The formation of many arctic wetlands is associated with the occurrence of polygon-patterned permafrost. Existing scenarios to describe and explain surface landforms in arctic wetlands (low-center and high-center polygons and polygon ponds) invoke competing hypotheses: a cyclic succession (the thaw-lake hypothesis) or a linear succession (terrestrialization). Both hypotheses infer the predictable development of polygon-patterned wetlands over millennia. However, very few studies have applied paleoecological techniques to reconstruct long-term succession in tundra wetlands and thereby test the validity of existing hypotheses. This paper uses the paleoecological record of diatoms to investigate long-term development of individual polygons in a High Arctic wetland. Two landform processes were examined: (1) the millennial-scale development of a polygon-pond, and (2) the transition from low-center to erosive high-center polygons. Diatom assemblages were quantified from habitats associated with contrasting landforms in the present-day landscape, and used as an analog to reconstruct past transitions between polygon types. On the basis of this evidence, the paleoecological record does not support either of the existing models describing the predictable succession of polygon landforms in an arctic wetland. Our results indicate a need for greater paleoecological understanding, in combination with in situ observations in present-day geomorphology, in order to identify patterns of polygon wetland development and elucidate the long-term drivers of these landform transitions.
Introduction
The occurrence of permafrost at high latitudes enables the widespread development of arctic and subarctic wetlands (Tarnocai and Zoltai, 1988; Bliss, 1997); i.e. in low-lying areas, drainage of snowmelt is impeded by perennially frozen ground. In the High Arctic such wetlands are frequently referred to as oases, being comparatively rich in plant diversity with high productivity (Bliss and Matveyeva, 1992), and they provide important summer habitat for muskox and large populations of migratory birds (Bliss and Matveyeva, 1992; Bliss, 1997). The cold and wet soil conditions in arctic wetlands preclude decomposition, and as organic matter accumulates each year, the frost table rises upwards, incorporating peat into the aggrading permafrost (Billings, 1987; Gorham, 1991; Michaelson et al., 1996; Vardy et al., 2000). Peat-rich soils associated with arctic wetlands have contributed a net sink for carbon during the Holocene and are estimated to store >97% of the tundra carbon reserve comprising ca 180−190 × 1015 g of soil-C, (Post et al., 1982; Oechel and Vourlitis, 1994). Given the incidence of suitably wet soil moisture conditions, the formation of ice wedges has resulted in the development of patterned ground over vast areas of the arctic landscape (Leffingwell, 1915; Lachenbruch, 1966; Péwé, 1966; Mackay, 1990, 2000). The accretion of peat and eolian sediments is therefore accompanied by the upward growth of syngenetic ice wedges (Harry and Gozdzik, 1988; Mackay, 1990, 1999) to form polygon-patterned wetland (Tarnocai and Zoltai, 1988; Bliss, 1997).
The landforms associated with polygon-patterned wetlands can be broadly divided into three types (Péwé, 1966; Billings and Peterson, 1980): low-center polygons (raised polygonal ridges surrounding a lower, wetter center), polygon ponds (raised polygonal ridges surrounding a pool), and high-center polygons (a raised, domed center surrounded by a polygonal network of troughs) (Fig. 1). Scenarios describing the long-term development of these periglacial landforms are based primarily on observations made in the present-day landscape of what are inferred to be chronosequence stages. Accepted explanations for the long-term development of arctic wetlands invoke directional change (cf. Svoboda and Henry, 1987; Bliss, 1997), though they are contrasting in pattern. For example, the widely accepted “thaw-lake cycle hypothesis” (Fig. 2) invokes a cyclic succession of vegetation over ca. 3000 yr (Britton, 1957; Billings and Peterson, 1980), though it contrasts with an inferred linear sequence of terrestrialization (Zoltai and Tarnocai, 1975) in which peat accumulation in polygon centers is thought to cause the gradual and linear transition from wet low-center polygons to dry high-center polygons. Elements of the thaw-lake cycle are supported by studies in paleoecology and periglacial geomorphology: e.g. the gradual inundation of low-center polygons as ridges become higher (Seppälä et al., 1991) and the regeneration of ice wedges on drained lake beds (Mackay, 1986, 1999). In contrast, a linear sequence of terrestrialization involving the development of high-center from low-center polygons has been supported by paleoecological reconstruction (Ovenden, 1982; Vardy et al., 1997). Determining the relative importance of these contrasting successional patterns (e.g. linear versus cyclic) is an essential prerequisite to correctly understand the long-term function of arctic terrestrial wetlands. The shift between landforms is associated with major hydrologic and vegetation differences (cf. Fig. 1) and can have significant impact on ecosystem function (i.e. the C-flux; cf. Billings et al., 1982, 1983; Johnson et al., 1996).
This paper attempts to confirm or refute the validity of linear versus cyclical models describing long-term succession in a High Arctic wetland, using diatoms as a tool for paleoenvironmental reconstruction (cf. Battarbee et al., 2001). A quantitative description of diatoms from landforms in the present-day wetland is used as an analog to reconstruct past hydrologic change, i.e. the development of a polygon-pond and a series of local shifts to drier high-center polygon conditions. We thus examine the sequence of hydrologic change for key stages during landform transitions, and ask:
(1) Does the paleoecological development of a polygon pond match with a predictable hydrologic sequence consistent with the transition from low-center polygon, to pond, to thaw lake? (i.e. stages C → D → E; Fig. 2)
(2) Is the transition to a high-center polygon consistent with slow, linear terrestrialization? Alternatively, is the transition to a high-center polygon consistent with a preceding period of lake formation? (i.e. stages E → F; Fig. 2)
Study Site
The study site was located on terraces lying to the south of Qarlikturvik Valley in western Bylot Island (Fig. 3), within the Simirlik National Park (73°08′N; 80°00′W), part of the Canadian Arctic Archipelago between latitudes 72° and 74°N and longitudes 75° and 82°W. The general character of Qarlikturvik Valley and details of the study site are provided by Allard (1996) and Fortier and Allard (2004).
The valley terraces comprising the study site have developed during the post-glacial accretion of eolian sands and silts and the concurrent deposition of peat, initiated at ca. 3500–4000 yr BP; for process descriptions see Allard (1996) and Fortier and Allard (2004). Syngenetic growth of ice wedges in the aggrading sediments has resulted in the extensive development of polygon-patterned ground, and the terraces include a range of associated landforms (Fig. 1): low-center and high-center polygons, polygon-ponds, and thaw lakes (cf. Billings and Peterson, 1980). The polygon complex is fed by meltwater from the hills to the south, and the vegetation is dominated by sedges (e.g. Carex aquatilis var. stans, Eriophorum scheuchzeri), grasses (e.g. Arctagrostis latifolia, Dupontia fischeri, Pleuropogon sabinei), and fen mosses (e.g. Drepanocladus spp., Aulocomnium spp.).
Methods
Permafrost Cores
A single sediment profile 147 cm long in depth was extracted from the center of a frozen polygon pond in June 1999 (BY-Pond) (Fig. 3c). First, sediments from the active layer (to ca. 30 cm depth) were sampled as 2 cm slices cut with a knife from an unfrozen peat monolith (ca. 2 cm deep × 10 cm × 10 cm). Second, permafrost sediment was collected as contiguous 5 cm diameter, 50 cm segments, using a machine-driven corer (built by M. Allard, Université Laval). The corer was drilled vertically into the sediment until the chamber was full, recovered and the sediments removed as a single frozen column. This process was repeated until the entire depth of organic-rich sediment had been sampled down to an underlying layer of marine clay. Cored sections were examined in the field to ensure that sediments were horizontally bedded and undisturbed (cf. Fortier and Allard, 2004). The frozen 50 cm sections were cut where possible into 2 cm slices which were cleaned around their exposed outer faces (to prevent contamination), and sealed in labeled polyethylene bags. Samples were stored in a dark cooler box packed with ice for the remainder of the field season (eight weeks). On returning to the laboratory, samples were stored in the dark at 4 °C.
Three short monoliths were cut from the exposed flanks of eroded high-center polygons. Loose material was cleaned away from the eroding polygon flanks to expose a vertical face. Contiguous slices of sediment, ca. 1 cm deep × 10 cm × 10 cm, were sampled from the cleaned face, downwards from the polygon surface. Stratigraphic patterns around the back and sides of the each monolith were inspected to ensure that the sediments were approximately linear and undisturbed by cryogenic turbation (cf. Fortier and Allard, 2004). Slices were initially cut in early July 2000, from one polygon at the center of three different high-center polygon zones (Fig. 3c). Monoliths were revisited weekly, and further slices were collected during the summer as the active layer deepened. Final monoliths were labeled BY-hc1, -hc2, and -hc3, and varied between 22 cm and 28 cm in depth.
Analog Diatom Assemblages
To quantify wetland diatom assemblages, five samples were collected in late July and early August 2000 from five contrasting landforms within the present-day polygon complex: i.e. moss mats from high-center polygons, low-center polygon flanks and centers, and the upper sediment layers from polygon ponds and thaw lakes (cf. Figs. 1 and 2). Lentic sediments were collected using a surface sediment sampler (based on a design by Benoni and Enell; Fig. 8.3 in Aaby and Digerfeldt, 1986). Samples were sealed in labeled polyethylene bags and stored in a dark cooler box packed with ice for the remainder of the field season (four weeks). On returning to the laboratory samples were stored in the dark at 4 °C.
Laboratory Analysis—Diatoms
Diatom analysis was targeted to the analog samples collected from present-day wetland habitats and the horizontally bedded stratigraphic sections from core BY-Pond (0–92 cm) and monoliths BY-hc1, -hc2, and -hc3.
Siliceous microfossils (diatoms and chrysophyte cysts) were extracted from ca. 0.5 cm3 of sediment and placed into a 250 mL beaker with an initial 30 mL of hydrogen peroxide (30% H2O2). The sediment was macerated and gently heated. Hydrogen peroxide was added in small amounts until all organic material had oxidized and the remaining mineral residue topped-up to 80 mL with distilled water. This mixture was stirred gently and transferred by pipette (excluding therefore the larger mineral fraction) into a second beaker, which was covered and left to settle for 48 h. Supernatant liquid was gently removed with a pipette and the deposited sediment washed into a 20 mL glass vial with distilled water. Vials were sealed and stored at 4 °C until analysis.
The mixture of sediment-distilled water was gently shaken, and a drop of liquid was evaporated onto cover-slips and mounted onto slides using Naphrax© mounting medium. Diatoms were identified and counted as transects across the slide, at ×1000 magnification using phase contrast microscopy on an Olympus BX-40. The taxonomy of arctic diatoms is poorly defined, and identification is according to Krammer and Lange-Bertalot (1986, 1988, 1991a, 1991b) and Fallu et al. (2000). A minimum of 500 diatoms was counted for the present-day samples, though it was possible to count a minimum of only 150 diatoms for samples preserved in the stratigraphic record. Values for diatom taxa and chrysophyte cysts are calculated as the percentage of the total diatom count (% TDC).
Radiocarbon Dating
Material was radiocarbon dated from core BY-Pond and BY-hc1 to -hc3, at individual horizons delimiting a significant change in the stratigraphic record. Samples were pretreated before being submitted for radiocarbon analysis, to prevent contamination by carbon-rich Tertiary deposits (Gajewski et al., 1995), which occur on Bylot Island adjacent to and within the catchment area of the sampling site (Klassen, 1993). For BY-Pond a sediment sample of ca. 2 cm3 was washed in distilled water and sieved through a 0.3 mm sieve. In the absence of woody remains, bryophyte stems and leaves were collected from the organic residue and rinsed in distilled water. They were dried at 60 °C and samples of ≥0.1 g submitted to Beta Analytic Ltd (Miami, Florida, U.S.A.) for radiocarbon analysis by accelerator mass spectrometry (AMS). For each of the high-center monoliths (BY-hc1, -hc2, and -hc3), a bulk sediment sample was submitted to the radiocarbon laboratory at Université Laval for dating by liquid scintillation.
Results of the radiocarbon analysis are presented in Table 1. Conventional radiocarbon ages (yr BP) were calibrated using INTCAL 98 (Stuiver et al., 1998) and are presented in the text as calibrated years before present (cal. yr BP).
Table 1
Radiocarbon dates for permafrost core BY-pond and monoliths BY-hc1, -hc2, and -hc3, Qarlikturvik Valley, Bylot Island (73°08′N; 80°00′W). Calibrated ages were calculated using the computer program INTCAL 98 (Stuiver et al., 1998).
Statistical Analysis
The proportion of diatom taxa (% TDC) was arcsine–square root transformed prior to analysis (McCune and Grace, 2002). Variation in diatom communities from landforms in the present-day polygon complex was summarized first using detrended correspondence analysis (DCA: implemented in PcOrd; McCune and Mefford, 1999) to examine compositional change, and second using the Shannon index (H′) as a measure of assemblage diversity (Magurran, 2003). Contrasts between diatom communities (DCA axis one scores, and values of H′) were compared using one-way analysis of variance, with least square difference to identify significant contrasts (implemented using Genstat v. 7.1, 2003: VSN International Ltd., Oxford, U.K.). Indicator species analysis (Dufrêne and Legendre, 1997) was used to identify diatom taxa associated with contrasting wetland landforms (implemented in PcOrd; McCune and Mefford, 1999). Statistical significance of indicator species was examined using a Monte Carlo permutation test with 10,000 iterations used to estimate significance (p < 0.05).
We aimed to compare the diatom composition of paleoecological strata to the composition of analog samples, though basing this comparison on the passive treatment of diatoms preserved in sub-fossil assemblages and allowing for potential losses caused by decomposition. Diatom composition of paleoecological strata was compared to that of analog communities by applying weighted averaging (WA) ordination to sub-fossil assemblages (Curtis and McIntosh, 1950, 1951), where taxon weight (Wi) is the respective score along the primary orthogonal axis derived using DCA, and Aij is the abundance of diatom species i in stratigraphic horizon j:
The WA ordination of sub-fossil diatom assemblages can be used therefore to tentatively reconstruct landform and hydrologic setting, incorporating the comparison of WA scores against DCA sample scores derived for diatom communities in present-day habitats. The level of statistical significance was set at p < 0.05.
Results
Sixty-one diatom taxa were identified from moss mats and surface sediments collected from the present-day polygon complex (Appendix 1). Indicator species analysis identified 43 diatom taxa which were significantly associated with a given landform (Figs. 4a and 4b). Accordingly, analysis by DCA effectively summarized variation between diatom communities (Fig. 5). Orthogonal axis one (Δ = 0.545) explained 27% of variation in the dataset (R2 = 0.27), and there were significant differences in axis one scores between diatom assemblages sampled from contrasting landforms (F = 37.87, df1 = 4, df2 = 4, p < 0.001), with DCA axis one scores significantly different for four out of five landforms. There was no significant difference in axis one scores between low-center polygon centers and flanks. The diversity index (H′) identified only slight and non-significant differences in diatom community diversity between landforms representing low-center polygons, polygon ponds, and thaw lakes, with large over-lap in values of H′ (Fig. 5). However, there was a large and significant difference between the relatively more diverse communities associated with these landforms and less diverse communities sampled from high-center polygons (F = 52.92, df1 = 4, df2 = 4, p < 0.001).
Paleoenvironmental Reconstruction
It was possible to consistently count a minimum of only 150 diatoms from sub-fossil samples (compared to 500 diatoms for analog samples), pointing to an effect of decomposition on paleoecological information. For this reason, the use of WA scores to passively compare preserved diatom assemblages with modern analogs is appropriate, enabling a comparison based on sub-fossil diatoms remaining intact within the permafrost sediment, and allowing for decay-related change in the paleoecological record. Significant differences in DCA scores between diatom communities sampled from contrasting landforms (Fig. 5) support our application of species scores along DCA axis one, used as weights to calculate a WA score (Sj) for sub-fossil diatom assemblages. We believe WA scores provide for each stratigraphic horizon a reasonable estimate of past environmental conditions, enabling the summary reconstruction of contrasting landforms and hydrologic setting during polygon development. Applied to horizons sampled from BY-Pond (Fig. 6), the diatom paleorecord provides evidence of changing hydrologic conditions, with at least two periods of relative dryness (sub-fossil diatom assemblages comparable to present-day communities from low-center polygon centers) intercalated between periods of wetness (i.e. with sub-fossil diatom assemblages analogous to communities sampled from fully aquatic polygon ponds). This application of WA scores is supported by shifts in the stratigraphic record of diatom indicator species (cf. Fig. 4) during development of the polygon pond (Fig. 6), with peaks in the occurrence of diatom taxa indicative of drier conditions (e.g. Pinnularia brevicostata, P. sylvatica, and P. viridis) contrasted against % TDC of taxa indicating fully aquatic conditions (e.g. Cymbella ventricosa var. groenlandicum, Navicula vulpina, and N. pupula syn. Sellaphora). The paleoecological record is disrupted between ca. 60 cm and 40 cm depth by a layer of segregation ice, though the radiocarbon date at 61–63 cm depth (ca. 1930 cal. yr BP; Table 1, Fig. 6) suggests reconstructed variation in hydrology and landform have occurred over millennia.
Transitions to high-center polygon conditions in BY-hc1, -hc2, and -hc3 can be tentatively identified as a shift in the diversity of sub-fossil diatom assemblages, from values of H′ between 2.5 and 3.5 (comparable to communities from non-eroded polygons in the present-day wetland; Fig. 5) to less than 1.75 (Fig. 7). This shift in diversity is not corrected for time differences between samples (i.e. rates of sediment growth), or contrasting rates of decomposition between polygon types, and should be treated with appropriate circumspection. However, the shift in diversity is contemporaneous with a transition in WA scores: from evidence of polygon-pond conditions (preceding the shift to a high-center polygon) to a steep trend indicating drying of the polygon-surface and supported by an increase in % TDC of the indicator species Pinnularia borealis (cf. Figs. 4a and 7). The sampling failed to capture the full transition in diatom composition to high-center polygon conditions because too few diatoms were preserved in the uppermost strata. Nevertheless, radiocarbon dates identify the timing of a shift between polygon ponds and high-center polygons (Table 1, Fig. 7).
Discussion
Detailed geomorphologic study of the Qarlikturvik Valley polygon complex on Bylot Island has confirmed the local presence of horizontally bedded strata in the center of polygons (Fortier and Allard, 2004). These conditions provide an environment suitable to paleoecological reconstruction and the opportunity to study the long-term development of polygon-patterned tundra. Accordingly, previous studies in this polygon complex have sought to identify the relative importance of regional climate change and local geomorphologic and vegetation processes during recurrent development of low-center polygons (Ellis and Rochefort, 2004, 2006) and have attempted paleoenvironmental reconstruction of moisture and historic wind patterns (Fortier et al., 2006). This study extends previous investigations by examining paleoecological evidence for transitions between polygon landforms, and asks whether these transitions are consistent with existing successional scenarios.
The thaw-lake cycle hypothesis (Britton, 1957; Billings and Peterson, 1980) suggests that complete cyclical transition (cf. Fig. 2: low-center polygon → polygon pond → thaw lake → high-center polygon → low-center polygon) will occur over a period of ca. 3000 yr (Billings and Peterson, 1980). The developmental period encompassed by BY-Pond (>2000 yr; Fig. 6) should be sufficient to determine a sequence of landform shifts on a time scale that is comparable to the inferred process of cyclical succession. In contrast to the thaw-lake cycle hypothesis, our evidence suggests that the polygon pond examined has not (over an equivalent time scale) undergone the inevitable coalescence and increase in permanency leading predictably towards thaw-lake conditions. The pond examined appears to have persisted over millennia as a stable unit, though incorporating periods during which it has undergone a process of increased drying and regression towards low-center polygon conditions (Fig. 6). While there is abundant physical evidence for the active coalescence of ponds into thaw lakes across the wider surface of the polygon complex (Figs. 1 and 3c), this process may occur with less predictability than inferred by a strict interpretation of the thaw-lake cycle. The thaw-lake cycle may fail to account for the potential longevity of polygon ponds at our study site, and their fluctuation between wetter and transitional drier conditions following pond formation. In addition, diatom evidence from the high-center monoliths (BY-hc1, -hc2, and -hc3) suggests that in all three instances polygon-pond conditions preceded the rapid transition to high-center polygons (Fig. 7). This contrasts with both the thaw-lake cycle hypothesis (Fig. 2), and the proposed sequence of gradual terrestrialization (Zoltai and Tarnocai, 1975). Instead, the paleoecological reconstruction presented here indicates that high-center polygons may be formed by the degradation of ice wedges surrounding polygon ponds, and the subsequent collapse of ridges to form troughs. Following this transition, high-center polygons may persist for many centuries (or even millennia) without the subsequent rejuvenation of actively aggrading low-center polygons (Fig. 7).
We believe this study presents the first paleoenvironmental evidence to directly address specific landform transitions in a High Arctic polygon-patterned wetland. The results of the study weaken the widespread applicability of generalized scenarios describing the predictable succession of landforms in polygon-patterned tundra. Our results point instead to existing uncertainties which we believe highlight local effects underlying a dynamic and less predictable process of landform transition in patterned High Arctic wetlands. A more complete understanding of landform transitions in polygon wetlands is imperative in the context of global change. A great deal of research has focused on gradual climate-driven change in arctic vegetation, and consequences for ecosystem function (e.g. Arft et al., 1999; Bret-Harte et al., 2002; Epstein et al., 2000, 2004), and previous studies have identified the effect of relatively small changes in moisture content on the tundra carbon flux (Oechel et al., 1993, 1995; Weller et al., 1995). However, there is already evidence for major change in the spatial extent of contrasting tundra landforms (Stow et al., 2004) on a scale which heralds substantial implications for the tundra moisture regime and carbon cycling. These impacts include the potential for the large-scale degradation of permafrost (Beilman and Robinson, 2003; Lawrence and Slater, 2005), potentially resulting in widespread drainage and ice-wedge collapse (Jorgenson et al., 2006). A more robust framework describing landform transitions in polygon-patterned wetlands may now be needed to provide long-term context for existing and future landform change, and providing a background against which it may be possible to extrapolate climate impacts from short-term to long-term effects. We tentatively suggest that existing successional hypotheses may be inadequate to meet this challenge.
Acknowledgments
This research was funded by NSERC, FCAR, Centre d'études nordiques and Faculté des sciences des l'agriculture et de l'alimentation, Université Laval, as a university post-doctoral scholarship awarded to Christopher Ellis. Fieldwork was carried out from the Bylot Island Goose Camp, funded by a Canadian Wildlife Service Grant to Gilles Gauthier. We gratefully acknowledge logistic support and additional funding provided by Polar Continental Shelf Project. The Hunter and Trapper Association of Pond Inlet (Baffin Island) enabled access to Bylot Island.