt and Brown 1999, 99) seems also to be the case for assembly rules in community ecology (Weiher and Keddy 1999).  In the absence of information about historical trajectories, assembly rules are better thought of as patterns of co-occurrence that are statistically significantly different from patterms that are produced by randomly sampling (“assembling”) species from the appropriately delimited species pool (Kelt and Brown 1999).

Lying between local results and pattern analysis is the pragmatic use of statistical techniques—if the statistical analysis shows a significant pattern then ecologists hope it is worth their while to investigate what could be causing this pattern.  The risks, however, are that statistical terms—effect, source of variation, etc.—are read in causal terms and that alternative categories for data collection are not considered.

These same issues about pattern and process apply to landscape ecology (Klopateck and Gardner 1999) and geography (Olsson 2002), including studies that make extensive use of remote sensing and Geographic Information Systems (see papers in Turner and Taylor 2003 for critical analyses by practitioners).

Extracted from Taylor, P.J. (2005) Unruly Complexity: Ecology, Interpretation, Engagement (U. Chicago Press).


Austin, M.P.: 1980, ‘Searching for a model for use in vegetation analysis’,  Vegetatio 42, 11-21.

Austin, M. P. (1987). “Models for the analysis of species’ response to environmental gradients.” Vegetatio 69: 35-45.

Faith, D. P., P. R. Minchin and L. Belbin (1987). “Compositional dissimilarity as a robust measure of ecological distance.” Vegetatio 69: 57-68.

Kelt, D. A. and J. H. Brown (1999). “Community structure and assembly rules: Confronting conceptual and statistical issues with data on desert rodents,” in E. Weiher and P. Keddy (Eds.), Ecological Assembly Rules: Perspectives, Advances, Retreats.  Cambridge: Cambridge University Press, 75-107.

Klopateck, J. M. and R. H. Gardner (Eds.) (1999). Landscape Ecological Analysis: Issues and Applications. New York: Springer-Verlag.

Lewontin, R. C. (1974). “The analysis of variance and the analysis of causes.” American Journal of Human Genetics 26: 400-411.

Minchin, P. R. (1987). “An evaluation of the relative robustness of techniques for ecological ordination.” Vegetatio 69: 89-107.

Olsson, G. (2002). “Glimpses,” in P. Gould and F. R. Pitts (Eds.), Geographical Voices: Fourteen Autobiographical Essays.  Syracuse, N.Y.: Syracuse University Press, 237-268.

Turner ,M. and P. J. Taylor (Eds.) (2003). “Critical reflections on the use of remote sensing and GIS technologies in human ecological research.” Human Ecology 31(2): 179-306.

Underwood, A. J. (1997). Experiments in Ecology: Their Logical Design and Interpretation Using Analysis of Variance. Cambridge: Cambridge University Press.

Weiher, E. and P. Keddy (Ed.) (1999). Ecological Assembly Rules: Perspectives, Advances, Retreats. Cambridge: Cambridge University Press.


Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your 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