1The Hudson Bay Lowlands have been rising isostatically for the past 7000 years, creating a regional chronosequence as new land emerges from the sea. Rates of uplift are most rapid in the eastern portion of the lowlands near the lower Albany River study area.
2The stratigraphy of three raised bogs was investigated to determine rates and pathways of peatland development in the Albany River region. The bogs are distributed evenly along the regional chronosequence from the oldest site at Oldman (5980 ± 100 bp) to progressively younger sites at Albany River (4810 ± 70) and Belec Lake (3960 ± 60).
3Each bog had the same stratigraphic sequence, beginning with a basal tidal marsh assemblage that was rapidly replaced by a Larix-dominated swamp forest, followed by a Picea-forested bog, and ultimately a non-forested bog. The bog–fen boundary is marked by the disappearance of fen indicators, dominance of bog-forming Sphagna, and a sharp decline in nitrogen. Each of these successional stages was associated with different rates of vertical growth.
4The rate of successional change was more rapid at the younger sites, and their vertical growth curve was more curvilinear. The formation of a raised bog, for example, was 1.3 times more rapid at Albany River and 5.5 times more rapid at Belec Lake than at Oldman. Belec Lake reached its ultimate successional stage first, although it was the last site to emerge from the sea.
5The differential rate of isostatic uplift across this region rather than climate was the principal environmental driver for peatland development. The faster rate of uplift on the lower reaches of the drainage basin continues to reduce the regional slope, impede drainage and shift river channels, continually altering the local hydrogeological setting.
6Groundwater flow simulations based on the Dupuit equation show that the growth of these raised bogs was probably constrained by their local hydrogeological setting. Bog formation was first induced by the creation of interfluvial divides between headwardly eroding streams or shifting river channels, and further bog growth was ultimately constrained by the width of the interfluve and the depth of river incision. The Belec Lake bog was the first to approach its limiting height because its narrow interfluve could only support a low water-table mound.
7Although peatland succession largely followed the same conservative pathway at each site, both the pace and direction of these pathways were set by geological processes, which are probably the decisive drivers for the evolution of this large peat basin.
In large peat basins the deepest organic deposits generally occur under the convex landforms of raised bogs. The morphology of these bog landforms changes across geographical gradients, suggesting an underlying environmental control on peat growth (Ruuhijärvi 1960; Eurola 1962; Glaser & Janssens 1986). These relationships are expressed quantitatively by two different groups of theoretical models. One set assumes that the vertical growth of a bog is limited by climatic drivers that control the geometry of its internal water-table mound (Wickman 1951; Ingram 1982, 1983; Winston 1994; Hilbert et al. 2000). The other set assumes that bogs cease to grow when a steady state is reached between primary production and decomposition (Clymo 1984, 1991). In addition, the morphology of a bog landform may be a function of its developmental stage. Glaser & Janssens (1986), for example, suggest that bogs with a forested crest will change to non-forested domes or plateaux with pool networks as their upward growth reaches a limiting height.
A stratigraphic study was conducted in the Albany River drainage of the Hudson Bay Lowlands to test the predictions of the Glaser and Janssens model and also to investigate the dynamics of bog development across an isostatically rising landscape. The Albany River region contains a wide range of bog landforms that vary in size, morphology and surface patterns. Moreover, these bogs have developed across a chronosequence created by the continuing emergence of the lowlands from the sea during the past 7000 years. The rapid rate of isostatic uplift has altered the topography of the regional water table and the geometry of the drainage network. The effect of these rapid geological changes on bog development should be readily apparent in this region, given the exceptionally gentle relief of the lowlands and the nearly continuous cover of peat.
The Albany River study area is located in the south-eastern portion of the Hudson Bay Lowlands (Fig. 1a). This exceptionally gently sloping bedrock plain is composed of Palaeozoic limestone and dolomite mantled by calcareous glacial and marine sediments of variable thickness (Bostock 1970). The plain is dissected by the Albany River, which is incised within a narrow gorge 30 m deep and by a system of largely unbranched first-order streams. The drainage network is largely dendritic in the upper reaches of the study area and then shifts to parallel systems of rivers down-gradient. The glacial landforms include a fluted till plain and moraine and also drumlinoid landforms indicating former directions of ice advance (Fig. 1b). The climate is temperate-boreal, with a mean annual average temperature of −1.2 °C, mean annual maximum temperature of 5 °C, mean annual minimum temperature of −7 °C, and mean annual precipitation of 750 mm (Sims et al. 1979). No evidence for permafrost exists within the study area. Approximately 90% of this landscape is presently covered by peatlands, of which 55% are fens and 35% bogs, with the remainder mineral soil or standing water (Glaser et al. 2004).
Materials and methods
The entire field area was surveyed with aerial photographs and Landsat TM imagery prior to fieldwork. A base camp was established in 1985 and 1992 along the lower Albany River, and from there a helicopter provided comprehensive access to sites within a 160-km radius. Field sites were selected on the basis of their landform patterns, hydrogeological setting and distance from the sea. It was assumed that the age of the land surface increased with increasing distance from the coast. Peat cores were collected from three raised bogs with a piston sampler equipped with a stainless steel barrel 10 cm in diameter with a serrated cutting edge (Wright et al. 1984). In the laboratory the core sections were described according to the methods of Troels-Smith (1955) and the major types of macrofossils were identified from the important stratigraphic zones. Peat samples were dated by 14C analysis by Beta-Analytic Inc, Miami, Florida, USA. Samples were also collected at approximately 10-cm intervals for analysis of carbon, nitrogen, hydrogen and sulphur with a Carlo Erba EA 1108 elemental analyser.
Pollen samples were taken at 20-cm or closer intervals from each core and prepared by standard procedures (Faegri & Iverson 1989). The results were plotted in percentage diagrams, with Sphagnum and Equisetum percentages based on a pollen sum of 300 plus Sphagnum or Equisetum spores, respectively. Conifer stomata were identified according to Hansen (1995) and Trautmann (1953), and concentrations were expressed as stomata gram−1 wet sediment.
A sensitivity analysis was conducted to determine the effect of the local hydrogeological setting on water-table elevations and bog development. This analysis is based on an analytical solution to the Dupuit equation that calculates the topography of the water table irrespective of the actual land surface and thus estimates the ultimate limiting height for peat accumulation. An analytical solution to the Dupuit equation from Fetter (2000) was used to simulate the topography of the water table between two parallel streams:
where H = elevation of the water table at point x, H1 = elevation of the water table in stream 1, H2 = elevation of the water table in stream 2, L = the distance from stream 1 to stream 2, w = recharge, and K = hydraulic conductivity of the aquifer.
This equation was solved for 20 evenly spaced intervals between the two streams by inserting values for L, K and w that bracket the average values obtained for the study area from either aerial photographs, field data or the literature. The effect of stream incision, for example, was simulated by progressively lowering the water-table elevation in H1 and H2 using the depth of the Albany River gorge (30 m) as the lower limit for incision. In addition, the model was also run by varying the average values reported for recharge rate (0.35 m year−1) by Hare (1997) and for hydraulic conductivity by Reeve (1996).
Three raised bogs were selected for stratigraphic analysis: (i) Oldman bog (51°01′ N, 84°34′ W) near Albany Forks, approximately 250 km from the coast of James Bay; (ii) Albany River bog (51°26′ N, 83°37′ W) near the confluence of the Chemahagan and Albany Rivers, approximately 170 km from the coast; and (iii) Belec Lake bog (51°37′ N, 82°17′ W), approximately 80 km from the coast (Fig. 1). All three sites are on interfluvial divides between streams. The vegetation and surface-water chemistry of these sites is described in detail by Glaser et al. (2004).
The 73.9 km2 Oldman bog occupies an interfluve 9 km wide between the Kenogami and Chemagan rivers (Fig. 2). Its landforms are transitional between the southern forested and the northern non-forested bogs of continental North America (sensuGlaser & Janssens 1986; Zoltai et al. 1988; Glaser 1992). The central axis of this large bog complex has a forested crest to the east marked by lines of Picea mariana that radiate downslope onto non-forested bog lawns. To the west, however, the forested crest grades into a non-forested bog plain with systems of pools, mud bottoms and hummocks. Fen water tracks arise from the lower edges of this bog plain and drain downslope to tributary streams. The coring site (8502) is located near the summit of the non-forested bog plain.
The 28.0 km2 Albany River bog covers a plateau 7.6 km wide bordered by the Albany River gorge to the south and U-shaped valleys to the north and east (Fig. 3). The bog has a narrow forested crest along its southern margin. This crest forms the upper rim to a broad non-forested bog plain that slopes gently to the east and north. The plain contains systems of bog pools, mud bottoms and hummocks, but it is also finely dissected by fen water tracks with orientated pool networks. The coring site (8501) is located on the upper reaches of this non-forested bog plain close to the southern bog margin.
The 12 km2 Belec Lake bog occupies a narrow interfluve 2.6 km wide between the Stooping River and a reticulate fen with networks of pools and peat ridges (Fig. 4). This reticulate fen is a local discharge zone for groundwater (Reeve 1996; Glaser et al. 2004) and therefore functions like a stream or drainage ditch. In contrast, the bog is nearly completely covered by a concentric network of large bog pools with narrow peat ridges and a few isodiametric bog hummocks. Its morphology seems to conform to the end-state of raised bog development, in which expanding pools cover the entire bog plain (Glaser & Janssens 1986). The coring site (8507) is located on a non-forested bog hummock near the summit of the bog.
The three study sites correspond to a regional gradient of increasing age and peat depth with increasing distance from the coast. Oldman is the oldest (5920 ± 90 bp) and deepest (446 cm) site, Albany River is intermediate in age (4810 ± 70 bp) and peat depth (266 cm), and Belec Lake has the youngest (3960 ± 60 bp) and shallowest (236 cm) basal peat (Fig. 5, Table 1). Despite these differences in age, depth and location, the cores from each site contain the same stratigraphic units. In ascending order these units are: (i) a basal clayey silt; (ii) non-woody herbaceous peat; (iii) woody peat with Larix needles and fen indicator species; (iv) Sphagnum-woody peat with no fen indicators; and (v) non-woody Sphagnum peat with no fen indicators (Fig. 5, Table 2).
Table 1. Radiocarbon dates for Oldman, Albany River and Belec Lake bogs
Sample depth (cm)
14C age (bp)
1110 ± 70
2020 ± 70
2450 ± 90
2680 ± 80
3750 ± 70
4810 ± 70
280 ± 90
870 ± 80
990 ± 70
1010 ± 70
1520 ± 70
2010 ± 70
2940 ± 70
2610 ± 70
3060 ± 90
3690 ± 70
4910 ± 100
5310 ± 80
5980 ± 100
5920 ± 90
900 ± 70
1350 ± 60
2280 ± 50
2770 ± 70
3070 ± 70
3430 ± 110
3960 ± 60
Table 2. Peat stratigraphy for the 8501, 8502 and 8507 cores. The symbols for sediment description follow Troels-Smith (1955)
(a) Albany River (8501)
4Tb° + Th° + Tl°
4Db2 + Dh2 + Dl2 (with more humified bands 3Db21Dg3)
The basal clayey silt contains fragments of Equisetum rootlets and other organic debris that decrease in abundance downwards from the peat/mineral contact. Weathering horizons are not apparent in the mineral substratum and the individual grains have unweathered surfaces under the microscope. The peat/mineral contact is sharp at each site. The thin layer of non-woody herbaceous peat just above the peat/mineral contact consists of a finely decomposed matrix with abundant remains of Equisetum and larger herbaceous fragments. This basal peat layer is overlain by woody peat containing Larix needles and various fen indicator species. The bryophyte remains are consistently feather mosses (Amblystegiaceae), except at the top of this zone, where leaves of Sphagnum subsecundum appear in strata with needles of Larix. This transitional zone grades upwards into woody-Sphagnum peat as first the Sphagnum subsecundum and then the Larix is replaced by remains of Picea and species of Sphagnum typical of a raised bog. No fen-indicator species occur in this zone. The wood layers and Picea needles are absent from the upper part of the peat profile, which consists largely of Sphagnum peat.
The bog–fen boundary is conservatively defined by the absence of fen-indicator taxa and the occurrence of both Picea mariana and Sphagnum. This boundary occurred at 197 cm depth at Belec Lake, 182 cm at Albany River, and 255 cm at Oldman bog. The remainder of the peat column is dominated by Sphagnum peat, although wood is missing from the profile above 83 cm at Belec Lake, 43 cm at Albany River, and 108 cm at Oldman.
All three study sites have the same general pollen stratigraphy (Figs 6–8), which is initially dominated by regional arboreal types, primarily Pinus diploxylon type (20–40%). The basal assemblage zone is distinguished by relatively high percentages of Equisetum (40–60%), Chenopodiaceae/Amaranthaceae (up to 10%), Cyperaceae (10–40%), and Poaceae (c. 5%). Other tidal marsh or fen taxa also characterize pollen zone 1, including Typha spp., Sparganium t., and Triglochin/Potamogeton t. Towards the top of the assemblage zone the percentages of Larix and Picea pollen and the total pollen concentration increase, whereas conifer stomata are generally absent.
The basal assemblage zone is overlain in zone 2 by a pollen assemblage distinguished by increasing percentages of Larix and Picea mariana and by significant concentrations of Larix stomata (2000–4000 stomata g−1 of sediment). Picea mariana pollen percentages reach their maximum values in zone 2 at all sites, but Picea stomata peak at the bottom of zone 2 at Albany River (8501) and Belec Lake (8507) bogs, whereas at Oldman (8502) stomata concentrations peak at the top of the zone. In contrast, marsh-fen taxa (e.g. Chenopodiaceae/Amaranthaceae, Equisetum, Poaceae and Cyperaceae) decline in representation in zone 2 concomitant with increasing percentages of Betula, Ericaceae and Sphagnum. Pinguicula pollen is prominent in zone 2 at Oldman, but not at the other sites. Regional pollen declines from zone 1 to 2.
Zone 3 is distinguished by the absence of Larix stomata and by declining Picea stomate concentrations. Picea mariana pollen percentages remain high (20–40%), and Larix pollen percentages are unchanged from the previous assemblage zone at all three sites. Further differentiating pollen zone 3 are the relatively high percentages of Betula, Ericaceae and Sphagnum (up to 80%), while Cyperaceae pollen percentages decline to less than 2%. The regional pollen input remains stable and pollen concentrations decline mid-zone.
Stomata of Picea and Larix essentially disappear in zone 4, which is still dominated by high percentages of Picea mariana pollen (40–60%). Sphagnum spore percentages decline in zone 4 from their highs in zone 3, whereas Betula (10–15%), Ericaceae (c. 10–50%) and Cyperaceae (up to 20%) gain in importance. Pollen from taxa associated with land disturbance (e.g. Ambrosia, Chenopodiaceae/Amaranthaceae) is evident in the uppermost portion of zone 4.
carbon, nitrogen and sulphur
The carbon content of all three cores ranges between 45.7 and 55.1% of dry mass, with an average value of 50.4%. Hydrogen is linearly related to carbon and has an average value of 5.4% of dry mass. Sulphur remains below 0.2%. Nitrogen, however, falls from greater than 2% of dry mass in the fen zone to much lower values (0.5–1.5%) in the bog zone. The sharp change in nitrogen precisely corresponds to the bog–fen boundary identified by plant macrofossils. The carbon : nitrogen ratios show corresponding changes from close to 25 in the fen peat to much higher values between 40 and 115 in the bog peat (Fig. 9a,b,c).
The Dupuit simulations show the response of the water table in an interfluvial divide to: (i) distance between the streams (L); (ii) relative water elevations in the streams (H1 and H2); (iii) recharge rate (w); and (iv) hydraulic conductivity (K) of the aquifer (Fetter 2000). The model consistently calculated a water-table mound in an interfluvial divide regardless of the values used for L, H1 and H2. However, the elevation of the water-table mound decreased with lower values for L, H1, H2 and w, whereas a decline in K raised the height of the mound when the other factors were held constant (Fig. 10a,b,c). The inflection point for the calculated curves in these plots consistently corresponds to an interfluvial width (L) of about 5.2 km. Stream incision lowers the simulated elevation of H1 and H2, which in turn lowers the height of the water-table mound while increasing its convexity. Different rates of incision will shift the water divide towards the stream with the higher water level if one stream erodes faster into its bed than the other (Fig. 11).
The simulations show that the hydrogeological setting of the Oldman bog (L = 9 km) can sustain the highest water-table mound relative to those at the Albany River (L = 7.6 km) and Belec Lake (L = 2.6 km) bogs. The water table was only slightly mounded within the narrow interfluvial divide at Belec Lake, and only a small amount (> 2.5 m) of downcutting by its streams would depress the top of the mound below the original land surface. The model simulations also show that the top of the water-table mound is located equidistant from both streams when H1 and H2 are at the same elevation. However, if one stream is lowered faster than the other, the water divide shifts up gradient towards the higher stream.
The Albany River peatlands contain a 6000-year record of primary succession and geomorphic change on an isostatically rising landscape. The evolution of this landscape was largely shaped by its close proximity to the centre of the former Laurentide Ice Sheet, which disintegrated about 7800–8000 years ago (Dredge & Cowan 1989; Dyke et al. 1989). The load from this ice mass depressed the earth's surface, permitting the Tyrell Sea to flood across the Hudson Bay Lowlands immediately after deglaciation (Lee 1960). Once the loading from the ice mass was removed, the land surface began rising isostatically with the maximum rates of uplift centred around the mouth to James Bay (Hunter 1970; Webber et al. 1970; Andrews & Peltier 1989). Isostatic rebound has continued to the present as a result of both a flexing of the lithosphere and the movement of crustal materials.
The postglacial evolution of the Albany River landscape has been driven by the pattern and rate of uplift. As new land emerged from the sea the coastline migrated towards the zone of maximum uplift in James Bay. Higher rates of uplift near the advancing coast decreased the grade of the rivers, promoting rising water levels and rapid spread of peatlands across the lowlands. The decreasing gradient also tended to back up flow in the Albany and Kenogomi rivers, which periodically spilled over and cut new drainage channels. Abandoned channels of these rivers are distinguished by U-shaped valleys with underfit streams and paludified beds (Fig. 1b). The lower reaches of the Albany drainage are also distinguished by dense networks of parallel streams that follow systems of mega-flutes (sensuClark 1993) aligned to the former direction of ice advance.
The development of peatlands on this rapidly changing landscape was driven by two different sets of factors. Biotic factors produced a nearly continuous blanket of peat, which impeded local drainage and separated the vegetation from the calcareous glacial/marine sediments. The most conspicuous evidence for these processes is the abundance of mire pools on the modern landscape and the relatively dilute acidified surface waters in the peatlands (Glaser et al. 2004). However, isostatic uplift raised the piezometric surface by 60–120 m within the study area, whereas the headward erosion and incision of streams altered the elevation of the water table in the interfluvial divides. The degree to which these external drivers were buffered by biotic processes is indicated by both the manner and the rates of change in the stratigraphic successions preserved in the peat profiles.
land emergence and regional peat formation
The clayey silt that underlies the three study sites was probably deposited along an estuarine-dominated coast similar to the modern coastline of south-eastern James Bay (Martini et al. 1980a,b; Martini 1981). The abundant remains of Equiseum in this clayey silt indicate that the salinity of the waters was low, probably as a result of discharge from the major rivers. This interpretation is supported by the occurrence of other freshwater taxa, such as Menyanthes, Typha, Rhynchospora, Hippuris and Dryopteris. The higher amounts of grass and chenopod pollen in this silty clay may be a response to storm surges carrying saline waters inland to areas where poor drainage and evaporation created zones of higher salinity (Martini et al. 1980a,b).
The sharp peat/mineral contact and absence of weathering profiles within the mineral substratum indicate that peat immediately began to accumulate at each site after emergence from the sea. The basal peat assemblages indicate a tidal marsh or brackish meadow dominated by sedges, grasses and chenopods. The most important indicator types, such as Triglochin/Potamogeton, Typha, Chenopodiaceae and Equisetum are characteristic of the brackish marshes that now dominate the coast between the Attawapiskat, Albany and Moose rivers (Glooschenko 1980; Martini et al. 1980a,b; Riley & McKay 1980). This basal herbaceous layer is thin at all three sites, indicating that a rapid rate of uplift separated the recently emerged land from tidal waters and favoured the invasion of a fen-swamp forest.
stages of peatland development
The tidal marsh assemblage was quickly replaced by a swamp-forest assemblage dominated by Larix laricina and various fen-indicator taxa, such as Equisetum, Betula pumila var. glandulifera, Calliergon giganteum, Meesia triquetra, Sphagnum subsecundum and Phragmites. The presence of forest is confirmed by large pieces of wood, abundant Larix needles, Larix stomata, and other pollen or macrofossil assemblages characteristic of a modern swamp forest in this region (cf. Glaser et al. 2004). This Larix assemblage zone was most diverse at the Oldman site, where three sub-zones could be distinguished and the section is over 1.5 m thick.
The Larix swamp forest was subsequently invaded by the minerotrophic species Sphagnum subsecundum just below the bog–fen boundary. Species characteristic of a modern raised-bog forest then replaced Larix and all fen-indicator species disappeared. Sphagnum recurvum usually appeared first, followed by needles of Picea mariana. The remainder of the forested bog zone is dominated by Sphagnum fuscum and S. magellanicum, with occasional seeds or leaves of Chamaedaphne calyculuata, Vaccinium and other Ericaceae. The presence of forest cover is supported by the abundant needles, stomata and wood of Picea mariana. The fossil assemblages in this zone correspond to the modern species assemblages described for bog forests in the region, which are always dominated by Picea mariana (Glaser et al. 2004). The bog–fen boundary is also distinguished by the sharp fall in the nitrogen content of the Sphagnum-bog peat, which coincides with the disappearance of fen-indicator taxa and dominance of bog-forming Sphagna. Similar floristic and/or nutrient changes at the bog–fen boundary were reported for other North American peatlands (Glaser et al. 1990; Janssens et al. 1992; Kuhry et al. 1993).
Towards the upper portion of the peat column the raised-bog forest assemblage was replaced by peat characteristic of a non-forested bog. This transition is marked by the disappearance of large wood layers, absence of Picea needles and stomata, higher concentrations of Cyperaceae pollen, and appearance of bryophytes characteristic of non-forested lawns, such as Warnstorfii fluitans (=Drepanocladus fluitans). This zone is also distinguished by interbedded layers of relatively undecomposed Sphagnum alternating with darker more humified layers probably derived from lichens that characterize the hummocks of non-forested bogs in the Maritime Provinces of Canada (Glaser & Janssens 1986).
The stratigraphy at each site indicates a common developmental pathway leading from a brackish tidal marsh to a non-forested bog. Peat accumulation began during the tidal marsh stage and therefore peatland genesis did not conform to the classical model of terrestrialization (lake-infilling) or paludification (swamping of dry land) as defined by Weber (1902, 1908) and Cajander (1913). Peat formation in the lowlands was driven by the faster rate of uplift near the coast, which impeded regional drainage, raised water levels, and maintained waterlogged soils from the time of land emergence. Peatland succession then followed a conservative pathway similar to that described for other bogs in the Hudson Bay Lowlands (Sjörs 1963; Klinger & Short 1996) and other regions (Glaser & Janssens 1986; Janssens et al. 1992).
rates of peatland development
The three study sites are evenly distributed across the regional chronosequence in terms of both distance (c. 80 km) and time of land emergence (1000 years). In all, the vertical growth rate was initially slow (0.04–0.07 cm year−1) during the Larix-swamp stage, increased significantly (0.08–0.1 cm year−1) during the Picea-bog forest, and then declined (0.04–0.05 cm year−1) during the final stage of non-forested bog. The similarity between the successional pathways and vertical growth rates of these bogs indicates a conservative response by the biotic assemblages to a changing environment. However, the markedly faster rate of succession from tidal marsh to non-forested bog at the younger sites provides strong evidence for external forcing by environmental factors (Tables 3 and 4).
Table 3. Duration of the various developmental stages
Table 4. Chronology of the major stratigraphic transitions
Basal peat/land emergence
5980 ± 100
4810 ± 70
3960 ± 60
3060 ± 70
2680 ± 80
3430 ± 110
Origin of non-forested bog
1540 ± 70
1110 ± 70
2280 ± 60
The tidal-marsh stage was brief at each site because the rising landscape rapidly receded from the tidal zone, and only a thin layer of peat was sufficient to isolate the vegetation from saline pore waters in the glacial marine sediments. However, formation of a raised bog was 1.3 times more rapid at Albany River and 5.5 times more rapid at Belec Lake than bog formation at Oldman. Moreover, the terminal stage of non-forested bog appeared first at Belec Lake even though this site emerged from the sea 2000 years after the Oldman site. Apparently the rates of isostatic uplift and geomorphic change were greater at the younger sites, intensifying the signal produced by external drivers.
models of bog development
In large peat basins, raised bogs tend to develop where rivers create local watershed divides (Kulczynski 1949; Heinselman 1970; Glaser et al. 1997). Kulczynski suggested that these water partings produce zones of stagnant flow that are disconnected from lateral flow systems transporting mineral solutes, a hypothesis later adapted by Heinselman (1970). Siegel (1983) and Glaser et al. (1997), however, proposed that the water-table mounds under these watershed divides drive local recharge (downwards) flow systems that separate the bog vegetation from geogenous waters. This view is supported by the hydrogeological setting of raised bogs in the study area (Glaser et al. 2004).
Once established, bogs in the Albany River region seem to develop according to the sequence proposed by Glaser & Janssens (1986) for a continental climate. According to this model changes in the rate of vertical growth produce corresponding transformations in peat landforms, hydrology and vegetation. The initial bog stage is marked by rapid vertical growth focused along a ridge-like crest. The relatively steeper slopes and porous peat along the crest improve drainage, promoting the growth of a Picea mariana forest. However, as a bog approaches its maximum limiting height the rate of vertical growth slows and the crest is transformed into a dome or gently dipping plateau. The change in slope and the accumulation of denser less porous peat inhibits drainage, driving a rise in the water table. The bog forest is then replaced by sedge assemblages and pool networks.
This conceptual model assumes that the rate of vertical growth slows as a bog approaches its maximum limiting height. Two different types of theoretical models predict the upper limit for bog growth. The most influential mass-balance models assume that a steady state is eventually reached between the new plant mass added from the peat surface and the cumulative loss of organic mass over the entire peat column (Clymo 1984, 1991). Cumulative carbon losses will therefore limit the maximum thickness of any peatland. Although peat depths were never greater than 6 m in the Albany River study area, a steady-state model fails to predict the more rapid rate of development at the Belec Lake bog with its thinner peat profile. It also cannot account for the variable peat depths and developmental rates in the two other Albany River study sites.
Alternatively, the vertical growth rate may be limited by the ability of the water table to migrate continually upwards and saturate new layers of peat. Saturation is an essential factor for peat formation (Clymo 1984, 1991), but peat mounds can only remain saturated under local hydrological conditions that counteract the tendency for water to drain downgradient to the regional base level (Freeze & Cherry 1979). Two different types of hydrological models have been employed to simulate the geometry of the water-table mound in raised bogs. Numerical models (e.g. Siegel 1983; Waddington & Roulet 1997; Reeve et al. 2000, 2001) are computationally intensive but have the flexibility required for modelling heterogenous hydrological settings. Analytical models (Ingram 1982, 1983; Clymo 1991; Winston 1994), on the other hand, are isotropic and therefore cannot be adjusted to account for local changes in model parameters (e.g. K, H1, H2). This type of model, however, is well suited for a sensitivity analysis to determine the relative effect of each model parameter.
Raised bogs are usually found on the interfluvial divides between parallel streams in the Albany River region (Glaser et al. 2004). An analytical solution for the Dupuit equation calculates the topography of the water table within these divides irrespective of the actual land surface. This theoretical surface therefore represents a maximum limit above which the water-table mound in a growing bog can no longer rise. Model simulations show that a water-table mound will always form under these interfluvial divides even before the streams erode deeply into their beds. This event will trigger the formation of a raised bog by: (i) favouring more rapid rates of vertical peat growth; (ii) producing vertical-flow reversals in which surface waters flow downwards into the deeper peat; and thus (iii) creating zones of dilute waters at the peat surface that favour the invasion of bog-forming Sphagna.
The model simulations also show that wider interfluves can support higher water-table mounds. Accordingly, vertical peat growth would proceed longer at Oldman bog (L = 9 km) than at Albany River bog (L = 7.6 km) or Belec Lake (L = 2.6 km). In addition, bogs in wider interfluves are less affected by river incision because the drawdown of the water table decreases with increasing distance from the downcutting rivers. The model simulations predict, for example, that only a small drop in H1 or H2 will lower the elevation of the Belec Lake water-table mound below the original land surface, whereas a much greater amount of stream incision is required to produce the same effect at Oldman. It also accounts for the restriction of the largest bogs in the Albany River region to the broader interfluvial divides (Glaser et al. 2004).
These model simulations are sensitive to changes in recharge (w) and the hydraulic conductivity (K) of the peat. No significant changes in recharge can be inferred from the 6000-year pollen stratigraphy, but K does change with depth in the Albany River bogs (Reeve 1996; Reeve et al. 2000, 2001). A sensitivity analysis indicates that the height and convexity of the water-table mound in an interfluve increases with decreasing K (or increasing w) for any given value of L. The model always calculates a rise in Hmax as L increases regardless of the values used for K, w, H1 and H2. The inflection point for these curves consistently occurs at an L-value of 5–5.2 km, indicating a threshold above which the land surface is prone to flood. This critical value corresponds to the lower size limit for bogs that are dissected by internal water tracks (sensuGlaser 1987, 1992; Glaser et al. 2004), which apparently drain surplus recharge.
Nevertheless, the Dupuit model simulations indicate that the local hydrogeological setting imposes a theoretical upper limit to peat growth. Peat cannot accumulate above the maximum simulated water table, which marks the ultimate capacity for the water table to rise into new layers of peat. Initially, the vertical growth of peat was limited by the absence of a local water divide at each of the study sites. In this setting the water table remained nearly horizontal, dipping gently towards the coast or the nearest discharge point. The slow rate of vertical growth during the Larix-swamp stage was probably sustained by the continued reduction of the regional gradient produced by greater isostatic uplift along the emerging coast.
The transformation of each site from a Larix-swamp to a Picea-bog forest occurred as headwardly eroding streams (or shifts in the channels for the Albany and Kenogomi rivers) formed local divides with water-table mounds. These sites could sustain a rapid rate of peat growth only as long as the maximum potential height of their water-table mounds exceeded that of the actual peat surface. Rapid vertical growth was promoted by the poor litter quality of the bog-forming Sphagna and woody tissue (Clymo & Hayward 1982; Bridgham et al. 1998), whereas the various woody remains also added structural support to the accumulating peat mass. However, as the streams incised their beds the simulated height of the water table declined until further peat growth was limited. The shift to slower rates of vertical growth was associated with the transition from Picea-bog forest to non-forested bog. Belec Lake was the first site to approach this limit, because its narrow interfluve could only support a low water-table mound. River incision should continue to lower these water table mounds, but this effect can be counteracted as higher decomposition rates simultaneously lower the values for K in the surface peat.
palaeo-shifts in the albany river channel
The Dupuit equation also predicts the distance of the water-table divide from the nearest streams (Fig. 11). The model simulations show that a divide forms in the middle of an interfluve if both streams have the same water level. The closest fit for these simulations is the centrally located crest or summit of the Oldman and Belec Lake bogs. In contrast, the model simulations also show that the divide shifts closer to the stream with the higher water level if one stream erodes into its bed at a faster rate than the other. This result does not agree with the present hydrological setting of the Albany River bog, where the bog crest is located near the downgradient margin overlooking the 30 m deep gorge of the Albany River. Surface drainage in contrast flows towards wide valleys to the north and east that are higher than the Albany River. The modern bog surface is therefore out of equilibrium with water levels in the adjacent streams and must represent an adjustment to a former drainage pattern.
The peat stratigraphy indicates that this change probably occurred about 1000 years ago when the rate of vertical peat growth slowed at the bog crest (Fig. 5). Prior to that time the Albany River probably flowed through the valley systems just to the north and east of the Albany River bog (Fig. 3). This valley system has approximately the same width as the modern Albany River gorge and presently contains underfit streams and a partially paludified valley floor. At approximately 1000 bp the Albany River probably spilled over and cut its present channel in response to the greater differential uplift of the lower reaches of the river and reduction in its grade. It is noteworthy that geomorphic evidence exists for other shifts in the Albany and Kenogomi River channels across the Albany River region (Fig. 1b).
Peatlands have an unusual capacity to modify their physical setting by raising the land surface and decreasing horizontal-hydraulic gradients. As a result biotic processes are often viewed as a primary determinant for peatland succession, which under suitable climatic conditions will produce a raised bog with a self-regulating flow system (Weber 1902; Von Post & Granlund 1926; Ingram 1983; Tallis 1983). In the Hudson Bay Lowlands bog development generally conforms to the simple pathway predicted by Glaser & Janssens (1986), indicating a conservative response of the biota to the regional environment. However, geological factors rather than climate determine the rates and directions of peatland succession in this peat basin. Peatland genesis in the lowlands is apparently driven by the differential pattern of uplift, which reduces the regional gradient and raises water levels. In addition, these peatlands are continually adjusting to changes in the rapidly evolving river systems, which control the water-table topography within interfluvial divides. Even in regions with more stable landscapes and more variable climates, the local hydrogeological setting must impose similar constraints on peatland development. The development of patterned peatlands may therefore be predictable by coupling groundwater to ecosystem models.
We thank John Almendinger, Howard Mooers, Jay Gilbertson and Michael Stone for assistance in the field. Robert Bissel and Daniel O'Donnell piloted the helicopters and G.M. Veilleux the float plane. Kerry Kean assisted with the modelling, Jan Janssens made the bryophyte determinations, and H.E. Wright Jr and two reviewers helped improve the manuscript. This work was supported by grants from the National Aeronautics and Space Administration and National Science Foundation.