Over the course of the last century, the 9000-km2 Ellesmere Ice Shelf (82–83°N, 64–90°W) fragmented into six main ice shelves now totaling 1043 km2. This ensemble of thick ice environments lies along the northern coast of Ellesmere Island in the Canadian High Arctic and provides a cryohabitat for microbial communities that occur in association with eolian and glacially entrained sediments on the ice surface. We undertook a comparative analysis of physical, chemical, and biological characteristics of five of the remnant ice shelves including geographic information system (GIS) mapping of ice types. Each of these remnants is a thick (>20 m) mass of ice with substantial sediment overburden that promotes the formation of oligotrophic meltwaters in the summer. Microbiota occurred in all sampled sediment, forming a continuum of abundance from sparse to loosely cohesive and pigmented microbial mats. Using digital images from over-flight transects we determined that 8% of the combined ice-shelf area was suitable microbial mat habitat, and contained an estimated 34 Gg of organic matter stocks for the entire system. A gradient of increasing chlorophyll a, organic content, and conductivity was found from west to east. This is likely related to the surface ice type (meteoric versus marine) and to the relative availability of sediment. Our results indicate that differences in phototrophic community structure (microalgae and cyanobacterial morphotypes) were associated with different ice and microbial mat types. In addition, the relative abundance of dominant taxa was significantly associated with environmental gradients of conductivity, soluble reactive phosphorus, and nitrate and ammonium concentrations. There were distinct differences between each ice shelf with regards to ice type and sediment availability but no differences in taxonomic richness or diversity, indicating little effect of habitat fragmentation on these community attributes. However, the ensemble of ice shelves that compose this unique cryoecosystem remains vulnerable to habitat attrition and complete loss with ongoing climate warming.
Ice shelves, thick masses of coastal ice floating on the ocean, are found in both the North and South polar regions and have been traditionally viewed as abiotic components of the cryosphere. However, numerous articles have shown that a rich diversity of microbial life exists in seasonal meltwaters on the surface of certain Antarctic ice shelves (Vincent, 1988; Howard-Williams et al., 1990; Suren, 1990; Hawes et al., 1993; Mountfort et al., 1997). More recently, the importance of similar microbial communities on the Ward Hunt Ice Shelf (lat. 83°N, long. 74°W) in High Arctic Canada was recognized (Vincent et al., 2000).
The Ward Hunt Ice Shelf is the largest remaining ice shelf in the Northern Hemisphere and is accompanied by five other substantial ice shelves along the northern coast of Ellesmere Island in the Canadian Arctic: Markham, Ayles, Milne, Petersen, and Serson (formerly known as Alfred Ernest Ice Shelf) ice shelves (Fig. 1). In addition to the main ice shelves, there are a few minor ice shelves (a few square kilometers in area) and scattered shelf ice fragments locked in multiyear landfast sea ice. These ice shelves are remnants of the Ellesmere Ice Shelf that spanned a ~400 km stretch of the northwestern coast of Ellesmere Island at the beginning of the last century (Koenig et al., 1952). Much of the fragmentation of this large ice shelf occurred prior to the 1950s; however, since then several calving and break-up events have taken place, further fragmenting this feature into smaller pieces (Hattersley-Smith, 1963; Jeffries and Serson, 1983; Jeffries, 1986a; Mueller et al., 2003).
The existence of ice shelves in the Northern Hemisphere is limited to the aforementioned Canadian Arctic examples, several in the Russian Arctic (Dowdeswell et al., 1994; Williams and Dowdeswell, 2001), and some potential candidates in northern Greenland (Higgins, 1989). This is in marked contrast with Antarctica that has 40% of its coastline fringed with ice shelves, some as large as half a million square kilometers (Jeffries, 1992b). There, they are typically formed by floating glacier tongues that aggregate into broad expanses of ice up to hundreds of meters in thickness. In contrast, Arctic ice shelves are mostly formed by in situ ice accumulation of marine basement ice and meteoric (atmospherically derived) iced firn, and in some cases, this is added to by glacial ice.
The surface of Arctic ice shelves is marked by an east-west trending ridge and trough topography with an amplitude of 2 to 7.5 m and a wavelength of 87 to 450 m (Jeffries et al., 1990). This regular undulating surface becomes more chaotic closer to coastlines (Hattersley-Smith, 1957) and the relative importance of wind versus tidal and other potential controls on Arctic ice shelf morphology has not yet been conclusively determined (Holdsworth, 1987). In this northern region, air temperatures can approach or exceed 0°C beginning in late June and continuing until late August (Alert, Nunavut; Meteorological Service of Canada, Climate Normals 1971–2000). During this period, meltwater flowing into the troughs accumulates in long (up to 15 km), thin (10 to 200 m) and shallow (up to 3 m) lakes. Liquid water is then seasonally available for biological processes, and this period constitutes the main growth season for resident microbiota.
In addition to snow, ice, and water, sediment is a conspicuous constituent on the ice shelf surface. It can be delivered in three ways or by a combination of these modes: across the ice surface via eolian distribution (Crary, 1958), pushed with the ice as glacial moraine debris (Lemmen et al., 1988), or through the ice via basal accretion and subsequent surface ablation (Crary, 1960). Microbial consortia are associated with sediment at the bottom of meltwater lakes and are arranged in a loosely cohesive microbial mat (organosedimentary matrix). Microbial mats may be found in discrete melt holes up to 50 cm in diameter (termed cryoconite holes on glaciers; see Mueller et al., 2001) or they may be found in more extensive and interconnected patches of sediment. In some areas, sediment and microbial mats may be located on the ice but not overlain by water. Overwinter measurements made on the Ward Hunt Ice Shelf (2001–2002) suggest that these mats become covered with snow and/or ice and remain below 0°C for 300 days of the year reaching minimum temperatures of −17°C (Mueller and Vincent, 2006). Oscillatorian cyanobacteria dominate the microbial assemblage but diatoms, green algae, other protists, heterotrophic bacteria, and viruses are also present. In addition, microinvertebrates such as rotifers, tardigrades, and turbellarian worms have been identified within the microbial mats (Vincent et al., 2000). This contrasts with microbial mats on the McMurdo Ice Shelf, which lack turbellarians, and are generally highly cohesive and layered (Howard-Williams et al., 1990; Vincent et al., 2000).
The cryohabitats of Arctic ice shelves have been little characterized and all work to date has focused on the Ward Hunt and Markham ice shelves. This includes a study of the environmental conditions at the microbial scale (Mueller and Vincent, 2006), pigment analysis of microbial mats (Vincent et al., 2004; Mueller et al., 2005) and an assessment of the microbial mat ecophysiology (Mueller et al., 2005). No analysis has been undertaken to assess the microbial ecology of the entire Ellesmere Ice Shelf cryoecosystem. It is therefore not known how much habitat exists across all these ice shelves, what the limnological properties of the meltwater lakes are, and the extent of spatial variability of the microbial communities. In the present study, we redress this by way of detailed transects, remote sensing and sampling of microbial mats and surface waters from five of the largest Ellesmere ice shelves. Our objectives were to quantify habitat availability (at two spatial scales), limnological conditions and taxon relative abundance over the entire cryoecosystem. We used these data to estimate the current standing stocks of microbial mats on the Ellesmere Ice Shelf remnants and to address the hypothesis that the microbial mat relative abundance and biomass is associated with environmental gradients (i.e., features of the environment that vary spatially). Our observations allowed us to delineate three categories of microbial mat as well as two ice types and to evaluate the hypothesis that differences exist in the relative abundance and environmental conditions among these nominal groups. Finally, we considered the ecological ramifications of the ongoing habitat fragmentation of the Ellesmere Ice Shelf and the implications of cryohabitat loss and isolation between glacial epochs.
Fieldwork was undertaken in late July and early August of 2001 and 2002. In 2001, five major ice shelves were visited by helicopter. We did not visit Petersen Ice Shelf or the minor ice shelves and shelf ice fragments locked in multiyear landfast sea ice. Sampling of surface waters and microbial mats (if present) took place on each ice shelf (Fig. 2). Samples were distributed as evenly as possible (given logistical constraints such as aircraft range and weather) down the centerline of each ice shelf (inshore/offshore). The eastern and western portions of the Ward Hunt Ice Shelf were also sampled, owing to this ice shelf's overall size and breadth. The Ward Hunt Ice Shelf was also accessible on foot, allowing supplementary sampling near Ward Hunt Island (Fig. 2a). In addition to the samples from each ice shelf, an auxiliary sample was taken on the Ward Hunt Ice Rise (a grounded portion of shelf ice) and another was taken in an area of multiyear landfast sea ice (MLSI; this ice type can be considered as a nascent ice shelf) to the west of the Ward Hunt Ice Shelf (Fig. 2a). In situ measurements of surface water pH, conductivity, and temperature were taken using portable instruments (Hydrolab Quanta, Hach Environmental, Loveland, CO in 2001 and pH/Con 10 Series, Oakton Instruments, Vernon Hills, IL in 2002) and GPS positions were recorded. Surface waters were sampled in 2-L HDPE bottles (rinsed three times with sample water) and kept cool and dark prior to subsampling and preservation at our field laboratory. Microbial mat type was noted at the time of sampling and classified into four categories: orange, matlet, sediment, and none, if no macroscopic mat or sediments were visible. This mat classification system was designed as a rapid, visual census only (Fig. 3). Microbial mat samples were scooped into clean plastic containers, kept cool and dark until they were processed (subsampled and preserved for transport to Quebec) in the field laboratory.
Taxon Relative Abundance
Subsamples of microbial mat were preserved with a gluteraldehyde/formaldehyde solution (Lovejoy et al., 1993) and refrigerated prior to analysis in Quebec. These samples were vortexed, diluted with water, dispersed by repeatedly passing them through a syringe and subsequently examined on an inverted microscope (Zeiss Axiovert 10) at 1000 times magnification (Villeneuve et al., 2001). Taxa were identified or assigned to a morphotype and their relative abundance was determined by counting at least 500 fields or, more commonly, enumerating no less than 200 natural units (individuals or colonies, depending on habit) per sample. Phototrophs were found to be the dominant group in terms of productivity and biomass in previous analyses (Mueller et al., 2005) and we therefore focused on cyanobacteria, green algae, diatoms, and other protists. We recorded microinvertebrates whenever encountered but we did not include heterotrophic bacteria, archaea, or viruses in our analyses. Diatoms were enumerated only if cell contents were visible. Due to the morphological variability of oscillatorian cyanobacteria, three morphotypes were used to classify filaments that were not identified to genus level. The morphotype Osc. 1 was defined as filaments up to 2 µm wide with a clear sheath and no discernable cross walls, Osc. 2 was as above but with cross walls clearly visible, and Osc. 3 was similar to Osc. 1 but with a diffuse, often ragged looking, sheath. For these filaments, the presence of phycobiliproteins was verified under green excitation epifluorescence microscopy and, in the same study, an HPLC analysis did not detect any bateriochlorophyll, which confirmed that these filaments were cyanobacteria and not green nonsulfur bacteria (Mueller et al., 2005). Given the uncertain taxonomy of oscillatorians, molecular analysis would be required to define these taxa at a more precise level (Castenholz, 1992). Our analysis therefore provides a first order estimate of phototrophic diversity and does not encompass the full biodiversity of the microbial mats, which could be revealed in future using molecular techniques.
Microbial Mat and Water Analyses
Microbial mats were subsampled for pigment determination and were stored frozen and in the dark for transport. In the laboratory, samples were thawed in 90% acetone∶water (v/v) and sonicated (Microson XL2000, Misonix, Farmingdale, NY) at approximately 10 W for two 30-s bursts with a pause of 30 s between sonication steps or samples were homogenized using a tissue grinder (Camafro R2R1, Wiarton, ON) for 1 min (there was no significant difference between these methods: paired t-test, t = 0.867, P = 0.415, n = 8). Pigment extraction took place in the dark for 1 h at 4°C and terminated with centrifugation at 4000 rpm for 5 min at 4°C. The supernatant was removed and chlorophyll a concentrations were determined at 663 nm and 750 nm, before and after acidification, while total carotenoid content was determined at 480 nm in a Cary 300 Bio UV-Visible spectrophotometer (Varian, Mulgrave, Australia) (Strickland and Parsons, 1972; Britton, 1985). This procedure was carried out three times to ensure complete extraction, and total chlorophyll a and total carotenoid content was determined by summing each extraction (Villeneuve et al., 2001). Total carotenoids were postcorrected for scytonemin (Garcia-Pichel and Castenholz, 1991) based on spectrophotometric analysis of similar samples. These analyses showed that the absence of such a correction would result in an overestimate of carotenoid content by 57% (SE = 4%, n = 70). Meltpond water (420 ml) was subsampled and filtered onto GF/F equivalent filters (AMD Manufacturing, Mississauga, ON) and preserved frozen in aluminum foil for transport. These plankton chlorophyll a samples were subsequently extracted with boiling ethanol extraction and measured using a model 450 Sequoia-Turner fluorometer (Turner Corporation, Mountain View, CA) (Nusch, 1980).
Microbial mat organic content was determined in subsamples collected from a known surface area and frozen for transport. In the laboratory, the sediments were weighed and then dried to constant weight at 95°C. The mass of organic and inorganic fractions were determined by loss on ignition at 500°C for 7 h.
Nutrients (NH4+, NO3−, NO2−, soluble reactive phosphorus [SRP], dissolved organic and inorganic carbon [DOC and DIC], and total nitrogen and phosphorus) were determined from filtered (cellulose acetate, 0.2 µm) surface water subsamples by the National Laboratory for Environmental Testing (NLET), Environment Canada, Burlington, Ontario. Methods were UV digestion and infrared detection (dissolved organic and inorganic carbon, respectively), ion chromatography (nitrate), and automated colorimetry (total nitrogen and ammonium by the indophenol method; nitrate/nitrite by cadmium reduction; soluble reactive phosphorus by the ammonium molybdate/stannous chloride method). Sample water (420 ml) was filtered through GF/C equivalent filters (AMD Manufacturing) to remove particulate carbon and nitrogen, which were determined by elemental analysis at NLET. Particulate phosphorus was calculated by subtraction of unfiltered and filtered total phosphorus determinations (NLET: stannous chloride colorimetry). Major ion concentrations at some sample sites were also determined by NLET using ion chromatography (Cl−, SO42−), atomic absorption (Na+, K+, Mg2+, Ca2+,) or heteropoly blue colorimetry (SiO2). Bicarbonate concentration was determined from pH and DIC concentration (Wetzel and Likens, 2000). If measurements were below the limits of detection, not detected as substituted by half the detection limit for statistical tests (detection limits were: 5 µg NH4+–N L−1 for NH4+, 5 µg NO3−–N L−1 for NO3−, 1 µg NO2−–N L−1 for NO2−, 50 µg L−1 for Ca2+, 200 µg L−1 for SO42−, 20 µg L−1 for SiO2).
Satellite Remote Sensing
The spatial extent of each ice shelf was determined using radiometrically-corrected RADARSAT-1 synthetic aperture radar (SAR) images that were projected to a common geographical system (UTM zone 18N) in ArcMap 8.0 (ESRI, Inc., Redlands, CA). Ice type identification was based on our field observations, air photographs, and SAR images, following Lemmen et al. (1988) and Jeffries (1986a, 1986b, 1992a, 2002). In this paper, we define shelf ice as floating ice having a freeboard of 2 m or more (Jeffries, 1992b). In contrast, MLSI is relatively thick and stable sea ice with a freeboard of less than 2 m. Meteoric ice includes glacial ice and iced firn. Iced firn is a term specific to the Ellesmere ice shelves that refers to the product of a rapid transformation of snow to ice by meltwater percolation and refreezing (Marshall, 1955). It accumulates in place but is not subject to flow as occurs on glaciers. Marine ice includes basement ice (sea or brackish ice, which accreted on the underside of the ice shelf; Lyons et al., 1971) and MLSI. A nonfloating ice feature related to ice shelves is called an ice rise, which is formed by the grounding of an ice shelf on subjacent land (Lyons et al., 1972).
Polygons were traced over SAR imagery raster layers using vector National Topographic System (NTS) 1∶250,000 scale maps (Natural Resources Canada: 120F, 120G, 340E, 340F, 340H) to confirm shoreline position. Ice shelves were digitized from the following image numbers and reflect the spatial extent at the latest image (Serson Ice Shelf: r1_24186_211, June 22, 2000; r1_22728_211, March 12, 2000; Petersen Ice Shelf; r1_23185_213, April 13, 2000, Milne Ice Shelf: r1_29859_214, July 25, 2001; Ayles Ice Shelf: r1_37462_215, January 07, 2003; r1_34062_209, May 14, 2002; Ward Hunt Ice Shelf: r1_35605_210, August 30, 2002; r1_35477_208, August 21, 2002; Markham Ice Shelf: r1_35261_215, August 6, 2002). All of these images were obtained in standard beam mode with the exception of one fine beam image (r1_35605_210). The image used to delineate Petersen Ice Shelf was a low resolution product and was not suitable for ice type discrimination. Surface ice type (marine vs. meteoric) was determined for the Ward Hunt Ice Shelf by performing the following operation on the radar backscatter amplitude from a Radarsat image at peak melt (A: r1_35261_215, August 6, 2002) and from an image recorded in winter (B: r1_37462_215, January 7, 2003) on a pixel by pixel basis in ENVI 3.2 (Research Systems Incorporated, Boulder, CO):
The resultant image was digitized as above keeping ice types in contiguous zones. Meteoric ice, which has a relatively small radar backscatter during the melt season and a relatively large radar backscatter in the winter, had a digital number of less than 128 in this classification scheme. Marine ice areas were denoted by lighter areas (digital numbers of 129 to 255) due to a high radar backscatter in August and a low backscatter in January.
In 2001, six helicopter transects were flown at a constant speed, bearing and altitude across the ice shelves and perpendicular to the Ellesmere Island shoreline (Fig. 2; Ward Hunt West, Ward Hunt Middle, Ward Hunt East, Serson, Milne, and Markham). The transects were flown along the centerline of each ice shelf with extra transects halfway between the center and shorelines for the Ward Hunt Ice Shelf due to its greater size and breadth. The Ayles and Petersen ice shelves were not surveyed in this manner due to logistical constraints. Images were extracted from a digital video sequence (every 40th frame, or 1 image every ~1.3 s, which ensured that images did not overlap). The images were de-interlaced and the cover of each class (sediment, water, and snow/ice) was estimated to the nearest tenth or, if present at less than 0.5 tenths cover, was recorded as a trace amount and was represented by 0.25 tenths in subsequent calculations.
The mean percentage area occupied by sediment was scaled-up to the size of the ice shelves to estimate the potential extent of microbial habitat over all the ice shelves. An estimate of the mass of chlorophyll a, organic matter and sediment on each ice shelves was made by combining potential habitat extent and mean determinations of these quantities (±SE).
Point transect analyses over different areas on Ward Hunt and Markham ice shelves were undertaken to provide a detailed examination of potential habitat at a smaller scale in marine and meteoric ice areas (Fig. 2). The substrate was characterized at 10-cm intervals by visually recording the presence or absence of snow and sediment. These transects were between 120 and 360 m long and ran perpendicular to ridges and troughs. At the Ward Hunt Ice Shelf marine ice area, the main point transect was flanked by two auxiliary transects running parallel to and positioned at approximately 50 m from the main transect.
Several approaches were used to compare ice shelves, ice types, and mat types. Univariate analyses were undertaken for several variables using one-way ANOVAs or t-tests and nonparametric ANOVAs or t-tests when warranted (non-normal distributions and/or unequal variances). For multivariate analysis, variables with more than several data points missing were removed and the remaining missing data points (25 out of 684 data points) were replaced by the mean for that particular variable (Legendre and Legendre, 1998). Non-normally distributed environmental and pigment data were normalized using a Box-Cox transformation (conductivity, DIC, total nitrogen, total phosphorus, soluble reactive phosphorus λ = 0, chlorophyll a, and carotenoids λ = 0.25, NH4+ λ = −0.25). Some variables were not transformable and therefore remained unaltered (temperature, DOC, NO3−, and NO2−). Taxon relative abundance data were not transformed. A detrended correspondence analysis (DCA) was undertaken using the abundance data and the dispersion was found to be <2, which indicated that a linear ordination technique should be employed (ter Braak and Smilauer, 1998). Therefore, redundancy analysis (RDA) was used to investigate the association between environmental and taxon relative abundance data. An overall solution was calculated using all 28 samples that were enumerated microscopically and 13 environmental variables. This was followed by a manual forward selection of variables that yielded a more parsimonious model.
To examine differences between ice shelves, ice types and microbial mat types, a one-way analysis of similarities (ANOSIM, Primer 8.5) was used (Clarke and Warwick, 1994). Similarity matrices were constructed using Bray-Curtis distances for taxon relative abundance data and normalized Euclidean distances for environmental and pigment data (the latter matrix contained several samples that did not have equivalents in the taxon abundance matrix). The ANOSIM is a multivariate procedure that tests for differences between groups determined a priori (ice shelf, ice type, mat type). The average ranked similarity value of pairs of replicates (R statistic) is calculated and the probability of obtaining a higher R with a given dataset is based on 999 randomly selected permutations of the dataset.
Ice Shelf Comparisons
The six largest remnant ice shelves in the Canadian High Arctic totaled 1043 km2 at the time of observation, and 8% of this surface provided a sediment cryohabitat for microbiota. The area of each ice shelf and their constituent ice types are given in Table 1. All the ice shelves except Milne were found to have substantial areas of marine ice exposures. Transect data showed that Ward Hunt and Markham ice shelves contained the greatest amounts of potential microbial mat habitat (10% and 44%, respectively) and this translated to high estimates for total chlorophyll a and organic matter relative to the other ice shelves (Table 1). This trend was also reflected in the estimates of sediment load for each ice shelf.
Comparison of the six Canadian Arctic ice shelves. Mean values followed by ±SE in parentheses.
The five ice shelves had similar values for many of their limnological properties (Table 2) and differed greatly in others (Fig. 4). The western ice shelves had less chlorophyll a, carotenoids, and organic matter per unit area than the Ward Hunt and Markham ice shelves (Fig. 4b; ANOVA results: H = 12.6, P = 0.005; H = 15.8, P = 0.001; F = 5.9, P = 0.003; respectively). This pattern was mirrored by conductivity, but there were no significant differences in sediment load (H = 5.8, P = 0.22) and pH (F = 1.4, P = 0.25) among the ice shelves (Fig. 4a). Nutrient concentrations were significantly different among the five ice shelves with the exception of total nitrogen (H = 6.3, P = 0.18) and ammonium (H = 3.9, P = 0.42) concentrations. DIC concentrations (H = 10.1, P = 0.039) were highest in Markham Ice Shelf sites (Table 2), while DOC (F = 3.99, P = 0.016) concentrations were highest in both Markham and Ward Hunt ice shelves. Nitrate (H = 11.4, P = 0.023) and nitrite (H = 16.3, P = 0.003) concentrations were highest in Serson Ice Shelf samples and these variables were lowest in the easternmost ice shelves. Total phosphorus (H = 9.5, P = 0.05) and soluble reactive phosphorus (H = 11.0, P = 0.027) concentrations were highest on Ward Hunt Ice Shelf (Table 2) and did not vary systematically across the ice shelves. For samples taken from marine ice, ionic composition indicated evaporative concentration (high Mg2+/Ca2+) consistent with a seawater source of major ions. Most samples taken from meteoric ice were enriched with magnesium relative to sodium, a proxy of rock interaction (Vincent and Howard-Williams, 1994), which suggests that in situ chemical weathering of surface sediments overshadowed the geochemical influence of marine aerosols in these environments (Fig. 5).
Ice shelf water column nutrients, major ions, and suspended particulates. Averages followed by standard error in parentheses; sample sizes are as follows: Markham Ice Shelf major ions 2, nutrients 6; Ward Hunt Ice Shelf major ions 15, nutrients 21 (except PP: 19 and Total P: 18); Ayles 1; Milne Ice Shelf major ions 3, nutrients 4; Serson Ice Shelf major ions 2, nutrients 4.
The dominant taxa in the ice shelf microbial mats were Oscillatorian cyanobacteria, specifically morphotypes Osc. 3 and Osc. 1 followed by Osc. 2, the diatom Chamaepinnularia begeri, and the palmelloid green alga cf. Bracteacoccus sp. The former three taxa are all filamentous in habit and produce extracellular polymeric substances (EPS) which suggests that the overall structure of the microbial mat is influenced by the intertwining of their trichomes and the cohesive, binding effect of the EPS (see Fig. 3e in Vincent et al., 2000). In all, 32 taxa were found (Table 3), and samples frequently contained empty diatom frustules of taxa that were observed with cellular contents elsewhere in the sample or in other samples in this study. The planktonic diatom Asterionella sp. and unidentified cysts were occasionally observed in samples but with no visible cellular contents. A list of each observed taxon and their relative abundance is given in Table 3. There were no significant differences between the ice shelves with regards to the number of species (F = 2.3, P = 0.10) or Simpson's diversity index (F = 2.6, P = 0.072).
Abundance of taxa found in Arctic ice shelf microbial mats. All samples: average and standard error for all samples. Ice shelf, Ma: Markham Ice Shelf, WH: Ward Hunt Ice Shelf, IR: Ward Hunt Ice Rise, Di: MLSI near Discovery Ice Rise, Ay: Ayles Ice Shelf, Mi: Milne Ice Shelf, Se: Serson Ice Shelf. Ice Type: Marine: basement ice and MLSI, Meteoric: iced firn and glacier ice. Mat Types: Matlet (small flakes of microbial mat), Orange (thick mat with orange surface layer), Sediment (no discernable biological aggregates). Osc. = oscillatorian cyanobacteria. Taxon relative abundance is coded as follows: 5 indicates dominant (>50% of total algal counts in the sample), 4 indicates subdominant (50–25%), 3 indicates common (25–5%), 2 indicates rare (<5%), 1 indicates not detected (0%).
The multivariate test ANOSIM, revealed that the microbial mat relative abundance was significantly different (Global R = 0.142, P = 0.048) among the five ice shelves. A pairwise comparison indicated that Markham and Serson ice shelves have the most dissimilar communities (P = 0.012) while Markham and Milne ice shelves also had significantly different microbial communities (P = 0.036). For environmental variables, as well as organic matter content and pigment data, the Global R was 0.408 (P = 0.001) with four significantly different pairs (Markham and Milne, P = 0.01; Markham and Serson, P = 0.005; Ward Hunt and Milne, P = 0.002; Ward Hunt and Serson, P = 0.003).
Direct ordination of the taxon relative abundance data and the 13 environmental variables revealed a non-significant association between these data matrices, with 53% of the relative abundance explained by variation in the environmental variables (P = 0.32). After a manual forward selection procedure, a more parsimonious model using 4 environmental variables (Conductivity, NO3−, NH4+, and SRP) was constructed that explained 40% of the relative abundance variance (P = 0.003).
Ice Type and Mat Type Comparisons
Samples from marine ice areas had significantly higher (P < 0.05) water column nutrient concentrations (except total phosphorus), conductivity, and pH as well as biodiversity indicators, microbial mat pigment concentrations, and organic content than their meteoric ice counterparts (Table 4). The ANOSIM result (Global R = 0.447, P = 0.001) also indicated that these ice types differed significantly in their habitat characteristics. The taxon relative abundance was also significantly different between these two habitat types (Global R = 0.187, P = 0.016).
Comparison of ice shelf ecosystem characteristics based on ice type and mat type. Averages followed by standard error in parentheses, for n = 13–14 (meteoric ice), 20–24 (marine ice), 9–11 (orange mat), 7–10 (matlet), 11–12 (sediment), 5 (no mat).
The microbial mat types had significantly different pigment and organic matter concentrations; however, the sediment concentration was not significantly different among these groups (three-level ANOVA, not including the no mat category). A four-level ANOVA revealed that water that was not underlain by microbial mats was significantly colder, less saline, and had lower pH than water overlying sediment. However, among the three mat types, conductivity was significantly different, whereas temperature and pH did not differ significantly. With the exception of total phosphorus, nutrient levels were significantly different from each other among the four groups and among the three mat types there were significant differences in both the number of taxa and Simpson's diversity index (Table 4). The ANOSIM analysis indicated a significant difference among the three mat types for relative abundance (Global R = 0.463, P = 0.001) with significant pairwise differences between orange mat and matlet (P = 0.002) as well as orange mat and sediment (P = 0.001). Significant differences were also found for both the three-level (Global R = 0.334, P = 0.001) and the four-level (Global R = 0.488, P = 0.001) ANOSIMs on environmental and pigment data among microbial mat categories. All pairwise comparisons were significantly different for each of these tests.
Over-flight transect estimates of percent sediment cover (Markham, 44%; Ward Hunt Middle, 19%; Ward Hunt East, 2%; Ward Hunt West, 12%; Milne, 2%; Serson, 3%) and meltwater cover (Markham, 17%; Ward Hunt Middle, 14%; Ward Hunt East, 22%; Ward Hunt West, 16%; Milne, 18%; Serson, 14%) indicated that substantial differences exist among ice shelves with regards to potential microbial habitat. The percent cover of sediment varied markedly throughout each transect, but overall patterns can be seen in Figure 6. Markham Ice Shelf had very little sediment cover at its northern and extreme southern end (Fig. 6a), the Ward Hunt Ice Shelf had greater sediment cover in the southern portions of all transects, and only the central transect, which terminated at Ward Hunt Island showed an increase in sediment cover farther north (Fig. 6b, 6c). The Milne and Serson ice shelves had very little sediment cover throughout, yet the southern portions of these transects did have slightly higher sediment cover (Fig. 6d). When sediment on these latter ice shelves was present in high percentages, it was of glacial origin (moraines), which explained large departures from the overall trend (e.g., Fig. 6d; Serson Ice Shelf, image number 260 on the abscissa).
On a smaller scale, sediment cover point transects also showed a high degree of variability in the ice-shelf surface cover. The point transect on the Ward Hunt Ice Rise revealed that this environment, while not strictly part of the ice shelves, also contained sediments, albeit at very low percentages (Fig. 7d). Meteoric ice (Fig. 7d, 7e) had less sediment exposed at the surface due to snow cover than the other point transects.
Overall System Properties
The major remnants of the Ellesmere Ice Shelf surveyed in this study are large masses of ice with substantial yet variable surface water and sediment cover. At the times of sampling the meltwaters had temperatures near freezing, near neutral pH values, and oligotrophic nutrient levels, with the exception of elevated phosphorus levels at some sites (Table 1). The water column seston was also in sparse concentration as indicated by particulate organic carbon and nitrogen and planktonic chlorophyll a levels (Table 1). In contrast to the water column, the sediments on the ice surface contained large amounts of organic material and algal pigments. An enumeration of these benthic microbial communities revealed numerous taxa including cyanobacteria, green algae, diatoms, other protists, and microfauna that were common to all habitats in this ensemble of ice shelves. Despite these similarities, there were important and systematic differences across the Ellesmere Ice Shelf cryoecosystem, as outlined below.
Total sediment estimates were higher than measurements made by Crary (1958), suggesting that sediment concentration may have increased since these analyses due to ongoing surface wasting of the ice shelf (Braun et al., 2004; Mueller and Vincent, 2006), which would also act to bring immured microbial mats to the surface (Mueller et al., 2005). Spatial variability and sampling bias limit the accuracy and precision of the estimates; however, our results allow a first order approximation of total biomass that can be compared to other ecosystems. In one study, samples from the Central Arctic pack ice averaged an organic carbon content of 112 mg C m−2 (Gradinger, 1999), which is more than 2 orders lower than our spatially averaged estimate of ice shelf organic matter. Chlorophyll a concentration was 14 mg m−2 averaged over all the ice shelf remnants in this study. This compares to a mid-summer chlorophyll a concentration in pack ice (>2 m thick) of 1.6 mg m−2 (Gradinger, 1999), although far higher chlorophyll a biomass may be attained beneath landfast sea ice (Michel et al., 2002). Under optimal conditions, Arctic ice shelf microbial mats have a gross photosynthetic rate of 129 mg C m−2 d−1 when averaged over the ice shelf (Mueller et al., 2005). This productivity is five times higher than average rates per unit area for the Central Arctic pack ice (Gradinger, 1999) and over ten times higher than the productivity of polar hypolithic communities (Cockell and Stokes, 2004). However, the total surface area of ice shelves is small relative to the 5.5 million km2 of sea ice found in the Arctic Ocean (Comiso, 2002), and also relative to the vast expanse of Arctic polar desert.
Consistent with our hypothesis regarding environmental gradients, there was a striking gradient in conductivity, biomass, and pigment concentrations from west to east along 250 km of the coastline. This broad gradient is likely due to the more common occurrence of marine ice with high sediment content in the eastern two ice shelves and the greater proportion of meteoric ice with low sediment content in the west. It is also possible, however, that this trend was influenced to some degree by sampling bias towards marine ice sites on the Ward Hunt Ice Shelf (Fig. 2a). Our results indicated that marine ice types were associated with higher amounts of organic matter and pigments as well as increased nutrient concentration and conductivity than meteoric ice types. Sediment mass per unit area of microbial mat is not necessarily greater for marine ice; however, higher sediment percent cover in the marine ice areas of Ward Hunt and Markham ice shelves makes them more important with regards to potential microbial habitat (Table 1).
We suggest that this east-west gradient is due to a combination of two factors: (1) sediment delivery onto the ice shelf, affected by the local availability of sediment and the efficiency of delivery mechanisms such as eolian transport and ice-shelf grounding (see below); and (2) local climatological differences in snowfall, fog cover, ice shelf albedo, and thus ablation rates (Mueller and Vincent, 2006). Highest surface coverage will occur where the supply rates of sediment are high, and where the sediment is exposed and concentrated on the ice surface due to low albedo and perennial net surface ablation (Crary et al., 1955). In this manner, ice-type surface exposures are related to a temporally integrated mass-balance history of the ice shelf. Over time, if the surface mass balance has been negative and basal mass balance has been positive, then marine ice will become exposed at the surface of the ice shelf. If the converse is true, then the surface ice type will be derived from precipitation. Given this set of processes, the exposure of either marine or meteoric ice would reflect historical differences in climatic parameters such as precipitation, cloud cover, albedo, and air temperature. Sediment concentration on the ice shelf surface serves in turn to enhance ablation through reduced albedo and provides more habitat for microbial mats.
The lithology of coastal areas surrounding the ice shelves likely determines the availability and composition of sediments. Markham and Ward Hunt ice shelves are surrounded by the Cape Columbia Group (biotite, feldspar, hornblende), the M'Clintock Group (andesitic and basaltic lava, breccia, tuffs, greywacke) as well as outcrops of limestone and sandstone. The western three ice shelves are bordered by the Cape Columbia Group as well, but also by harder rocks such as granite and gneiss (Christie, 1957; Trettin, 1991). The relative surface area of outcrops and differences in the friability and weathering of these formations determines the grain size distribution and availability of sediment. Prevailing winds move from west to east along the northern coast of Ellesmere Island (Crary, 1960). For eolian transport of these sediments to the ice surface, source areas must be aligned with wind patterns and there must be sufficient fetch and appropriate topography to move sediments efficiently (winds crossing steep-walled fiords will eddy, thereby decreasing the linear transport of materials). Finally, bathymetry may play a role in determining whether marine ice is exposed at the surface of an ice shelf. Marine sediment can be delivered to the ice-shelf surface via basal freezing and subsequent surface ablation. In this manner, intact marine mollusks, sponges, and sediments are raised to the surface of the ice shelf over a period of several decades to centuries. This has been noted on the ice shelf to the east of Ward Hunt Island (Crary, 1960; Jeffries, 1992b), and we have observed the remains of marine invertebrates on the surface of southern Markham Ice Shelf and on the Ward Hunt Ice Shelf to the southwest of Ward Hunt Island. This mechanism is also known in Antarctica (Debenham, 1920; Vincent, 1988), and it occurs in areas where the bottom of the ice shelf touches the sea floor. A high proportion of sediment on the ice surface will also encourage greater surface ablation via an albedo feedback, thereby accentuating this sediment delivery mechanism. Although little is known about the near shore bathymetry of the region (see Fig. 1 in Crary, 1956), it is probable that certain areas near the shore of Ward Hunt Island and Ellesmere Island have a suitable bathymetry for this to play a role in the delivery of sediments to sections of the Ward Hunt and Markham ice shelves.
From an inshore/offshore perspective, sediment cover patterns are consistent with assumed gradients caused by sediment source areas and modes of transport. Over-flight transects confirmed that sediment distribution is skewed towards coastal areas of the ice shelves with the exception of the termini of coalesced glacier tongues that have transported debris far from shore.
Gradients also exist in the biological domain of the Ellesmere ice shelves, as might be expected from such large gradients in the physical characteristics of this cryoecosystem. Greater chlorophyll a and organic content was associated with the more saline ice with higher sediment content in the two eastern ice shelves, whereas the western ice shelves had less sediment and lower biomass content. A multivariate approach showed that the relative abundance was strongly associated with environmental variables, which was supported by a significant RDA ordination with the variables conductivity, soluble reactive phosphorus, ammonium, and nitrate. Using ANOSIM, the taxon relative abundance data were found to be significantly different between marine and meteoric ice types; therefore, it is not surprising that the RDA solution shows the influence of conductivity gradients. The oligotrophic nutrient status of the water columns and the inclusion of nutrients in the RDA solution suggest that nutrient supply may be a limiting factor for certain microbial mat species. However, there is evidence that the nutrient status within microbial mats is much less impoverished than the overlying water column and that microbial mat organisms may not in fact be nutrient limited (Bonilla et al., 2005; Mueller and Vincent, 2006). In this case, if diffusion and water column mixing were comparable among samples, the water column nutrient concentrations may be influenced by microbial mat interstitial waters and may serve as a proxy for the relative nutrient status of the mats.
In the present study, we classified mat types in discrete categories, while recognizing that these communities span a continuum in structure and pigmentation. However, there were important taxonomic differences between the matlet, orange, and sediment mat types, which suggest that these visually discernable mat characteristics may continue to serve as convenient and useful categories for future work. It is possible that these microbial mat types are related to the ice type of the sample site. For instance, orange mat was not found on meteoric ice whereas matlet and sediment mat types were found in both marine and meteoric ice types. The biodiversity and nutrient data suggest that a progression may exist in microbial mat development. Inorganic sediment may first become colonized with a number of pioneer taxa (notably Osc. 3, Osc 1, and Ancyclonema nordenskioldii) and over time this may develop into matlets with more biodiversity and biomass. Finally, if conditions are right, an orange surface layer may develop in late succession with an increase in prominence of taxa such as Chamaepinnularia begeri, cf. Bractaeococcus sp., Navicula cf. phyllepta, and cf. Pseudanabaena sp. Conversely, the microbial mat types may represent three divergent end members that are not successionally related to each other. As well, the ANOSIM results for ice type and mat type were significantly different for both taxonomic and environmental data, suggesting these categories are meaningful and possibly related. It is reasonable to assume that the environmental gradients exert some degree of control on relative abundance of taxa; however, the reverse may also be true in that certain taxa or mat types may influence local conditions, for example through nutrient release, pH control and albedo effects.
Ice Type Determination
The overall area of ice shelves has not changed substantially since the last determination was made in 1998/1999 (Vincent et al., 2001). Two relatively minor changes are the loss of 6 km2 of shelf ice and 20 km2 of MLSI from the eastern seaward edge of the Ward Hunt Ice Shelf in 2002 (Mueller et al., 2003). This was preceded by the break-up of the ice shelf into two fragments, although they are still held in place by islands and ice rises. Other discrepancies from the previous literature can be explained by the ice type reclassifications. For example, the inner portion of the Milne Ice Shelf is known to be epishelf lake ice (Jeffries, 2002; Van Hove, 2005) and was therefore not included in this study. As well, we have recently determined that the dark ice at the southern end of the western Ward Hunt Ice Shelf was ice shelf covered by water during our over-flight transects (Figs. 1b, 2a). We therefore include this area as part of the Ward Hunt Ice Shelf in the present study.
Marine and meteoric ice areas were relatively easy to demarcate; however, in the case of one sample in the southeast corner of the Ward Hunt Ice Shelf (Fig. 2a), the ionic composition was closer to marine ice meltwater than any of the other meteoric ice water samples (see Fig. 5). This likely reflects an error in ice type delineation, probably due to a prevalence of ice type unconformities (Lyons et al., 1971) in this particular region. Two other outliers in this graph were from the glacial section of Serson Ice Shelf and the floating portion of the Milne Glacier (see Fig. 5). These samples may have differed from other samples underlain by meteoric ice due to a high degree of flushing with glacial surface meltwaters that had not contacted rock debris and were therefore marine-influenced relative to other meteoric ice meltwater that had longer contact times with eolian sediment.
The Effects of Habitat Fragmentation
Our results show that gradients exist across the extent of the Ellesmere Ice Shelf cryoecosystem, although these spatial trends are discontinuous because the former ice shelf is now fragmented into several remnant cryohabitats. According to ecological theory, habitat fragmentation affects population dynamics through a reduction in total habitat, fragment size, and connectivity, yet effects due to habitat fragmentation are highly species specific, depending on their dispersal ability and habitat specificity. For this reason, habitat fragmentation is a concern for species with narrow ecological niches, such as certain plants, mammals and birds (e.g., Terborgh et al., 2001) and has been little explored in microbial ecology. The physical disintegration of the largest known ice shelf in the Northern Hemisphere can also be viewed as habitat fragmentation since the remaining ice shelves today form an archipelago of isolated habitats. However, many microbial mat photoautotrophs that inhabit cryoecosystems are easily wind-dispersed and are physiological generalists with a relatively wide optimum range (Mueller et al., 2005) that may not be adversely affected by habitat fragmentation. It is not possible to compare the ecological properties of the original Ellesmere Ice Shelf to the ensemble of present-day remnants, however these two competing hypotheses may be evaluated by comparing biodiversity among the isolated fragments.
There were no significant differences in species richness or diversity among these remnants, suggesting that fragment size does not control the biodiversity of the microbial mat phototrophs as would be expected if habitat fragmentation were deleterious. Furthermore, the ice shelf cryohabitat is naturally heterogeneous at the sub-ice shelf scale and microbial mats are distributed in patches as shown in Figure 7. Therefore, the latter hypothesis appears to be more plausible, and there is likely considerable connectivity and dispersal between many of the ice-shelf fragments as well as between habitat patches (especially in the direction of the prevailing wind). In addition, habitats from Ellesmere Island serve as a source for ice-shelf biota. Most of the organisms found in ice mats are present on nearby islands in either microbial mats from tundra ponds (Villeneuve et al., 2001), glacial cryohabitats (Mueller et al., 2001), or soil ecosystems (Elster et al., 1999), although they are not organized into the distinctive assemblage types and mat architectures observed on the ice shelves. It is also possible that other factors masked the effects of habitat fragmentation in our study as was observed in studies on High Arctic soil microarthropod communities (Coulson et al., 2000) and soil decomposers in Finland (Rantalainen et al., 2004), where resource quality and environmental stress overshadowed expected community changes due to habitat patchiness.
The cryosphere in general provides many microbial habitats throughout the biosphere, and the availability of these habitats has varied greatly throughout geologic time. During interglacial epochs, the cryosphere becomes physically fragmented until cryohabitats are relegated to high altitudes and latitudes. This habitat fragmentation may be detrimental to microbial biodiversity and population dynamics or, as we have observed in Arctic ice shelves, the dispersal abilities and broad tolerance ranges of these organisms may overcome the effects of habitat fragmentation and loss and allow the development of similar communities at multiple, noncontiguous sites.
Our observations across the 250-km extent of the Ellesmere Ice Shelf cryoecosystem show that each of the five remnants sampled contains a rich microbial flora. This amounts to an estimated 34 Gg of organic matter, with average areal productivities that are well above values in the Central Arctic pack ice. The surface exposure of meteoric versus marine ice is an important factor in structuring these ice shelf assemblages. However, the optimal habitat appears to be in areas where inorganic sediments have accumulated over exposed marine ice. Regions where such conditions exist, on Ward Hunt and Markham ice shelves, differ greatly from areas elsewhere on the ice shelves that are characterized by sparsely distributed sediment and matlet microbial mats in isolated cryoconite holes and glacial debris. Our results underscore the broad gradient of environments that can be found across this ensemble of northern ice shelves and our RDA analysis showed that microbial mat relative abundance was significantly associated with environmental variability. There was no evidence that the 20th century fragmentation of the Ellesmere Ice Shelf cryohabitat has affected biodiversity and richness of the microbial mat communities, but the continued existence of this extreme Arctic ecosystem is vulnerable to ongoing climate change.
We thank the Natural Sciences and Engineering Research Council of Canada for a graduate scholarship (DRM) and a discovery grant (WFV), the Polar Continental Shelf Project (this is PCSP/ÉPCP publication number 01405), the Canada Research Chair in Aquatic Ecosystem Studies, the Networks of Centres of Excellence program ArcticNet and the Northern Scientific Training Program for financial and logistical support. We acknowledge the Alaska Satellite Facility at the University of Alaska Fairbanks and a NASA data grant for providing RADARSAT images. Laboratory and field assistance was provided by Marie-Josée Martineau, Katie Breen, Sébastien Roy, Denis Sarrazin, and Jeffrey Kheraj. Sylvia Bonilla assisted with taxonomic identification. We acknowledge Vicki Sahanatien, Ross Glenfield, and the staff of Quttinirpaaq National Park for their support and use of facilities. We thank Dominic Hodgson, Martyn Tranter, and an anonymous reviewer for helpful comments on the manuscript.