METHODS

Sampling

From July through September 1989, 55 phytoplankton samples were collected from 51 NOCA lakes. Four lakes were sampled twice during the field season. The lake sampling schedule was limited by logistical problems associated with working in rugged terrain and much of the sampling effort was constrained by the timing of the open-water periods. For these reasons, low-forest and high-forest lakes were sampled before subalpine and alpine lakes. Some subalpine and alpine lakes were not sampled until September. All phytoplankton samples were collected with a 1.5-L Van-Dorn-style sampler from a depth of 1 m near the deepest location in each lake. The samples were preserved in Lugol's solution, and 500 individual cells were identified to species and counted in each sample using an inverted microscope (McIntire et al. 1996). The minimum size of cells counted was 1 µm.

Unfiltered water samples obtained concurrently with the phytoplankton samples were analyzed for pH, alkalinity, conductivity, and total phosphorus. Water samples also were analyzed for concentrations of orthophosphorus-P, Kjeldahl-N, nitrate-N, and ammonia-N after they were filtered through 0.7 µm prewashed Watman GF/C filters. The water samples were stored in acid washed polypropylene bottles. Water samples were transported out of the field, frozen, and then shipped to the Cooperative Chemical Analytical Laboratory, Oregon State University, Corvallis, Oregon, for analysis. Near-surfacewater temperatures were recorded with an Omega HH70 series hand-held thermometer. Lake elevation, area, depth, water quality, and nutrient data were provided by Liss et al. (1995).

Data Analysis

For analytical purposes, the data were organized into two files, a matrix of species abundances and a corresponding matrix of environmental data. The species-abundance matrix contained cell density values for 93 species; the environmental data included measurements of 12 chemical/physical variables, the initial date of the open-water period, and a classification index for vegetation zone. Species richness, species heterogeneity, and total cell densities for all taxa in each sample also were calculated and treated as a separate set of variables for regression analysis. Species heterogeneity was expressed by the Shannon measure of information (Pielou 1975), and species richness was specified as the number of species in a sample of 500 cells. In addition, the niche breadth of each taxon (B_i) was calculated from the expression:

In this case, values of B_i ranged from 1, when a taxon was present in only one sample, to 55, when it had the same relative abundance in all of the 55 samples.

The chemical/physical variables of the environmental data matrix were analyzed by a principal components analysis (Ludwig and Reynolds 1988). For this analysis, the data were centered, and each variable was standardized by the corresponding standard deviation. An examination of the component loadings indicated that the 12 variables could be reduced to four interpretable components that accounted for 75.6% of the total variance. Relationships between phytoplankton cell density, species richness, and species heterogeneity, and the environmental variables were analyzed by regression analysis. For this analysis, cell density, species richness, and species heterogeneity were designated as the response variables and the explanatory variables were the four principal components derived from the environmental data.

Relationships between the taxonomic composition of the phytoplankton and environmental variables were investigated by ordination methods (Jongman et al. 1987). Preliminary analysis of the species-abundance data indicated that a unimodal response model was appropriate for this purpose and that correspondence analysis without detrending (CA) and canonical correspondence analysis (CCA) provided satisfactory approaches to the analysis of distributional patterns and environmental gradients. For the ordination analyses, cell densities were relativized and expressed as a proportion of the total density in each sample. This transformation gave the samples equal weight, so lakes with relatively high cell densities did not dominate the analysis. Cell densities also were transformed to their logrithm and relativized, but the results of this analysis were no more interpretable than the analysis without the log transformation. Consequently, the results obtained for the relativized data without the log transformation were used for the presentation below.

To examine similarities in the phytoplankton flora among lakes, the species-abundance matrix was analyzed by CA. In this analysis, the lake (sample) ordinations were displayed, and the axes were scaled to approximate chi-square distances. CCA was used to reveal relative positions of individual taxa and lakes along gradients of selected environmental variables. CCA generated a constrained ordination by selecting linear combinations of environmental variables that maximized the dispersion of species scores along the ordination axes. In this case, the axes were scaled so that the species points were at the centroid of the samples in which they were found, and inter-species distances approximated their chi square distances. A Monte Carlo permutation test for a forward selection of environmental variables was performed to help determine a minimum set of variables that explained the species- abundance data. All ordination analyses and permutation tests were performed by the computer program CANOCO (ter Braak 1987, 1990).