Phytoplankton are a focal point in management strategies to improve or maintain lake water quality. Although much research has focused on macroscopic phytoplankton attributes such as total biomass and bloom-forming cyanobacteria, more subtle responses may be also be important for trophic relationships and may serve as indicators of ecosystem change. The objectives of this dissertation were to (1) quantify macroscopic and subtle responses of phytoplankton community structure to experimental manipulations of nutrient inputs and food web structure, and (2) determine what phytoplankton variates are reliable indicators of nutrient enrichment.

These objectives were achieved using data from whole-lake experiments conducted in Paul, Peter, West Long, and East Long Lakes (Gogegic County, Michigan) during summers 1991-1994. In May 1991, contrasting food webs were established by manipulating fish communities, then each lake was monitored weekly for two summers under background nutrient conditions. In 1993 and 1994, Peter, West Long, and East Long Lakes were enriched with nitrogen and phosphorus at 5-10 times background rates and at ambient nitrogen:phosphorus (N:P) ratios. Chlorophyll, primary productivity, and phytoplankton species were measured routinely; I created additional phytoplankton variates by aggregating the species data into taxonomic (genera, divisions), allometric (average size, size classes, size spectra) and community variates (diversity, richness, evenness).

I first quantified the effects of the nutrient and food web manipulations on these phytoplankton variates using ARIMA time series analysis. Nutrient enrichment increased cyanobacteria, chlorophytes, cryptomonads, mean size, chlorophyll and primary productivity, but decreased species diversity, chrysophytes and dinoflagellates. Larger zooplankton decreased small phytoplankton and species diversity, but increased large phytoplankton and mean size. Effects of zooplankton on chlorophyll and taxonomic divisions differed among lakes due to differences in species composition and size structure among plankton communities. Overall, the magnitude and direction of nutrient effects on all variates were more consistent among lakes than food web effects, suggesting that phytoplankton responses to nutrient perturbations should be more predictable than responses to food web perturbations. In addition, the effects of both perturbations on allometric variates were more consistent among lakes than effects on taxonomic variates, suggesting that responses of size-based variates to these perturbations should be more predictable than responses of taxonomy-based variates.

I then determined what phytoplankton variates were reliable indicators of the experimental nutrient enrichments using Bayesian time series analysis. Variates with high sensitivity to enrichment and low background variability were considered reliable indicators, while variates that failed to change following enrichment or changed without enrichment were considered unreliable indicators. Contrary to expectations from other perturbations, community and ecosystem variates were more reliable indicators of enrichment than species and genera. Productivity, biomass, diversity, and evenness changed significantly only during the first year of enrichment and only in enriched lakes. In contrast, changes in species and genera were unreliable due to high background variability: most taxa were not present often enough during a single year to reliably assess whether they had increased or decreased compared to previous years. Although species change is a useful indicator of toxic chemical stress, this conclusion does not appear to extend to nutrient enrichment.