Study site

The Dry Valleys are among the coldest and driest habitats on the planet and comprise the largest ice-free region on the Antarctic continent. This topographically diverse landscape contains a series of valleys punctuated by mountain ranges running from the Ross Sea to the Polar Plateau (Fig. 1).

Fig.1. 

Overview of Lake Fryxell basin, Taylor Valley, Antarctica. Locations of algal-mat transects analyzed in our study are marked with diamonds. Circles indicate locations of stream discharge gauges. Bowles Creek, marked with an arrow, lies west of Green Creek and does not have a stream gauge. Map adapted from Gooseff et al. 2003.

Dry Valley stream beds are dry for most of the year and become active for 6 to 12 wk during the austral summer. During this period of 24-h sunlight, incoming shortwave radiation generates glacial meltwater, which drains via ephemeral streams into closed-basin lakes on the valley floors. The streams are important sites for nutrient cycling (Gooseff et al. 2004, McKnight et al. 2004), support a relatively diverse suite of organisms and biological activities, and influence the water budget of the Dry Valley ecosystem. Streams retain meltwater through hyporheic storage and lose water via evaporation and biological metabolism (Conovitz 2000, Gooseff et al. 2003, Cozzetto et al. 2006). Weathering reactions in the hyporheic zone are the primary sources of ions, such as Si and K, to the water column (Lyons et al. 1998, Gooseff et al. 2002). Stream water temperature is highly variable, ranging from 0.1 to 15°C, and stream flow can vary 5 to 10× in a 24-h period depending on the sun angle (Conovitz et al. 1998, Cozzetto et al. 2006). Stream flow has been monitored at gauging stations on streams in Taylor Valley since the 1990/1991 summer season. These records demonstrate the high variability in annual stream flow across streams and seasons. For instance, the total annual volume of water flowing in Canada Stream is 2 to 17× higher than that flowing in Von Guerard Stream (Fig. 2). Also, infrequent flood years, such as occurred during the 2001/2002 season, can increase total annual discharge as much as 15× above the average for the recorded period.

Fig.2. 

Total annual discharge records for Canada Stream, Delta Stream, Green Creek, and Von Guerard Stream from 1990 to 2008. Seasons in which diatom samples were analyzed are indicated with hatches. Solid bars indicate seasons in which diatom data were not analyzed. N/A  =  no discharge data available, N/F  =  no flow observed.

In Antarctic algal mats, cyanobacteria are numerically dominant and have been studied with regard to taxonomy (Howard-Williams et al. 1986, Vincent et al. 1993, Alger et al. 1997, Taton et al. 2003, Comte et al. 2007, Komárek et al. 2008), primary production (Howard-Williams and Vincent 1989, McKnight and Tate 1997), and responses to desiccation (Davey 1989). In fine-scale examination of cyanobacterial mats, distinct layers were found that may prevent UV damage and desiccation and retain nutrients (Davey and Clarke 1992, Vincent et al. 1993). Spatial variability in stream algal-mat microbial communities is evident based on mat color, which changes depending on dominant cyanobacterium present (Howard-Williams et al. 1986). Within the main channel, algal mats are dominated by Oscillatoriales and appear orange or grey/brown. Along the stream margins, algal mats appear black because of high abundances of members of the order Nostocales. Additional mat types exist within the main channel on the under side of rocks or in the main channel as thick, rubbery mats, and combinations of mat types can be found in regions of variable flow.

The diatom flora of Dry Valley streams has been well characterized morphologically (Kellogg et al. 1980, Spaulding et al. 1997, Sabbe et al. 2003, Esposito et al. 2008). A complete list of Antarctic stream diatom taxa and currently known distributions can be found in Esposito et al. (2008). Of 41 stream diatom species, 17 have a current distribution that does not extend beyond the Antarctic continent (Esposito et al. 2008). These endemic species are distributed throughout Dry Valley streams at high relative abundances.

Algal mat samples were collected from 5 glacial meltwater streams—Von Guerard Stream, Delta Stream, Green Creek, Bowles Creek, and Canada Stream (Fig. 1). Algal mat transects located on each stream were sampled every 2–3 y as part of the MCMLTER program (Alger et al. 1997). The transects were chosen to represent similar habitats and are near the discharge gauging stations.

Von Guerard Stream is the 2^nd-longest (4.9 km) stream within the Fryxell Basin. The source glacier is at a high elevation in the Kukri Hills, resulting in a later flow season and higher discharge variability within the season and interannually (Table 1). The transect site is in a braided region of the stream channel where the stream has a low gradient and the rocks are arranged in a stable stone pavement (McKnight et al. 1999) that becomes less stable toward the stream margins. Algal mats have a patchy distribution and are predominantly black and orange.

Table 1. 

Physical characteristics of streams examined in this study. Mean total annual discharge (TotQ) represents the average total daily discharge measured during a summer season from 1990/1991 to 2007/2008, excluding 1992/1993 and 2000/2001. Qvarhist  =  maximum annual discharge/minimum annual discharge for the 1990–2008 time period.

Delta Stream is fed from the Howard Glacier and is the longest (11.2 km) stream in the Fryxell Basin. Flow begins later in the season but persists longer into February because of greater storage of water in the extensive hyporheic zone and in upland ponds. This stream has the highest historical variation in discharge of the 5 streams examined in our study (Table 1). At the transect site, the channel widens and seeps develop that connect with the main channel along the western edge of the stream bed. Some larger cobbles are present at a riffle, whereas the rest of the stream bed consists of coarse sand and has a shallow gradient. Orange mats occur in a narrow zone within the thalweg, whereas filamentous mats dominated by green algae and black mats occur in patches throughout.

Green and Bowles Creeks are on the southwestern side of Lake Fryxell and are fed by meltwater ponds at the base of Canada Glacier. They are the shortest streams with lengths of 1.2 and 0.9 km, respectively. Both creeks have shallow gradients and well defined channels consisting of a stable stone pavement, but Green Creek has higher discharge and a much wider stream channel (Alger et al. 1997). Orange and black algal mats are abundant at each transect. Discharge is not actively monitored on Bowles Creek, but discharge in Bowles Creek is correlated with discharge in Green Creek (Alger et al. 1997).

Canada Stream is ∼1.5 km long from its source on the tongue of Canada Glacier. It is one of the first streams in the Dry Valleys to begin flowing each summer and is one of the major sources of inflow to Lake Fryxell. The stream bed contains mostly cobbles and rocks that are arranged in a stable pavement. Along the stream margins, the stream bed becomes less stable with smaller particles. Orange, black, and green algal mats are abundant at this transect. Upstream of the transect, meltwater flows through a large pond at the base of the glacier.

Hydrologic and chemical data collection

Biological and environmental data used in our study were collected as part of the MCMLTER (available from:  www.mcmlter.org). Discharge data were collected continuously by pressure transducers installed at gauge sites on Von Guerard Stream, Canada Stream, Green Creek, and Delta Stream (Fig. 1) and logged at 15-min intervals throughout the austral summer. Discharge at Bowles Creek was estimated based on the correlation between 3 manual discharge measurements collected from Green Creek and Bowles Creek during the 2009/2010 flow season. The linear equation is: y  =  0.1847x − 0.0229 (r²  =  0.93).

Stream water was collected during weekly hydrologic site visits conducted throughout the flow season. Raw water samples were collected in triple-rinsed 250-mL Nalgene® bottles, and water for dissolved organic C (DOC) analysis was collected in 125-mL amber precombusted glass bottles. At the field laboratory, water for nutrient analysis was filtered from the raw water sample on glass-fiber filters and frozen for later analysis. DOC was filtered and acidified using concentrated HCl and stored at 4°C for later analysis. All chemistry analyses were performed at the Crary Laboratory at McMurdo Station. The mean values of all water-chemistry samples collected during a season were used in our study.

Algal-mat collection and preparation

Algal-mat samples analyzed in our study were collected early to mid-January from similar habitat and mat types to control for variation that might be attributable to these characteristics. Algal-mat samples were orange and were collected from the stable stone pavement areas of the main channel in each stream. Mats were collected using a brass cork borer (1.7 cm diameter) and placed in Whirl-Pak® bags containing 10 to 15 mL stream water. Samples were preserved in 10% formalin and shipped to the University of Colorado at room temperature. Replicate samples were limited for some streams, so an equal representation of samples from each stream and season was maintained by analyzing 1 sample from each stream during each season for which data were available. Representative mat samples were chosen at random or to maximize the availability of complementary biomass data. Samples were analyzed during the following seasons: 1993/1994 (all 5 streams), 1994/1995 (Canada Stream and Green Creek), 1997/1998 (Canada Stream, Delta Stream, Green Creek, Von Guerard Stream), 2002/2003 (all 5 streams), 2006/2007 (all 5 streams), and 2007/2008 (Canada Stream, Delta Stream, Green Creek, Von Guerard Stream). More samples were available from Green Creek and Canada Stream than from other streams. Therefore, replicate algal mat samples collected from these 2 streams were used in a separate analysis to determine the effect of interannual variation on diatom community composition. Replicates were collected at the same time along the transect. For this analysis, 1 to 4 mat samples were analyzed for each time point to the extent that samples were available.

Chlorophyll a (Chl-a) and ash-free dry mass (AFDM) analyses were conducted on algal-mat samples that were collected in conjunction with diatom sampling events or as ancillary data for other stream studies. Whether or not diatom samples were collected, all algal-mat samples for biomass analyses were collected in January and at the same transect sites with similar coring methods as described above. Samples were dewatered on precombusted Whatman GF/C® filters, wrapped in foil, and stored at −20°C. Prior to the year 2000, Chl-a was extracted in buffered acetone and analyzed spectrophotometrically with the trichromatic method. After 2000, Chl-a was extracted in buffered acetone and analyzed using a Turner Designs 10-AU field fluorometer (Turner Designs, Sunnyvale, California). The 2 methods produce comparable Chl-a results (Lorenzen and Jeffrey 1980). For AFDM analysis, samples were dried at 100°C for 24 h, weighed, burned at 450°C for 4 h and reweighed, and then rewetted and dried to determine mass loss caused by hydration of sediments. Both Chl-a and AFDM analyses were done at Crary Laboratory in McMurdo Station.

Algal-mat samples used for community analysis were digested using heat and H₂O₂ and rinsed several times with distilled water. Examination of oxidized material limits the ability to distinguish live from dead cells, but this method was necessary to make accurate species-level identifications. Based on the presence of chloroplasts in intact cells, 50 to 77% of the diatoms were alive at the time of sampling. A subset of the digested material was dried onto cover slips and permanently mounted on glass microscope slides with the mounting medium Zrax (W. P. Dailey, University of Pennsylvania). Relative abundances of diatom species were determined using an Olympus Vanox light microscope (Japan) at 1250× magnification, with ≥250 valves enumerated per slide.

Taxonomic identifications were done according to the descriptions of Sabbe et al. (2003), Van de Vijver et al. (2004), Esposito et al. (2008), Van de Vijver and Mataloni (2008), and the Antarctic Freshwater Diatoms database ( http://huey.colorado.edu/diatoms). The taxonomic identity of Psammothidium chlidanos, as defined by Spaulding et al. (1997) and Esposito et al. (2008), is under revision. It is referred to here as Psammothidium sp. #1 and may constitute a new species. In comparison to Psammothidium chlidanos, our specimens have coarser striae (25–28/10 µm vs 27–33/10 µm) and a more elongated central area in the pseudoraphe valve. Compared with its Antarctic relative, Psammothidium metakryophilum, our specimens are larger (length 14.3–16.2 µm × width 5.0–5.8 µm vs 9–12 µm × 4.3–5.2 µm).

Data analysis

Diatom community composition was analyzed using a variety of multivariate methods. Statistical analyses were done with Primer-E (Clarke and Gorley 2006) and the R software package (R Development Core Team, Vienna, Austria). For analyses that required a distance matrix of diatom communities, rare species occurring at <0.4% relative abundance (<∼1 valve in a given sample) were removed, and a distance matrix of relative abundance counts was calculated based on Bray–Curtis dissimilarity. Diatom community diversity (SW) was calculated with the Shannon–Wiener index

where p_i is the relative abundance of species i and S is the number of species. Evenness was calculated as the ratio of H′ to H′_max, which equals lnS.

Nonmetric multidimensional scaling (NMDS) is an ordination method that reduces the complexity of community data into fewer dimensions. This method can use any distance matrix, making it useful for community data. The nonmetric ordination method replaces intersample distances with their ranks, which makes the method less sensitive to nonlinear relationships and discontinuous distributions. The method is iterative and repeats the ordination calculations at random starting points. Species are plotted at the center of their distribution across samples. NMDS was done on a Bray–Curtis dissimilarity matrix of diatom community data using the vegan package in R. A 3-dimensional model produced a goodness-of-fit value of 6.4% using Kruskal's stress formula. A Shepard plot of calculated vs raw dissimilarities showed strong nonmetric (r²  =  0.996) and linear (r²  =  0.974) fits.

Principal Components Analysis (PCA) was used to examine variation in environmental variables across streams. A Euclidean distance matrix of all environmental data was calculated using either log(x)- or √(x)-transformed and scaled data.

A metric for determining the relative size of diatoms within a mat community (diatom relative biovolume) was calculated based on the equation

where x is the relative abundance of species i and b is the biovolume of species i. This metric differs from the standard sample biovolume calculation in that it represents only the biovolumes of the diatom species, which constitute a relatively minor proportion of the biomass of Dry Valley stream algal mats. Diatom biovolumes were calculated based on the geometric shapes for diatom genera outlined in Hillebrand et al. (1999). Mean lengths and widths were taken from the Dry Valley diatoms database ( http://huey.colorado.edu/diatoms). Determining frustule depth is technically challenging, so all species were assumed to have the same average cell depth of 2 µm. This assumption may have led to underestimation of the sizes of larger species.

Hierarchical cluster analyses were done on diatom community data using Bray–Curtis distance matrices. The average linkage method was chosen because it produced a classification tree with the highest cophenetic correlation coefficient (which describes the similarity of the modeled distances to the original pairwise distances).

Sixteen physical and chemical variables were considered to develop an explanatory model for diatom community composition. The physical variables and their definitions are listed in Table 2. Historical variation in discharge (Qvarhist) represents the maximum factor of interannual variation in total annual discharge (TotQ) for a stream by dividing TotQ of the highest flow season on record by the TotQ of the lowest flow season on record to give a single value for each stream. The chemical variables considered were NO₃⁻, NO₂⁻, NH₄⁺, soluble reactive P (SRP), DOC, and Si. The correlations between environmental variables and diatom community composition were determined using BEST analysis (PRIMER-E; Clarke 1993). This method maximizes the rank correlation between 2 distance matrices using Mantel-like tests, in which significance is tested via permutation (Mantel 1967). For environmental data, a Euclidian distance matrix was calculated on log(x)-transformed, normalized data.

Table 2. 

Description of hydrologic variables.

Stream-wide variation in environmental conditions and diatom communities

PCA of the environmental variables separated Von Guerard and Delta Streams from Green Creek, Bowles Creek, and Canada Stream (Fig. 3). PCA axis 1 was most strongly related to differences in hydrologic variables, whereas axis 2 was driven by differences in nutrient concentrations, flow variability, and flow frequency (Fig. 3). In general, Von Guerard Stream and Delta Stream had higher nutrient concentrations and higher interannual variability in stream flow than Green Creek, Bowles Creek, and Canada Stream (Fig. 3, Table 1). Conversely, the 3 latter streams have longer growing seasons and, for Green Creek and Canada Stream, total annual discharge is higher (Table 1). Numerous chemical and physical variables were correlated. Qvarhist varied significantly with stream length (R²  =  0.90, p < 0.005) and with ion concentrations, such as Na⁺ (R²  =  0.91, p < 0.05; Table 3). In contrast, SRP and other nutrients were not correlated to discharge variables (Table 3).

Fig.3. 

Principal components analysis (PCA) of all environmental variables considered in our study. See Table 2 for description of variables and abbreviations.

Table 3. 

Correlation coefficients of selected physical and chemical variables considered in our study. Mean values were used when correlating water-quality variables to maximum annual discharge/minimum annual discharge for the 1990–2008 time period (Qvarhist) and stream length. See Table 2 for variable descriptions. SRP  =  soluble reactive P, *  =  p < 0.05, **  =  p < 0.005.

Algal-mat diatom species were ubiquitous across streams, and most species were present in most streams (Fig. 4). However, their relative abundances varied greatly among streams. Hantzschia species occurred in higher relative abundances in Bowles Creek, Delta Stream, and Von Guerard Stream. Species of the genus Diadesmis occurred more abundantly in Canada Stream throughout the study period and in Green Creek after the 2001/2002 flood year. Luticola species were in highest relative abundances in Delta Stream and Von Guerard Stream, but also dominated Green Creek before the flood.

Fig.4. 

Dot blot representation of diatom relative abundance counts. Dashed lines separate counts data by stream. The distinctive shift in diatom community composition in Green Creek following the 2001/2002 flood season is marked with an arrow. Streams are coded by name and the last 2 digits of the sampling year. VonG  =  Von Guerard, Atay  =  Achnanthes taylorensis, Aolig  =  Amphora oligotraphenta, Ccyma  =  Chamaepinnularia cymatopleura, Cmole  =  Craticula molestiformis, Dcon  =  Diadesmis contenta, Dcvpar  =  Diadesmis contenta v. parallela, Dperp  =  Diadesmis perpusilla, Fpell  =  Fistulifera pelliculosa, Habun  =  Hantzschia abundans, Hamph  =  Hantzschia amphioxys, Hansp5  =  Hantzschia sp. #5, Hsubr  =  Hantzschia subrupestris, Hmuell  =  Hantzschia amphioxys f. muelleri, Laustr  =  Luticola austroatlantica, Lgaus  =  Luticola gaussi, Llaet  =  Luticola laeta, Lmutic  =  Luticola mutica, Lmuticop  =  Luticola muticopsis, Lcapi  =  Luticola muticopsis f. capitata, Levol  =  Luticola muticopsis f. evoluta, Lredu  =  Luticola muticopsis f. reducta, Ldoli  =  Luticola dolia, Lsp23  =  Luticola spp. #2/3, Matom  =  Mayamaea atomus, Matomv1  =  Mayamaea atomus v. #1, Mmeri  =  Muelleria meridionalis, Mpera  =  Muelleria peraustralis, Msupra  =  Muelleria supra, Ncinc  =  Navicula cincta, Nshac  =  Navicula shackletoni, Nwest  =  Nitzschia westii, Pbore  =  Pinnularia borealis, Pdelta  =  Pinnularia deltaica, Pquat  =  Pinnularia quaternaria, Pgerm  =  Psammothidium germanii, Psamsp1  =  Psammothidium sp. #1, Slati  =  Stauroneis latistauros, Scons  =  Stauronella constricta.

To develop a model that best explained the observed diatom community results, combinations of environmental variables were correlated with a distance matrix of diatom community data with BEST analysis (Clarke 1993). The model that best explained the variation in diatom communities contained 2 variables, TotQ and Qvarhist (ρ  =  0.463, p  =  0.01). Canonical correspondence analysis of all environmental variables produced similar results (data not shown).

Different streams had different sized diatoms. Based on the relative abundance counts and estimated biovolumes for each diatom species, a diatom relative biovolume measurement was calculated for each sample. Overall, a significant negative relationship between diatom relative biovolume and average annual stream flow explained 86% of the variability in diatom relative biovolume (p  =  0.004; Fig. 5A). In general, the relative abundances of small diatoms increased with increasing total annual discharge (Fig. 5B). One species in particular, Psammothidium sp. #1, increased exponentially with increasing stream flow (p < 0.0001). Analysis of variance results indicate that diatom relative biovolume differed significantly among streams (p < 0.0001). Most streams had characteristic ranges of diatom relative biovolumes. Green Creek was an exception and had much more variable diatom sample biovolumes that were significantly correlated with stream flow (R²  =  0.61, p  =  0.04). These biovolume estimates assume that all diatoms have the same depth, so these estimates probably underestimate diatom relative biovolume, particularly in streams containing larger diatoms. Thus, with more detailed biovolume measurements, the negative slope of the relationship between diatom biovolume and total annual discharge would be expected to increase.

Fig.5. 

A.—Mean diatom relative biovolume vs mean total annual discharge for each stream (R²  =  0.86). Mean discharge is based on the continuous flow record between 1990/1991 and 2007/2008, excluding 1992/1993 and 2000/2001, in which data are incomplete. B.—Dot blot representation of diatom relative abundances for common species (≥5%). Samples are arranged by stream along the x-axis from lowest to highest total annual discharge. Species are arranged along the y-axis from smallest to largest biovolume. See Fig. 4 for diatom abbreviations.

Diatom community composition across streams

NMDS analysis was used to determine stream-scale and temporal variation in diatom community composition. Each stream had a characteristic diatom community that broadly distinguished it from its neighbors (Fig. 6A). Von Guerard and Delta Streams clustered more tightly with each other than other streams. Canada Stream points were farthest from the other streams on NMDS axis 1. Green Creek diatom communities were distributed across the 1^st axis, indicating higher interannual variability. Bowles and Green Creeks were the closest geographically and most similar in source and stream characteristics, but their diatom communities did not cluster together. PERMANOVA results showed significant differences in diatom communities among most streams except when comparing Delta Stream to Von Guerard Stream or Bowles Creek (Table 4).

Fig.6. 

Nonmetric multidimensional scaling (NMDS) of diatom communities (stress  =  0.064) showing plots for sample coordinates (A) and diatom species with relative abundances ≥5% (B) displayed at the centers of their distributions. All species with relative abundances ≥1% were included in the ordination. The distance matrix was calculated based on the Bray–Curtis algorithm. Symbols are color-coded based on stream.

Table 4. 

PERMANOVA results testing for differences in diatom communities between streams. Statistically significant correlations (p < 0.05) are highlighted in bold.

Along NMDS axis 1, samples separated based on the relative abundances of Psammothidium sp. #1 and Diadesmis spp. on the left vs Hantzschia and Luticola species on the right. Fistulifera pelliculosa separated a Delta Stream sample collected during the 2007/2008 season from samples containing Hantzschia and Luticola species along axis 2 (Figs 4, 6B). Clear grouping of species by genus also occurred.

Diatom species richness varied spatially and temporally (Table 5). Canada Stream had the highest mean diatom richness of 27.7 species, whereas the other streams had similar richness values of ∼25 species, but these differences were not significant. Significant differences arose when examining species diversity and evenness. Von Guerard Stream had the highest SW diversity and evenness values. These values were significantly higher than values for Canada Stream, which had the lowest values because of the dominance of Psammothidium sp. #1. Delta Stream generally had high evenness, but variability was high because of the dominance of F. pelliculosa during the 2007/2008 season. Green and Bowles Creeks had similar SW diversity and evenness values throughout the study period.

Table 5. 

Diatom richness, Shannon–Wiener diversity, and evenness metrics for algal-mat samples from the available flow seasons. Mean values were calculated with data from seasons in which samples were collected from all streams. Postflood (2002/2003 season) values were not significantly different from mean values. Pairwise comparisons of mean diversity metrics for each stream were done with Student's t-tests on all seasons for which data were available for both streams. Results of significance tests are represented with lettered superscripts: streams with different letters had significantly different (p < 0.05) values for that metric.

Microbial-mat responses to a discrete flood event

During the cooling trend in the late 1990s, Chl-a concentrations in algal mats generally increased in Delta Stream, Green Creek, and Bowles Creek and decreased in Von Guerard and Canada Streams (Fig. 7A, B). In the most extreme example, during the 2000/2001 season, no flow occurred at the Von Guerard site and no orange mats were visible (DMM, personal observation; Fig. 7A, C). A similar pattern was observed for AFDM (Fig. 7C, D).

Fig.7. 

Interannual variation in algal-mat chlorophyll a (Chl-a) in Von Guerard Stream, Green Creek, and Canada Stream (A) and in Delta Stream and Bowles Creek (B) and ash-free dry mass (AFDM) in Von Guerard Stream, Green Creek, and Canada Stream (C) and in Delta Stream and Bowles Creek (D). Open circles indicate the lack of orange algal mats in Von Guerard stream during the 2000/2001 season, in which no flow was observed at the transect. Additional biomass data were obtained from the McMurdo Dry Valleys Long-Term Ecological Research (MCMLTER) database.

Following the flood of 2002/2003, Chl-a and AFDM concentrations decreased for all streams. Overall, average Chl-a values for all streams dropped from 10.48 to 3.97 µg/cm, and average AFDM values decreased from 13.3 to 4.0 mg/cm. Chl-a and AFDM values of algal mats from Canada Stream and Green Creek were only slightly lower after the flood, whereas Chl-a and AFDM in Delta Stream and Bowles Creek decreased drastically (Fig. 7A–D). Chl-a values declined again in the 2006/2007 season in Green and Bowles Creeks and then rebounded in 2007/2008. AFDM values increased slowly in all streams, and all streams except Bowles Creek returned to preflood values by the 2007/2008 season. No mats were visible in Von Guerard Stream during the season leading up to the flood, but the increase in Chl-a values between 2002/2003 and 2007/2008 was similar to the pattern observed in the other streams.

The significant decreases in algal-mat biomass did not correspond with major changes in diatom diversity metrics (Table 5). Nonetheless, SW diversity generally increased in less variable streams, particularly high-flow streams, after the flood. In contrast, increases in SW diversity were minimal in the highly variable Von Guerard and Delta Streams. The same pattern occurred when comparing richness and evenness values. Richness and evenness increased in higher-flow, less-variable streams after the flood, whereas essentially no change in these metrics was observed in more variable streams.

Species responses to the flood varied according to stream (Fig. 4). Relative abundances of L. muticopsis and Hantzschia sp. #5 temporarily increased in Delta Stream mats after the flood and had returned to preflood levels by 2006/2007. Von Guerard Stream experienced a temporary spike in the relatively rare species Achnanthes taylorensis, a decrease in Luticola spp., and an increase in Hantzschia abundans and Hantzschia amphioxys. Green Creek had the most striking changes in species abundance after the flood. Numerous Luticola species decreased, and the smaller species Psammothidium sp. #1 and Diadesmis contenta var parallela increased.

Unlike in the other streams, the community changes in Green Creek seem to have persisted after the flood. Hierarchical cluster analysis of the diatom community distance matrix showed Green Creek clustering with Delta Stream and Von Guerard Stream before the flood and clustering with Canada Stream after the flood (Fig. 8A). Our analysis is based on a single sample at each transect for each time period, so a more detailed analysis of Green Creek and Canada Stream was carried out with numerous replicate samples. Cluster analysis of diatom communities from these 2 streams showed that, with the exception of a single sample collected in 2008, Green Creek samples collected before the flood formed a distinct cluster, whereas samples collected after the flood clustered with Canada Stream (Fig. 8B). PERMANOVA results showed that season exerted a significant influence on diatom community composition in Green Creek (p  =  0.004, n  =  15), but not in Canada Stream (p  =  0.29, n  =  18). Diatom communities in Canada Stream did not vary as much as those in Green Creek, and the most different communities were found during a cold, low-flow season in 2000/2001 rather than after the flood (Fig. 8B).

Fig.8. 

Dendrograms resulting from hierarchical cluster analyses of diatom community data from all 5 Dry Valley streams (A) and Canada Stream and Green Creek with all replicates included (B). Boxes highlight the clusters containing Green Creek samples before and after the flood event. Labels are written as stream name followed by the sampling year in panel A and as stream name followed by the sampling year and sample number in panel B. VonG  =  Von Guerard.