PERIPHYTIC DIATOM ASSEMBLAGES FROM BATHURST ISLAND, NUNAVUT, CANADIAN HIGH ARCTIC: AN EXAMINATION OF COMMUNITY RELATIONSHIPS AND HABITAT PREFERENCES

Sample collection

Water and periphyton were sampled over a short period of time (July 14–20, 1994) so as to minimize seasonal differences that might confound some of the spatial trends in the data set. Sampling techniques for all abiotic and biotic limnological components followed routine procedures as detailed by Pienitz et al. (1997) and Douglas and Smol (1994). A total of 38 sites were sampled, with 23 being shallow ponds and the remainder deeper lakes.

Environmental samples

A total of 39 environmental variables were determined for each site, although 6 of these variables were consistently below the detection limit and were consequently eliminated from the active data set (i.e. leaving a total of 33 active variables). A detailed description of water-sampling techniques used for this study is published elsewhere (Lim et al. 2001b), although a brief synopsis is detailed below.

Water temperature at approximately 0.3 m depth was measured to the nearest 0.5° C at each site using a handheld thermometer. Water samples for chemical analyses were also collected from approximately 0.3 m water depth. At each site, precleaned 250-mL polypropylene bottles were rinsed three times with lake/pond water before sample collection. Unfiltered samples for major and minor ion analysis were collected in these bottles. Similar samples were collected for trace metals.

Unfiltered samples for total unfiltered phosphorus (TPU) were collected in acid-prewashed 250-mL glass bottles. A 1-L polypropylene bottle was filled with lake/pond water and was subsequently partitioned and prepared for future analyses of chl a, particulate organic carbon (POC), particulate organic nitrogen (PON), further nutrient analyses (for details, see Lim et al. 2001b, pH, and conductivity measurements.

Upon returning to base camp, unfiltered samples were measured for pH with a Hanna Instruments pH meter (Hanna Instruments, Inc., Woonsocket, Rhode Island) calibrated to pH 7 and 10 buffers daily. Conductivity and salinity levels were measured using a Yellow Springs Instrument (YSI model 33, Yellow Springs, Ohio) temperature/conductivity/salinity meter corrected to 25.0°C.

Some samples were filtered in preparation for subsequent nutrient analyses by the National Water Research Institute in Burlington, Ontario (i.e. total “dissolved” phosphorus, filtered [TPF], nitrite [NO₂] plus nitrate [NO₃], ammonia [NH₃], soluble reactive phosphate-phosphorus, dissolved organic carbon, dissolved inorganic carbon, and total Kjeldahl nitrogen [TKN]). Samples were filtered through 47-mm-diameter cellulose acetate filters (pore size, 0.45 μm), and the filtrate was placed in acid-rinsed 125-mL glass bottles.

Phytoplankton biomass was estimated by filtering water samples through glass microfiber filters for analysis of chl a. An additional 100 mL of each water sample was filtered through preashed 25-mm-diameter glass microfiber filters (particle retention, 0.7 μm) for future analysis of POC and PON. Both filters were stored frozen in plastic petri dishes wrapped in aluminum foil. All filtered and unfiltered samples were kept chilled and in the dark for the duration of the field period. Subsequently, samples were shipped directly to the National Water Research Institute (Burlington, Ontario) for all additional nutrient and major ion analysis, following standard methods (Environment Canada Manual of Analytical Methods 1994a,b).

Periphyton samples

For comparative purposes, sampling techniques were similar to those used by Douglas and Smol (1995) in their Ellesmere Island periphyton study. Periphyton diatom samples were collected from all representative microhabitats within a site (i.e. from submerged moss, sediment, and rock substrates when present). Some sites did not support all microhabitats and were therefore not sampled for all substrates. Of the 38 sites sampled, 34 contained epiphytic communities and 29 were sampled for surface sediment collections. A total of 35 sites were sampled for epilithic communities, of which a representative subset of 16 sites was analyzed. In total, 79 periphytic samples were examined for this study.

Epiphytic communities were collected from submerged bryophytes, because when present they constituted the most ubiquitous aquatic plant substrate. Mosses are well suited to the short arctic growing season and often form dense mats on the lake and pond bottoms down to great water depth (Sand-Jensen et al. 1999). Samples representing the dominant submerged moss substrates were collected from three to four different areas at each site. Once collected, the moss samples were stored in 15-mL plastic scintillation vials and preserved in Lugol's iodine solution (Lind 1974). Sites BG (lake), BAC (pond), BAI (pond), and BAL (pond) did not contain any readily apparent submerged moss habitat. The basins of these four sites were rock covered and had little to no exposed sediment.

Nine sites (BI, BL, BO, BQ, BS, BY, BAA, BAB, and BAF) could not be sampled for surface sediments for either logistical reasons (e.g. unsafe ice cover) or because sediments were not accumulating in some of the shallow ponds, leaving a total of 29 sediment collections. Nine of these sites were lakes (i.e. BC, BG, BH, BK, BM, BN, BZ, BAE, and BAJ) and the remaining 20 were ponds. The uppermost top centimeter of the lake or pond surface was scooped into a 15-mL plastic scintillation vial. Depending on the bathymetry, sediment was either collected several meters from the edge of the shore or as close to the center of the site as possible.

The epilithic algal communities were sampled by brushing the attached algae from five representative submerged rocks into a 15-mL plastic scintillation vial with a toothbrush and preserving with Lugol's iodine solution. Of the 16 sites analyzed for this study, 5 were lakes (i.e. BH, BK, BO, BS, and BZ) and 11 were ponds.

Slide preparation

Preparation of diatom slides followed standard diatom techniques (Battarbee 1986). Approximately 1 cm³ of the moss fragments or 3 mL of sediment and rock scrape samples from each site were placed in glass scintillation vials, and 9 mL of a strong acid solution (50:50 nitric/sulfuric) was added to digest organic material. These samples were placed in a hot water bath at approximately 100° C for 2 h to expedite the digestion process and were subsequently left to cool for 12 h. Next, a series of eight washes with distilled water was performed until all acidic residue had been removed from the remaining siliceous slurry. Two dilutions of each slurry were pipetted onto glass coverslips and dried on a slide warmer at low heat over a 24-h period. Coverslips were then mounted on glass slides using the permanent mounting medium Naphrax (refractive index [R.I.] = 1.74).

Microscopy

A Nikon Optiphot-2 microscope equipped with differential interference contrast optics (numerical aperture [N.A.] = 1.25) was used to identify and enumerate a minimum of 300 diatom valves per sample under oil immersion (1000×). Diatoms were counted along transects following counting procedures outlined by Kingston (1986). In four cases (moss samples BH and BS, sediment samples BH and BZ), diatoms were scarce, and as a result fewer than 300 valves were counted. All common diatom taxa have been arranged alphabetically, and relevant information has been compiled in Table 1. Undefined diatom species have been given a generic and numeric designation (e.g. Achnanthes sp. 1). A Nikon FDX-35 mm camera was used to take light micrographs of the diatom flora. Photographic plates of the common diatom taxa are in Lim (1999).

Statistical data analysis

The three data sets (i.e. moss, sediment, rock) were compiled individually and were also combined to create a 79-sample data set. Statistical analyses were performed on individual and combined data sets.

Only diatom species with a relative abundance ≥1% in a minimum of three sites were included in the analyses. However, during the individual microhabitat analyses, 10 species (e.g. Navicula hilliardi found in sediment samples) that did not meet the aforementioned criteria but dominated one to two sites were retained in the substrate specific data set. They were not, however, retained in the overall combined data set of species listed in Table 1. Species names along with corresponding Hills N2 values, taxon codes, and numbers (which are included in brackets following species names whenever mentioned in the text) are listed in Table 1. N2 values represent the effective number of occurrences of a species in each microhabitat and are a robust measure of abundance. High N2 values, such as those for Nitzschia perminuta (63) on moss (N2_moss = 12.73), sediment (N2_sediment = 11.22), and rock (N2_rock = 8.52), indicate that this taxon is common in all three habitats relative to other species such as Cymbella subaequealis (35; N2_moss = 5.03, N2_sediment = 2.59, N2_rock = 2.94). Along with the canonical correspondence analysis (CCA) ordination results, N2 values were used to quantitatively determine which taxa had the greatest relative abundance in each of the three microhabitats.

Samples were also screened for characteristics that had the potential to skew statistical results. Moss samples from sites BH and BS and sediment samples from BH and BZ had fewer than 300 diatom valves. These sites were included in the analyses, although they were run passively (i.e. without associated limnological data) in the ordinations. Sites BF, BT, and BX had no diatom valves present in the moss samples, and these samples were eliminated from the data set, leaving a total of 31 epiphytic samples for inclusion in the statistical analyses and a combined data set total of 76 samples. The pond bottoms of these three sites were predominantly rock covered, with very small and sporadic mosses growing among the rocks. Given the overall lack of mosses in these sites, it is possible that the absence of epiphytic diatoms is indicative of suboptimal substrate conditions.

To relate the diatom distributions to the limnological variables measured in Lim et al. (2001b) and to define diatom habitat specificity, multivariate techniques were performed (Birks 1995). A brief synopsis of these techniques follows.

Detrended correspondence analysis (DCA) was performed to determine whether species distributions followed a unimodal or linear pattern based on the maximum spread between sites and species (i.e. lengths of gradients). Results indicated that CCA, a unimodal technique, would be appropriate for both individual microhabitat analyses and the combined data set (see Table 2 for DCA results).

CCA was performed using CANOCO v. 3.15 (ter Braak 1990, 1997 to identify patterns of distribution and influence among species, sites, and environmental variables. CCA was also used to ordinate the diatom species with respect to the three habitat types (i.e. moss, sediment, rock). To run a CCA for the 29 site sediment data set and the 16 site rock data set, the size of the environmental data set (i.e. of 33 variables) was reduced to a maximum of n− 1 in each case, where n is the number of sample sites. Environmental variables of high collinearity and significant correlation (P≤ 0.05) were first eliminated with the exception of one representative variable, and then forward selection was used to identify the minimum number of variables that could effectively explain the greatest amount of variance in the species data sets. Essentially, forward selection allows for the reduction of a large environmental data set and the construction of an environmental model for the species data. This is accomplished by sequentially ranking the environmental variables in their order of importance as they affect the species data. The significance of each variable, and its potential inclusion in the construction of the model, was determined by testing the significance of the first canonical axis using implicit Monte Carlo testing (based on 999 unrestricted permutations) (ter Braak 1988. This process produced a reduced environmental data set for the moss, sediment, and rock habitats that explained the species data almost as well as the full set (ter Braak 1990.

To reduce distortion in the CCA ordination, rare and dominant species were given less weight (i.e. downweighting) in the proportion to their frequency (ter Braak 1988 while retaining them in the analyses.

The length of the CCA biplot arrows demonstrates the relative importance of each environmental variable, whereas the placement of each arrow relative to the other arrows and to the ordination axes represents their approximate correlation to these factors (Jongman et al. 1987). The arrows point in the direction of maximum variation (ter Braak and Prentice 1988. The projection of site and species scores against these arrows allows for inferences to be made about the limnological state of any given site, as well as the dominant environmental factors affecting species composition (ter Braak and Prentice 1988.

The relationship between diatom taxon distributions and habitat was identified by species scores and their ordination patterns. For example, in the combined moss, sediment, and rock data set, those species occurring primarily on moss substrates would ordinate in close proximity to each other. Moreover, their substrate affinities were indicated by their proximity to each of the substrate variables on the CCA ordination. Those species found within approximate equidistance from each of the substrate choices (i.e. in the center of the ordination) were deemed to be generalists versus those that showed a strong association with one of the substrates.

Correspondence analysis (CA) was used to ordinate samples only in relation to their similarities in diatom assemblages, ignoring the effects of habitat in defining their relationship. CA was chosen because of the unimodal nature of our data set and also because of its ability to ignore environmental variables (i.e. habitat specificity) and simply locate sample scores based on similarities in species assemblages (i.e. sites that ordinate close to each other would share similarities in their diatom assemblages).

Habitat specificity

The eigenvalues of the first two axes of the CCA constrained only to habitat were λ1 = 0.19 and λ2 = 0.16, accounting for 9.7% of the cumulative variance of the diatom species. Monte Carlo permutation tests (99 unrestricted permutations) of the first two axes indicated that both were statistically significant (P≤ 0.01). Factors other than habitat type, such as water chemistry variables, which may have also affected species variance, are explored below.

Qualitatively, the CCA (Fig. 2) constrained to habitat type for the 76-sample data set demonstrates both similarities and dissimilarities in habitat preference among the 68 common diatom species. Habitat preference on the ordination is indicated by a taxon's proximity to a habitat centroid (e.g. close ordination to the sediment centroid indicates affinity to this habitat).

Not surprisingly, many of the diatom species were found on more than one substrate (e.g. moss and sediment), suggesting that some species take advantage of more than one habitat type. Alternatively, taphonomic processes can lead to diatom frustules from other habitats collecting in the sediment. Furthermore, a few species (e.g. Cocconeis placentula[24] and Cyclotella pseudostelligera[25]) showed a strong affinity to only one habitat type (e.g. moss, and rock, respectively).

Diatom species found near the center of the 76-sample ordination (Fig. 2), such as Achnanthes minutissima (8), Cymbella arctica (27), C. subaequealis (35), Fragilaria capucina var. capitellata (44), Navicula cryptotenella (50), and N. vulpina (56), are considered to be common to all three habitats. Eleven diatom taxa (i.e. Achnanthes bioretii[3], A. oestrupii[10], A. ricula[13], Amphora veneta var. capitata[20], Navicula bryophila[48], Navicula jaernfeltii[51], N. pseudoscutiformis[53], N. salinarum[54], Neidium umiatense[58], N. iridis[57], and Nitzschia cf. perminuta[64]) were found exclusively on sediment. Douglas and Smol (1995) found nine taxa unique to the Cape Herschel surface sediment assemblages, of which three, Caloneis schumanniana (22), Diploneis oculata (42), and Fragilaria pinnata (47), were also found in close proximity (although not exclusively) to the sediment centroid in the Bathurst Island ordination (Fig. 2). Achnanthes chlidanos (4), A. laevis (6), Cymbella designata (30), Denticula elegans (38), Nitzschia frustulum (60), N. palea (62), and Pinnularia subrostrata (67) also had a strong affinity to the Bathurst Island sediment substrate. Of the 11 species unique to the sediment habitat, four (Achnanthes bioretii[3], A. oestrupii[10], Amphora veneta[20], Navicula jaernefeltii[51]) had their greatest relative abundance in site BZ, three (Navicula bryophila[48], Neidium iridis[57], Nitzschia cf. perminuta[64]) in site BT, two (Achnanthes ricula[13], Neidium umiatense[58]) in BH, and one each in sites BG (Navicula salinarum[54]) and BAD (Navicula pseudoscuteloides[53]). As reported by Lim et al. (2001b), with the exception of BH, each site had temperatures equivalent to or below the reported mean of 9.0° C. Therefore, these data indicate that, in addition to showing similar autecological characteristics, these epipelic taxa may also be found under specific limnological conditions.

Four species were unique to the moss substrate (i.e. Cocconeis placentula[24], Cymbella turgidula[37], Navicula minima[52], and Pinnularia sp.1 [65]). Although found on different plant substrates, Reavie and Smol (1997) also found Cocconeis placentula (24) to be quite restricted to the epiphyton in the St. Lawrence River. Cymbella turgidula (37) was, on the contrary, recorded only in the sediment substrate of the Cape Herschel habitat study, although this does not preclude its possible presence on moss substrates (Douglas and Smol 1995). Achnanthes sp.1 (1), Achnanthes sp.2 (2), Achnanthes peragalli (11), Achnanthes petersenii (12), and Caloneis silicula (23) all shared a close affinity to moss substrates. Achnanthes sp.1 was also found in close proximity to the moss centroid of the Cape Herschel study.

Only one diatom species was found exclusive to the rock substrates: Cyclotella pseudostelligera (25). Cyclotella species are typically considered to be planktonic; however, in this case it is possible that Cyclotella pseudostelligera (25) is tychoplanktonic or even benthic, thus accounting for its presence on the rocks. Amphora inariensis (17), Cymbella cesatii (29), and Nitzschia inconspicua (61) were all found to have a strong affinity for the rock substrates as well. Reavie and Smol (1997) also found Amphora inariensis (17) characteristic of rock substrates in the St. Lawrence River. Douglas and Smol (1995) commented on the apparent affinity of Amphora species for an epilithic habitat preference.

As stated earlier, some species ordinated between habitat types indicating their presence on two substrates. Achnanthes kryophila var. petersenii (9), Amphora pediculus (19), Cymbella latens (31), Eunotia arcus (43), Fragilaria construens (46), and Pinnularia balfouriana (66), for example, were all found in both moss and sediment habitats. The relative affinities of each of these species for one substrate over the other could be discerned from their position relative to a habitat centroid. For example, although Eunotia arcus (43) was found on both moss and sediment substrates, it has a greater affinity for mosses and its presence on the sediment may be due to taphonomy. Pinnularia balfouriana (66), on the other hand, was found almost at the halfway point between the moss and sediment centroids, indicating approximately equal representation on both substrates. However, given that P. balfouriana (66) is commonly found as an epiphytic diatom (Douglas and Smol 1999), its presence on both substrates likely reflects the sediment being an integration of diatom remains from the moss habitats. Fragilaria capucina var. vaucheriae (45), Diatoma tenuis (41), Achnanthes subatomoides (14), and Nitzschia alpina (59) are four examples of diatom taxa that were found to share a strong affinity for both the rock and the moss habitats. Amphora aequealis (16), Navicula cryptocephala (49), N. soehrensis (55), and Stauroneis anceps (68) are examples of taxa that were present on both rock and sediment but not moss substrates.

Similarities in the diatom flora between microhabitats were identified by CA (Fig. 3). Axis 1 and axis 2 explained 13.6% and 7.2%, respectively, of the variance in the species data. Qualitative examination of the ordination indicates distinct diatom assemblages associated with specific substrates, although there is a degree of overlap. Specifically, the moss samples tend to be grouped in the upper portion of the ordination and the epipelic communities in the lower right quadrant. However, the epilithic assemblages are less distinguishable.

The sediment samples are interspersed in all the quadrats of the CA to a greater degree than the moss and rock samples, which indicates that while having some distinct taxa, sediment samples most likely include diatom remains from the other two substrates. Eight species were common to sediment and rock substrates alone, 15 common only to the moss and sediment substrates, whereas 2 taxa overlapped between the moss and rock substrates alone (Table 1). These data further indicate that although there may be overlap among the diatom communities found on each substrate, the dominance of some individual taxa in sediment sample collections may be strong indicators of specific microhabitat presence in a lake or pond. Limiting biomonitoring efforts in high arctic lakes and ponds to only sediment collections may be a viable and robust method as the sediments incorporate the dominant taxa of all other substrates. However, substrate sampling is still important to identify which taxa are unique to a specific substrate. These autecological data have important paleolimnological ramifications.

Environmental influences

Three separate CCAs were performed for each individual habitat (i.e. sediment, moss, rock) to identify important environmental variables affecting the respective patterns of distribution among the diatom taxa. The quantitative and qualitative results of each of the CCAs are discussed below.

Moss samples

The eigenvalues of the first two DCA axes (λ₁ = 0.52 and λ₂ = 0.25) accounted for 20.2% of the cumulative percentage variance of the diatom species data (Table 2). After the removal of six highly collinear variables (i.e. Al, Mn, Zn, PON, NO₃-NO₂, NH₃) from the original data set (and before forward selection), the first two axes of the CCA accounted for 25.1% of the cumulative percentage variance of the species data. Furthermore, the first two axes of the CCA accounted for 26.6% of the diatom-environment relationship, whereas the diatom-environment correlations for axis 1 (0.99) and axis 2 (0.97) indicate a strong relationship between the diatoms and the 28 environmental variables.

Na⁺ and total nitrogen (TN, defined as the sum of TKN, NO₃NO₂, and PON) were identified as variables that significantly (P≤ 0.05) explained the species variance. Both variables significantly correlated with one or more environmental variables. For example, Na⁺ was significantly (P≤ 0.01) correlated with other ions such as Mg⁺, K⁺, Cl⁻, and Fe³⁺, as well other variables such as conductivity. Na⁺ and TN together accounted for 16.5% of the explainable variation and significantly (P≤ 0.01) contributed to all four canonical axes (Table 2).

CCA axes 1 and 2 ordinate sites and species along nutrient (represented by TN) and ionic (represented by Na⁺) gradients, respectively (Fig. 4). Eight (i.e. BH, BM, BN, BO, BS, BAA, BAE, and BAF) of 13 lake samples represented in this CCA ordinate at the low nutrient low ionic end of the gradient. All of these sites had [TN] values below the mean (mean, 577.19 μg·L⁻¹), whereas all but sites BM and BN had [Na] (mean, 3.1 mg·L⁻¹) and other ion concentrations below the mean. These sites were dominated by such species as Achnanthes sp.1 (1), Achnanthes sp.2 (2), Achnanthes subatomoides (14), Caloneis sp.1 (21), Cocconeis placentula (24), Cymbella arctica (27), C. tumidula (36), C. turgidula (37), Fragilaria construens (46), F. pinnata (47), Navicula minima (52), Nitzschia perminuta (63), and Pinnularia balfouriana (66). Cocconeis placentula (24) and Cymbella turgidula (37) were only found on the moss substrates. Three of the five samples containing Cocconeis placentula (24) were recovered from lakes, whereas two were taken from large ponds. However, all four sites were sampled close to the shore.

Cymbella turgidula (37) was in greatest relative abundance in three sampled lakes. Site BS had the highest abundance of this taxon (6.9%), was characterized by significantly lower ionic and nutrient concentrations than the mean, and was considered to be ultraoligotrophic. Douglas and Smol (1995) found this species only on sediment substrates; however, relative percentages were low, with the greatest abundance being 2.0% in pond 3, at Cape Herschel, Ellesmere Island. This pond lay within a sedge meadow and had lower Na⁺ concentrations relative to other sites sampled in that region (Douglas and Smol 1994). Whether Cymbella turgidula (37) is strictly epiphytic or epipelic in nature cannot be discerned from either of these studies for certain. However, it is possible that this is an environmentally sensitive species, with a narrow autecological range.

Fourteen of the 18 ponds represented by moss samples ordinated in close association with high nutrient, high ionic concentrations. Associated with these ponds are such species as Cymbella cesatii (29), Cymbella microcephala (32), Cymbella subaequalis (35), Eunotia arcus (43), Fragilaria capucina var. capitellata (44), Navicula cryptotenella (50), Navicula vulpina (56), Nitzschia alpina (59), and Nitzschia frustulum (60). This is a similar list to those taxa found on sediment substrate samples from the warm nutrient-rich ponds and small lakes (see below). With the exception of Eunotia arcus (43), each of these species was found on all three substrates (i.e. moss, sediment, rock). Douglas and Smol (1995) also found Eunotia species to be associated with mosses, whereas Weckström et al. (1997) found this epiphytic genus to favor warm surface waters. Furthermore, the Eunotia species could take advantage of the localized increases in acidity levels near the moss-substrate interface, which is caused by cation exchange (Glime and Vitt 1984). Because Eunotia is commonly found in low ambient pH environments, the moss substrates may confer a better substrate for this genus in these highly alkaline lake and pond waters.

Rock samples

The eigenvalues of the first two DCA axes are λ1 = 0.44 and λ2 = 0.19, accounting for 35.4% of the cumulative percentage variance of the diatom species (Table 2). Before forward selection, the first two axes of the CCA accounted for 38.9% of the cumulative percentage variance of the diatom data and 39.0% of the diatom-environment relationship.

TPF and pH were the variables that significantly (P≤ 0.05) explained the epilithic species variance. Along the first two axes, these variables accounted for 23.9% of the explainable variance in the diatom data and 100% of the species-environment relationship (Table 2). Furthermore, these two variables contributed significantly (P≤ 0.01) to all four axes.

Axis 1 of the CCA is driven primarily by a nutrient gradient represented by TPF, whereas axis 2 is related to a pH gradient (Fig. 5). All the lake samples (i.e. BH, BK, BO, BS, BZ) with the exception of BAJ clustered separately from the ponds in the lower right section of the CCA. Their position relative to the pH and TPF gradients indicates that they are characterized by low nutrient concentrations and lower alkalinity levels relative to the other sites. The diatom taxa associated with these deeper sites include Achnanthes subatomoides (14), Cyclotella pseudostelligera (25), Fragilaria capucina var. capitellata (44), F. construens (46), Nitzschia palea (62), and Pinnularia subrostrata (67). Cyclotella pseudostelligera (25) was unique to the rock substrates in this study and was not documented in the habitat study of other high arctic sites on Ellesmere Island (Douglas and Smol 1995). The larger ponds (e.g. BAK, BAL, BAN) that ordinate at the lower extreme of the TPF gradient are dominated by such species as Cymbella cesatii (29), C. arctica (27), Fragilaria pinnata (47), and Pinnularia balfouriana (66), which is a trend mirrored by the moss and sediment sample ordinations. Sites such as BD, BP, BAD, and BAJ are inferred to have higher concentrations of TPF relative to the other sample sites and ordinated in close association with, for example, Denticula elegans (38), D. kuetzingii (39), Diatoma tenuis (41), Diadesmis sp.1 (40), and Navicula cf. pupula. Species ordinating in association with higher pH levels and [TPF] include Cymbella cesatii (29), Cymbella microcephala (32), Cymbella subaequalis (35), and Nitzschia inconspicua (61).

Sediment samples

A detailed description and discussion of the CCA based on the surface sediment diatoms is in Lim et al. (2001a). The following is a brief summary of the key findings. The eigenvalues of the first two DCA axes are λ₁ = 0.61 and λ₂ = 0.31, accounting for 30.0% of the cumulative percentage variance of the diatom species data (Table 2). Forward selection and Monte Carlo permutation tests (999 unrestricted permutations) reduced the environmental data set from 29 to 5 variables (Fe³⁺, TPU, TN, temperature, and pH) that significantly (P≤ 0.05) explained the species variance. Along the first two axes of the CCA, these selected variables together accounted for 16.5% of the diatom data cumulative variance and explained 56.1% of the cumulative variance in the species-environment relationship (Table 2). These five variables significantly (P≤ 0.01) contributed to all four CCA axes.

Axis 1 is most strongly influenced by ionic concentrations (represented by Fe), nutrient (represented by TPU and TN), and temperature gradients, whereas axis 2 is driven by pH (Fig. 6). Similar to the findings of the Lim et al. (2001b) limnological analysis, the lakes and larger ponds tend to cluster separately from the ponds and shallower lakes particularly along the temperature gradient. Fragilaria construens (46), F. pinnata (47), and Pinnularia balfouriana (66) dominate in these larger, colder, less nutrient rich sites. Interestingly, these three taxa are also clustered together on the moss (Fig. 4) and rock CCAs, in close association with similar deeper colder sites to those in the sediment CCA. Therefore, given the strong association these small benthic species have with moss and sediment habitats (Fig. 2), it is possible that they thrive in the ultraoligotrophic conditions of these sites by remaining in the littoral zone and associating with the moss substrates and the warmer interstitial waters of the surface sediment, where an appreciable amount of heat may be absorbed through the summer months.

Overall, species such as Achnanthes flexella (5), A. laevis (6), Caloneis schumanniana (22), Cymbella cesatii (29), C. subaequalis (35), Denticula kuetzingii (39), Diatoma tenuis (41), and Eunotia arcus (43) ordinated together and in close association with the maximum rate of change of the TN and temperature gradients. Sites associated with this extreme in the ordination were ponds (e.g. BD, BJ, BW). Many of the species associated with these warmer sites were also found to have an affinity for both moss and sediment habitats. These two habitats were found in abundance throughout the shallow warm sites where light could penetrate to the bottom and allow for moss growth, for example, throughout the water body.

Diatom species associated with the pH gradient shifted from such predominant taxa as Cymbella designata (30), C. tumidula (36), Navicula cryptotenella (50), and N. soehrensis (55) near the arrowhead (higher pH values) to small Navicula (e.g. N. jaernefeltii[51], N. bryophila[48]), Achnanthes (e.g. A. oestrupii[10], A. ventralis[15]), and Amphora (A. aequalis[16], A. libyca[18]) species at the opposite extreme (lower pH values). This spread along the pH gradient suggests that the smaller Navicula and Achnanthes species may prefer less alkaline conditions. However, the pH range was small (8.1–8.6), and therefore any ecological conclusions inferred about the diatom pH optima should be treated with caution.