Appendix B. Additional details on the methods used to analyze changes in zooplankton community composition through time.
To visualize changes in cladoceran community composition through time, we used non-metric multi-dimensional scaling (NMDS) based on Bray-Curtis distances of the yearly mean cladoeran species abundances. NMDS ordination allows for the use of non-Euclidean distance measures (such as Bray-Curtis), which is useful when analyzing species abundance data. If Euclidean distance measures are used, the presence of double-zeros in species abundance data may result in two sites without any species in common being found more similar than other pairs of sites with shared species. All multivariate procedures were performed using PC-ORD (v4.25, McCune and Mefford 1999) and followed methods outlined by McCune and Grace (2002). In PC-ORD we ran the NMDS function in “slow and through” mode with 40 runs of real data, 50 runs with randomized data, and 88 iterations for the final solution. The final solution was 2-dimensional because adding a third dimension resulted in only a very minor reduction in the stress level (inverse measure of fit to the data). We performed a Mantel test to look for a correlation between cladoceran community composition and planktivore density, comparing the Bray-Curtis distance matrix of cladoceran community composition to the Euclidean distance matrix of planktivore density (significance tested by 1000 Monte Carlo randomizations).
To look for discontinuity in the response of the zooplankton community to changing planktivore levels through time, we plotted an index of zooplankton community dissimilarity against the density of planktivores in the lake. We used Bray-Curtis distances (calculated in R; R development core team (2003)) to quantify the degree of dissimilarity between the zooplankton assemblage found at beginning of the study (1989) and the zooplankton assemblages found in each of the fifteen subsequent years. This dissimilarity metric ranges from 0-1 and reaches its maximum value when no species are shared in common between assemblages. Other measures of community dissimilarity (Jaccard, Kulczynski) were highly correlated with the Bray-Curtis values.
McCune, G. and J. B. Grace. 2002. Analysis of Ecological Communities. MjM software Design, Gleneden Beach, Oregon, USA.
McCune, G. and M. J. Mefford. 1999. PC-ORD. Multivariate analysis of ecological data, v4.0. MjM Sotware Design, Gleneden Beach, Oregon, USA.
R Development Core Team (2003). R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. ISBN 3-900051-00-3, URL http://www.R-project.org.