RESULTS

Environmental Variables

A principal components analysis of the 12 chemical/physical variables reduced the interpretable pattern to four independent components of variance (Table 3.2). The first four principal components included the highest component loading (i.e., the correlation between the component scores and the variable) for each variable and accounted for 75.6% of the total variance. The first principal component had high-positive loadings (>0.7) for total Kjeldahl-N, pH, alkalinity, conductivity, and temperature, and a high-negative loading (<-0.7) for elevation. The second principal component, with high-positive loadings (>0.8) for depth and surface area, was an expression of lake size. The component loadings for the third and fourth principal components clearly indicated that these variables represent concentrations of phosphorus (orthophosphate-P and total phosphorus) and ammonia-N, respectively.

Table 3.2. Principal components analysis of environmental data collected from 51 lakes in North Cascades National Park in the summer of 1989. The analysis included 55 samples of 12 chemical/physical variables. Each variable was standardized by subtracting the mean from each observation and dividing by the corresponding standard deviation. The highest loading for each variable is indicated in bold type.

Community Properties

To examine relationships between the environmental variables and three aspects of phytoplankton community structure, total cell density, species richness, and species heterogeneity were regressed against the first four principal components of the chemical/physical variables. For this analysis, the components and response variables were standardized to zero mean and unit variance. Therefore, the explanatory variables were orthogonal, and the standardized regression coefficients were the simple correlation coefficients between each component and the response variable.

The regressions of total cell density, species richness, and species heterogeneity against the four environmental variables were significant (P < 0.05), but the multiple correlation coefficients were not particularly high: 0.56, 0.42, and 0.43, respectively (Table 3.3). The strongest relationship revealed by the regression models was between total cell density and the elevation-temperature-alkalinity-total Kjeldahl-N component (P < 0.01); there were no significant correlations between cell density and lake size, phosphorus concentration, or ammonia concentration (components 2, 3, and 4). Both species richness and species heterogeneity exhibited a significant positive correlation with the phosphorus component (P < 0.02), but were uncorrelated with the other components.

Table 3.3. Parameter estimates and ANOVA table for the regression of total cell density, species richness, and species heterogeneity against the first four principal components of the environmental variables listed in Table 3.2. The table also includes a multiple correlation coefficient for each model (R), test statistic (F), and associated probability (P) for each regression.

Taxonomic Composition of the Phytoplankton

Ninety-three taxa were observed in the 55 phytoplankton samples during the counting procedure. The relative abundance of each taxon, based on cell density and expressed as a percentage of the total relative abundance (%TRA), was calculated from the expression:

where p_ij was the proportional abundance of the i-th taxon in the j-th sample and k was the total number of samples, 55 in this case. When a taxon was a recognizable entity, but could not be identified to species, or in some cases to genus, it was given an identification number and a drawing was made for future reference. Forty-one taxa had total relative abundance values less than 0.1 % and were eliminated from the ordination analysis. Taxa that could be identified to the genus or species level and had relative abundance values greater than 0.1% are listed in Table 3.4. These taxa accounted for 80% of the total relative abundance in the species-abundance data matrix.

Table 3.4. List of taxa found in phytoplankton samples from 51 lakes in North Cascades National Park in the summer of 1989. The table also includes the relative abundances of each taxon in all samples (see text), the number of lakes in which a taxon was observed (B_o) and niche breadth values for each taxon (B_i). Taxa that could not be identified to the genus level and taxa with a relative abundance of less than 0.1% were eliminated from the table.

Of the 52 taxa with relative abundance values greater than 0.1%, there were 3 diatoms (Bacillariophyceae), 19 chrysophytes (Chrysophyceae), 14 chlorophytes (Chlorophyta), 4 cryptomonads (Cryptophyta), 8 cyanobacteria (Cyanophyta), 3 dinoflagellates (Pyrrhophyta), and 1 taxon of unknown taxonomic position. The corresponding relative abundances of these groups of taxa in the entire data set were 1.73% (diatoms), 19.63% (chrysophytes), 9.75% (chlorophytes), 3.25% (cryptomonads), 36.68% (cyanobacteria), and 1.54% (dinoflagellates). Taxa with the highest relative abundance values (Table 3.4) were Aphanocapsa delicatissima (14.0%), Chlorella (10.8%), and unidentified species of Diogenes (8.1 %). Other prominent taxa in the samples were Chromulina parvula (4.0%), Chromulina sp. (5.7%), Ochromonas sphagnalis (2.1%), Ochromonas sp. (3.6%), Aphanothece clathrata (5.9%), Gloeocapsa gelatinosa (3.1%). and Synechocystis sp. (4.7%). Of the ten taxa identified to the genus or species levels with relative abundance values greater than 2%, seven of these were found in 34 or more lakes (Table 3.4). However, only five of these seven taxa (Chlorella sp., Chromulina sp., Ochromonas sp., Rhodomonas sp., and Aphanocapsa delicatissima) had niche breadth values greater than 20. Aphanothece clathrata, a species with the fourth highest relative abundance value (5.7%), had a relatively low niche breadth (6.3), and was found in only nine lakes, eight of which were in the forest vegetation zone. Ochromonas sphagnalis, with a relative abundance of 2.1%, also had a very low niche breadth (2.7) and was found only in three subalpine lakes. Most of the rarer taxa in Table 3.4 (%TRA <2%) had relatively low niche breadth values, although a few of these occurred in 18 or more lakes and had niche breadth values greater than 12 (Chlamydomonas globosa, Oocystis solitaria, Chlorocloster sp., Chrysomonadales sp., Rhodomonas sp., and Peridinium inconspicuum).

At the level of taxonomic division, cyanobacteria had the highest relative abundance in the entire data set, i.e., the data pooled and considered as a single sample. Aphanocapsa delicatissima was the most abundant taxon in the pooled data, followed by unidentified species of Chlorella and Diogenes. The density of cyanobacteria decreased with increasing lake elevation; corresponding proportional densities of cyanobacteria in the vegetation zones were 49.3% (low-forest), 39.5% (high-forest), 34.1% (subalpine), and 24.8% (alpine; Table 3.5). Relative abundances of chrysophytes ranged from 25.7% (high forest) to 34.3% (subalpine). The relative abundances of chlorophytes ranged from 12.9% (low forest) to 25.0% (high forest).

Table 3.5. Proportional abundances of NOCA phytoplankton by vegetation zone in 51 lakes.

¹ BAC (diatoms), CHL (chlorophytes), CHR (chrsyophytes), CRY (cryptonomads), CYN (cyanobacteria), PYR (dinoflagellates), and UNK (unknown).
² 1 (alpine), 2 (subalpine), 3 (high-forest), 4 (low-forest)

Dinoflagellates and diatoms reached their maximum relative abundances in the alpine zone (7.8% and 7.9%, respectively), but were relatively rare (<4%) in the other vegetation zones. The relative abundances of cryptomonads were low in all vegetation zones, ranging from 0.9% (subalpine) to 3.4% (low forest). Only 14 taxa (15.1%) were found in 20 or more samples. These taxa had niche breath values of 12 or greater and only 5 taxa had values greater than 20 (range: 23.6 - 37.2). These results suggest that most taxa were limited in distribution and that their densities were uneven among lakes.

Lake Ordinations

CA sample (lake) ordinations were determined by the species-abundance data only, and therefore, the scores revealed the best latent gradients, independent of the measured environmental variables. CA axes 1 and 2 accounted for 19.5% of the variance in the phytoplankton data and had eigenvalues of 0.65 and 0.57, respectively. A plot of axis 2 against axis 1 (Figure 3.2) indicated that most alpine and subalpine lakes, and some of the lakes in the forest zones, had very similar phytoplankton floras. A group of lakes in the forest zones (NERT, THUN, RIDL, WILL, HOZO, and LS2) separated from the other lakes and were located on the right side of axis 1. These lakes all exhibited high relative abundances of Aphanothece clathrata, >50% of the total cell density in the case of WILL, RIDL, THUN, and NERT. The four alpine lakes were located on the left side of the tight cluster of subalpine lakes. Taxa that had high relative abundances in two or more of these lakes were Chromulina parvula, Gymnodinium sp., Aphanocapsa delicatissima, and Diogenes sp. The flora in 4 subalpine lakes was different enough to separate these lakes from the large cluster of alpine, subalpine and some lakes in the forested zones. Ochromonas sphagnalis had high relative abundances in three of these outliers (MR9, TRIL, and BEAR), particularly in TRIL where it represented 70.6% of the total cell density. TRIU separated from the cluster primarily because of a high relative abundance (41.4%) of Ochromonas pinguis. KLAW was dominated by the cryptomonad Chroomonas acuta (76.2%) and the chrysophyte Chromulina parvula (13.8%); the relative abundance of C. acuta in the other lakes was less than 4.5%. Because of the high dominance of C. acuta, Klaw was considered an extreme outlier, and consequently, was eliminated from the final ordination analysis and direct gradient analysis (see below).

Direct Gradient Analysis

CCA selected linear combinations of a subset of the environmental data that maximized the dispersion of the species scores along ordination axes. Therefore, in contrast to the indirect gradient analysis, the ordination axes were determined by both the species-abundance data and the environmental data matrix. An unrestricted Monte Carlo permutation test was used for the forward selection of environmental variables (ter Braak 1987, 1990). For this test, the number of permutations were 999, and the usual significance level of P < 0.05 was relaxed to P < 0.2 in order to retain at least one variable from each of the components of variance determined by the principal components analysis (Table 3.6). The degree to which each of the environmental variables individually affected dispersion of taxa along the first ordination axis was revealed by the corresponding eigenvalue (Table 3.6). The permutation tests reduced the number of variables to eight, each of which had an inflation factor of less than 3.0 (Table 3.6). Some variables with eigenvalues of 0.25 or greater (e.g., pH, temperature, and conductivity) were eliminated from the analysis because of colinearity with either alkalinity or total Kjeldahl-N.

Table 3.6. List of environmental variables evaluated by an unrestricted Monte Carlo permutation test for the CCA of NOCA phytoplankton data. The first eight variables in the table were retained for the analysis. Relationships between each variable and ordination scores predicted by the regression model are expressed by intraset correlation coefficients.

CCA axis 1 had an eigenvalue of 0.44 and accounted for 29% of the species-environment relation. The multiple correlation between the environmental variables and this axis was 0.86, which represented some improvement in comparison to the results from the indirect gradient analysis (Table 3.6). CCA axis 1 had relatively high intraset correlations with alkalinity (0.80), total Kjeldahl-N (0.79), and elevation (-0.78). CCA axis 2 had a relatively low eigenvalue (0.25) and was less interpretable. This axis had weak correlations with alkalinity (0.44), time of iceout (-0.49), and nitrate concentration (0.45). CCA axes 1 and 2 together accounted for 45.5% of the species-environment relation.

A graphic display of sample (lake) ordinations in relation to the eight environmental variables indicates that the most prominent gradient is characterized by changes in alkalinity and total Kjeldahl-N with elevation (Figure 3.3). Temperature, conductivity, and pH, variables eliminated from the analysis, have relatively high correlations with alkalinity and/or total Kjeldahl-N (>0.6), and also exhibit changes along an elevation gradient. For interpretative purposes, the axes in Figure 3.3 are scaled so the length of each arrow representing the environmental variables indicates the magnitude of the correlation between these variables and the ordination axes. Therefore, the projections of the sample points on each environmental arrow reveals the approximate positions of the lakes relative to the corresponding environmental gradient. For example, RIDL, WILL, THUN, and HOZO are lakes in the low- forest zone and have above average alkalinities and concentrations of total Kjeldahl-N. Lakes with relatively high concentrations of nitrate-N include TRIU, MRl3, TTAR, TRIL, and WILE. These lakes are found in either the subalpine or alpine zones. Environmental arrows can be extended through the origin in an opposite direction to explore the projection of samples with below average values of each variable. If this is done, the ordination predicts that many of the subalpine lakes and alpine lakes have below average alkalinities and concentrations of total Kjeldahl-N and orthophosphate-P. Futhermore, multiple samples from RAIN (1 and 2), PYRA (1 and 2), PAN (1 and 2), and WADD (1 and 2) indicate that seasonal changes of the taxonomic composition of phytoplankton assemblages were minimal in this study (Figures 3.2 and 3.3).

CCA ordinations of phytoplankton taxa that correspond to the lake configuration (Figure 3.3) are presented in Figure 3.4. Many of these taxa are located near the origin, indicating that they tend to occur in lakes with near average values for the environmental variables. A projection of the taxa points on the alkalinity and total Kjeldahl-N arrows place Aphanothece clathrata and Oocystis submarina at the positive end of the corresponding gradient, followed by Cryptomonas ovata, Chlorocloster sp., Ophiocytium cochleare, Cyclotella comta, Crucigenia tetrapedia, and Oocystis parva. Five taxa (Ochromonas sphagnalis, Dinobryon borgei, Haematococcus lacustris, Ochromonas pinguis, and Dactylococcopsis acucularis) tend to occur in lakes with above average concentrations of nitrate-N, whereas Dinobryon sertularia has high relative abundance in lakes with a high concentration of ammonia-N.

CCA also was used to examine relationships between phytoplankton taxa and the vegetation zones. For this analysis, the environmental data consisted of a matrix of zeros and ones that represented four nominal variables, one for each of the four zones (alpine, subalpine, high forest, and low forest). These zones were determined by the nature of the surrounding vegetation of each lake and were closely related to lake elevation (Lomnicky 1996). A two- dimensional plot of taxa scores and the centroids for each vegetation zone provided an indication of the affinities of the taxa for lakes in the different zones (Figure 3.5).

The general configuration of taxa in Figure 3.5 depicts a continuous distribution of taxa ordered from the subalpine zone, through the high-forest zone to the low-forest zone, with a large number of taxa occurring primarily in both the subalpine and high-forest zones. Taxa most closely associated with the alpine zone were Haematococcus lacustris, Dactylococcopsis acucularis, and Chroomonas acuta; whereas Oocystis parva, Oocystis submarina, Cryptomonas ovata, Cyclotella comta, Aphanothece clathrata, and Ophiocytium cochleare had a greater tendency to occur in the low-forest zone. Five taxa (Cyclotella stelligera, Gymnodinium sp., Peridinium inconspicuum, Rhodomonas minuta, and Chromulina parvula) that occurred in the subalpine and high-forest zones were able to extend their distribution into some of the alpine lakes.