Temporal fluctuations and experimental effects in desert plant communities.

In the Chihuahuan Desert of the southwestern United States we monitored responses of both winter and summer annual plant communities to natural environmental variation and to experimental removal of seed-eating rodents and ants for 13 years. Analyses of data on population densities of the species by principal component analysis (PCA) followed by repeated measures analysis of variance (rmANOVA) on PCA scores showed that: (1) composition of both winter and summer annual communities varied substantially from year to year, presumably in response to interannual climatic variation, and (2) community composition of winter annuals was also significantly affected by the experimental manipulations of seed-eating animals, but the composition of the summer annual community showed no significant response to these experimental treatments. Canonical discriminant analysis (CDA) was then applied to the data for winter annuals to more clearly identify the responses to the different classes of experimental manipulations. This analysis showed that removing rodents or ants or both taxa caused distinctive changes in species composition. There was a tendency for large-seeded species to increase on rodent removal plots and to decrease on ant removal plots, and for small-seeded species to change in the opposite direction. In the winter annual community there was a significant time x treatment interaction: certain combinations of species that responded differently to removal of granivores also showed opposite fluctuations in response to long-term climatic variation. The large year-to-year variation in the summer annual community was closely and positively correlated across all experimental treatments. The use of multivariate analysis in conjunction with long-term monitoring and experimental manipulation shows how biotic interactions interact with variation in abiotic conditions to affect community dynamics.