FIELD SITES

The Ontario James Bay Lowland (JBL), in which most of the study sites are located (Fig. 1), covers 221,000 km²; 36% of the wetlands present are bogs dominated primarily by Sphagnum (Riley, 2005). The Hudson Plains contain Picea mariana-dominated forests in the south (Fig. 2) and mixed P. mariana and Larix larcina forests in the north (Riley, 2005, 2011). These are interspersed with unforested bog and fen complexes dominated by Sphagnum mosses and Carex sp., respectively (Fig. 2). Northern permafrost bogs have drier surfaces and are often typified by a surface cover of Cladonia. The Hudson Plains formed after the last glacial maximum following the retreat of the Laurentide ice sheet commencing at approximately 10 ka. The Hudson Plains remained largely covered with the remaining ice sheet and glacial lakes Agassiz and Ojibway until the first land emergence around 9 ka (Dyke et al., 2003). Many peatlands in the southern lowlands were formed after the catastrophic drainage of the glacial lakes due to the breaking of the Laurentide ice dam around 8.4 ka (Lajeunesse and St-Onge, 2008). The rest of the land gradually became available for peatlands due to isostatic rebound. The Hudson Plains have had some of the fastest rates of isostatic uplift on earth, with some areas rising an average of 1.2 m century^-1 (Webber et al., 1970).

Permafrost occurs in the Hudson Bay coast as well as northern regions of the Hudson Bay and James Bay Lowlands. It dominates the landscape in the north within ∼80 km of the coast, and occurs along the coast in the south within a slimming margin approximately 20 to 40 km wide (Fig. 1; Riley, 2005). Along the coast, continuous permafrost and subarctic vegetation can be found anywhere where soil temperatures remain below 0 °C annually, and discontinuously south of this isotherm due to the efficient insulating qualities of peat soils (Riley, 2005).

The peatlands of central and northern Ontario are accessible almost exclusively by helicopter, boat, or winter ice road, and remain both sparsely populated and understudied. Some peatland basal dates and paleoecological information are available from studies focused near the Albany River (Glaser et al., 2004), Churchill Rail Line (Dredge and Mott, 2003), and Victor Mine (O'Reilly, 2011; Bunbury et al., 2012). We collected eight bog cores from eight sites by helicopter during the summer of 2008 (Fig. 1; Table 1). The sites form a latitudinal transect between 50°27′N and 55°24′N (Fig. 1; Table 1). All cores were collected from raised, ombrotrophic peat landforms (Table 1). MAP and MAAT data from three climate stations within the approximate range of the sites are included in Table 2 (Fig. 1; Environment Canada, 2013). In 2008 these climate stations were within the peatlands' temperature-precipitation space described in Yu et al. (2009), ranging from -4.88 to 3.03 °C, and 607.8 to 683.5 mm (Table 2; Environment Canada, 2013).

TABLE 2

Climate data for 2008 from 3 stations in the JBL and adjacent regions.

Six cores—JBL1, JBL2, JBL3, JBL4, JBL7, JBL8— were from Sphagnum fuscum-dominated hummock surfaces and were permafrost free (Table 1). These hummocks existed in complexes with S. angustifolium hollows and intermediate S. magellanicum sections. Hummock-hollow patterning reflects Sphagnum species optimal growth and community ecology (Gunnarsson, 2005). Notable vascular species include tree species (Picea mariana, Larix larcina), shrub species (Chamaedaphne sp., Vaccinium sp., Salix sp., Betula sp., Eriophorum sp.), as well as Carex sp. and Equisetum sp. in hollows. JBL2 and JBL3 existed as bog islands in larger poorfen complexes. JBL4 was from the discontinuous permafrost zone and contained examples of boreal vegetation as well as subarctic Cladonia sp. on higher microsites. Two cores, JBL5 and JBL6, were from Cladonia-dominated surfaces underlain by Sphagnum-dominated permafrost peat (Table 1). JBL5 was a degraded permafrost plateau with large cracks filled with meltwater. JBL5's surface cover was predominantly Cladonia sp., Ledum sp., and unvegetated surfaces of decayed peat.

At each core site, the first 0.8 to 1 m was sampled with a box corer and the remainder with a Russian auger in non-permafrost peatlands and a motorized SIPRE drill in permafrost. Seven of the cores contained complete profiles from the surface to the mineral/ peat interface. JBL6 was recovered without the basal section and is therefore omitted from the pre-2 ka C accumulation and total C accumulation data sets. Cores were wrapped in plastic wrap and aluminum foil and placed in a freezer on arrival at the University of California, Los Angeles. All cores were subsampled, while frozen, in 1 cm increments using a bandsaw.

We supplemented our 8 cores with a review of 17 other cores from a broader geographical region encompassing the Boreal Shield, Hudson Plains region, and Taiga Shield surrounding the Hudson and James Bays. The entire data set ranges in latitude from 50°27′N to 60°50′N (Fig. 1; Tables 1 and 3). MAP for the entire network ranged from to 238.0 to 861.50 mm, and MAAT ranged from -8.72 to 0.35 °C from 1975 to 2005 (Matsuura and Willmott, 2009). Out of the combined data set, 19 of the sites are from ombrotrophic bogs, whereas 6 are from permafrost plateaus or polygonal peat formations (Tables 1 and 3). We avoided analyzing cores from fens or sites with chronologies that were less than approximately millennially resolved.

TABLE 3

Site information from 17 Hudson Bay Lowlands-JBL and adjacent regions ombrotrophic peat profiles with millennially resolved age-depth models, listed by author.

SOIL C ESTIMATES

We estimated percentage C for every 1 cm increment as the product of dry bulk density (BD), loss on ignition at 550 °C (LOI₅₅₀), and a %C assumption of 0.5 g C 1 g OM^-1 (Turunen et al., 2002). Samples for BD and LOI₅₅₀ were taken from subsampled cores with a stainless steel tube with a 1 cm diameter. Samples were measured lengthwise with digital calipers to calculate volume. Dry BD was measured by dividing the dry weight by the sample volume after drying to a constant mass at 105 °C. LOI₅₅₀ was calculated as the mass lost after one hour ignition in a muffle furnace at 550 °C (Sheng et al., 2004).

AGE DEPTH MODELING

Chronologies for the eight new cores were established using multiple ¹⁴C dates (Table 4). Dates were selected in order to have approximately millennially resolved chronologies before 1 ka, and bicentennially resolved chronologies since 1 ka. Plant macrofossils were separated from peat with distilled H₂O. Fungal hyphae and plant rootlets were removed using forceps under a dissecting microscope. Macrofossils were identified down to the genus level when possible using a stereomicroscope and Canadian macrofossil identification guides (Lévesque et al., 1988). Bulk peat was used when no single macrofossils were identified. Samples were pre-treated using acid-base-acid treatments at 65 °C to remove carbonates, humic acids, and dissolved organic C (Olsson, 1986). Samples were vacuum-sealed in quartz tubes with CuO powder and Ag wire and combusted for 4 h at 900 °C. Samples were graphitized, and atomized in an accelerator mass spectrometer at the Keck Laboratory at the University of California, Irvine. The ratio of ¹⁴C to ^l2C was calculated relative to blank and standards from dendrochronologically dated wood.

Multiple ¹⁴C dates (Table 4; Appendix Table A1) were calibrated, and an age-depth model was created for each core using ‘Bacon’ (Fig. 3 and Appendix Fig. A1), a flexible Bayesian age-depth modeling software (Blaauw and Christen, 2011) that is coded using the computer language R (R Development Core Team, 2012). Dates were calibrated using INTCAL09 (Reimer et al., 2009), and NH1 for post-bomb dates (Hua and Barbetti, 2007). We assumed a date of -58 BP for the surface of each new core except for JBL5, which had surface peat that dated pre-modern (15 ¹⁴C age; Table 4). Seventeen age-depth models from the previously published sites were recalculated using “Bacon” to standardize methods. The surface date was used as a calibration point if it was listed in the literature.

FIGURE 3.

‘Bacon’ age-depth models for 8 new JBL cores. The best estimates for calibrated age are shown in black. The upper and lower estimates are shown in gray.

We preferred ‘Bacon’ to linear interpolation between dates because the latter can be too restrictive if cores are longer than 1 m and do not have high-resolution dating (Blaauw and Heegaard, 2012). ‘Bacon’ deals with outliers in a standardized way using a robust Student's t method (Christen and Pérez, 2011). Bayesian algorithms require priors. We used the program's default settings for the priors: shape (2), memory strength (4), and memory mean (0.7). For the accumulation rate prior we used a default value of 20 yr cm^-1, and used 50 yr cm^-1 if the default did not produce a parsimonious age-depth model. Only JBL7 required a different accumulation prior of 40 yr cm^-1.

AVERAGE APPARENT RATE OF C ACCUMULATION

We calculated the mass of C accumulated from initiation to the surface, 2 ka depth to the surface, and initiation to 2 ka (Beilman et al., 2009). For average Holocene apparent rate of C accumulation, we summarized the mass of C and divided the mass by the amount of time since initiation. For post-2 ka apparent rate of C accumulation, we summarized the amount of C accumulated since the closest approximation to 2 ka and divided the mass by that closest approximation. For pre-2 ka apparent rate of C accumulation, we summarized the mass of C accumulated from initiation to 2 ka and divided it by the difference between initiation date and the closest 2 ka date.

STATISTICAL ANALYSIS

We statistically tested hypotheses regarding the relationships between climate and 2 ka depths, as well as permafrost occurrence on both total depths, and 2 ka depths. To test the hypotheses that MAP, MAAT, GDD₀, and PAR₀ correlate significantly with 2 ka depth, we used linear regressions, and climate data from 0.5 × 0.5° integrated gridded databases for MAP and MAAT (1975–2005; Matsuura and Willmott, 2009). Initial analysis and previous research (Beilman et al., 2009) indicated the possibility of an exponential relationship between MAAT and 2 ka depth, so we included an exponential regression in our analysis of these two variables. For measurements involving seasonality we used 0 °C as a base temperature because Sphagnum species are adapted to growth at low temperatures (Asada et al., 2003). We calculated a 0.5 × 0.5° gridded database for GDD₀ from daily temperature values and PAR₀ from latitude, orbital parameters, and the fraction of sunshine hours (Prentice et al., 1993; Hijmans et al., 2005), using the CLIMATE 2.2 database (Kaplan et al., 2003). We also tested the hypothesis that permafrost peatlands are significantly shallower than nonpermafrost peatlands, using a one-way ANOVA with permafrost occurrence as the independent variable, and total depth, or 2 ka depth, as the dependent variable. All statistics were calculated in R (R Development Core Team, 2012).

TABLE 4

¹⁴C-AMS samples, depth, ¹⁴C ages, and best fit “Bacon” model calibration estimates for 8 new southwest JBL cores.

Continued.

TABLE 5

Peatland basal depths, basal ages, and 2 ka depths for 17 review and 8 new JBL and adjacent regions cores. Data for loss on ignition at 550 °C (LOI₅₅₀), mean bulk density (BD), and apparent rate of C accumulation; C mass data since inception, post-2 ka, and pre-2 ka are displayed as well.

PEAT INITIATION ESTIMATES, DEPTHS, AND CHRONOLOGIES

A full report of ¹⁴C dates in the eight new cores is available in Figure 1 and Table 4. The timing of peat initiation in the eight new sites followed major deglaciation, glacial lake drainage, and land emergence around 8.34 ka (Dyke et al., 2003). Basal dates from the eight new cores vary between 7.8 ka and 4.4 ka, and have an average initiation time of 6.3 ± (std. dev.) 1.2 ka (Fig. 1; Table 5). Peat depth averaged 256 ± 98 cm and ranged from 141 to 422 cm (Table 5). Core lengths from previously published northeastern Canadian studies ranged from 70 to 483 cm, and had a mean of 266 ± 109 cm (Table 5). These mean depths roughly correspond to the average depths estimated for northern peatlands (2.3 m; Gorham, 1991).

FIGURE 4.

Post-2 ka C accumulated in the 8 southwest JBL sites arranged by latitude. (-) Represent the upper and lower estimates of Post-2 ka C masses based on upper and lower estimates of 2 ka by ‘Bacon’ models. (*) Represents permafrost peat.

SOIL C MASS IN CENTRAL AND NORTHERN ONTARIO

BD in the new cores ranged from 0.0034 to 0.62 g cm^-3 and averaged 0.093 ± 0.041 g cm^-3 (Table 5). LOI₅₅₀ ranged from 8.7 to 100.0 %OM and averaged 94.5 ± 6.3 %OM. The highest LOI_₅₀ generally corresponded to the low-density acrotelm and active layer sections. The exception to this is JBL5, which has a densely packed rootlet/lichen active layer (Table 5). The lowest LOI₅₅₀ values generally correspond to highest BD values at the mineral dominated basal sections of JBL3 and JBL5. Our mean organic matter density, 0.087 ± 0.03 is slightly lower than the mean value for fens and bogs in western Canada, 0.094 g cm^-3 (Vitt et al., 2000a), and the range of measurements from the Mackenzie River Basin, Finland, and the WSL (0.092–0.094 g cm^-3; Makila, 1994; Sheng et al., 2004; Beilman et al., 2009). There were generally lower BD and higher LOI₅₅₀ in the top meter of peat, and higher BD and lower LOI₅₅₀ in deeper peat due to decay and compaction over time in deeper peat.

More C was older than 2 ka than was younger in the eight new cores. This is not surprising as the pre-2 ka portion of the cores typically represents >5 ka to 2.4 ka of deposition. Pre-2 ka stocks range from 28.3 to 108.7 kg C m^-2 and averaged 67.5 ± 30.0 kg C m^-2. Post-2 ka C stocks range from 16.3 to 62.5 kg C m^-2 and averaged 40.5 ± 14.3 kg C m^-2 (Fig. 4 and Table 5). In central and northern Ontario, about 40% of the total C present in the cores is younger than 2 ka. Notable exceptions to this trend were JBL4 and JBL8 where pre-2 ka apparent rate of C accumulation was drastically lower than post-2 ka (Table 5).

Apparent rate of C accumulation was generally higher post-2 ka than total in the eight new cores. Average apparent rate of C accumulation since initiation ranged from 12.8 to 26.7 g C m^-2 yr^-1, with a mean of 17.4 ± 5.0 g C m^-2 yr^-1 (Table 5). These values correspond to estimates of 13 to 30 g C m^-2 yr^-1 for North American peat bogs (Gorham, 1991; Turunen et al., 2002; Kuhry and Turunen, 2006), and fall in the range of observed global peatland C accumulation (8.4 to 38.0 g C m^-2 yr^-1; Yu et al. 2009). Post-2 ka apparent rate of C accumulation ranged from 8.5 to 30.8 g C m^-2 yr^-1 and had an average of 20.2 ± 6.9 g C m^-2 yr^-1. The lowest post-2 ka estimate in JBL5 (8.5 g C m^-2 yr^-1) roughly corresponds with projected rates for subarctic Canada (9 g C m^-2 yr^-1; Gorham, 1991), and the lowest observed arctic estimates from a global synthesis (8.4 g C m^-2 yr^-1; Yu et al. 2009). Pre-2 ka apparent rate of C accumulation was generally lower than post-2 ka apparent rate of C accumulation and ranged from 8.2 to 24.8 g C m^-2 yr^-1 and averaged 15.0 ± 5.6 g C m^-2 yr^-1 (Table 5).

In the eight new sites, both apparent rate of C accumulation and 2 ka depth correlate significantly and positively with MAP with r² values of 0.56 and 0.58, respectively (p <0.05; Table 6). 2 ka depths also correlate significantly and positively with MAAT (p < 0.05; r² = 0.52; Table 6).

CLIMATE AND 2 ka DEPTH

Based on the combined data set of our 8 new cores and the 17 previously published cores, peat depth at 2 ka correlated significantly and positively with MAP, MAAT, GDD₀, and PAR₀. Depths at 2 ka ranged from 8 to 173 cm with a mean of 88.5 ± 41.1 cm (Table 5). Overall peat depth ranged from 70 to 483 cm, meaning that the average proportion of peat vertical accumulation since 2 ka was 36.9 ± 19.1%, roughly similar to the 40% of C present that was younger than 2 ka in our eight new sites. Depth at 2 ka correlates significantly and positively with MAP (p < 0.05; r² = 0.20; Fig. 5; Table 6), MAAT (p < 0.001; r² = 0.46; Fig. 5; Table 6), GDD₀ (p < 0.0001; r² = 0.60; Fig. 5; Table 6), and PAR₀ (p < 0.0001; r² = 0.62; Fig. 5; Table 6). An exponential regression between MAAT and 2 ka depth increased the r² value to 0.49 (p < 0.0001; Fig. 5; Table 6). A linear regression defining PAR₀ as a predictor of 2 ka depth describes the most variance out of all of the variables we tested (Fig. 5; Table 6).

PERMAFROST OCCURRENCE AND PEAT DEPTH

Permafrost occurrence was found to have a significant effect on 2 ka depths, and on total depths in the 25 new and review cores. A one-way ANOVA determined that there was a statistically significant difference between the depths of permafrost peatlands, and non-permafrost peatlands (p < 0.05). A one-way ANOVA showed that permafrost peatlands have a significantly shallower 2 ka depth than non-permafrost peatlands (p < 0.0001; Fig. 6). Permafrost total depths averaged 182.8 ± 31.6 cm, whereas non-permafrost total depth averages 290.6 ± 112.0 cm. Permafrost 2 ka depths averaged 41.3 ± 23.6 cm for permafrost peatlands, while non-permafrost bogs averaged 103.4 ± 33.4 cm (Fig. 6).

PEATLAND INITIATION

Peatland initiation followed major deglaciation in central and northern Ontario. The oldest site (JBL3) occurred in an area that deglaciation models indicate was at, or near, major glacial lake drainage (Dyke et al., 2003). Basal dates show a lag between land availability and peat initiation in North America (Hasley et al., 2000; Gorham et al., 2007), and this is likely the case in eastern Canada. Previous studies describe an average 4000-year lag between deglaciation and widespread peat development due to the time it takes for land to become amenable to peat formation, and due to the random chance of dispersal and colonization of peat-forming species (Hasley et al., 2000). The drainage of post-glacial lakes around 8.34 ka is a likely lag factor (Hasley et al., 2000). Another factor may be a delayed Holocene Thermal Maximum in the North Atlantic caused by cooling from the remainder of the decaying Laurentide ice sheet as late as 6 ka (Kaplan and Wolfe, 2006).

FIGURE 5.

Four scatterplots correlating 2 ka depth as a function of environmental variables. Figures are arranged from left to right and top to bottom in order of the linear model's r² value. Solid lines represent the best fit for linear regressions. Dashed lines represent the best fit for an exponential regression. Triangles represent new JBL site data. Circles represent review data.

TABLE 6

Correlation statistics for James Bay Lowland (JBL) peatland 2 ka apparent rate of C accumulation, 2 ka depths, and Hudson Bay Low-lands (HBL)–JBL synthesis with mean annual precipitation (MAP), mean annual air temperature (MAAT), growing degree days above 0 °C (GDD₀), and photosynthetically active radiation integrated over the growing season (PAR₀).

FIGURE 6.

Bar and whisker plot of 2 ka depths in permafrost and non-permafrost peat. Central lines represent medians. Box edges represent the 25% upper and lower quantiles. Whisker lines represent least and greatest values.

There was a high rate of peat initiation between 6 and 7.8 ka in central and northern Ontario (Fig. 1). However, the eight new initiation estimates do not show a clear stratification with newer initiation occurring closest to the James Bay coast. The youngest core (JBL8) occurs at the southernmost end of the transect, whereas the oldest (JBL3) occurs in the middle (Fig. 1). Stratification of initiation was discussed by a study closer to the southwest coast of the JBL near the Albany River, where rates of isostatic uplift are the highest (Fig. 1; Glaser et al., 2004). This difference is likely due to the fact that catastrophic drainage of glacial lakes Agassiz and Ojibway were responsible for land emergence in some of the eight new sites, rather than gradual isostatic uplift that drives initiation in the Albany River region (Glaser et al., 2004).

C ACCUMULATION IN CENTRAL AND NORTHERN ONTARIO

C accumulation measurements in the eight new sites showed that 40% of C was younger than 2 ka in Sphagnum bogs. This is roughly comparable to estimates for WSL, where 41% of C is younger than 2 ka (Sheng et al., 2004; Beilman et al., 2009). Only two of our sites, JBL4 and JBL8, have more post-2 ka C than pre-2 ka C, which may have been caused by low relative pre-2 ka accumulation, relatively high belowground C mineralization, or a combination of both. Apparent rate of C accumulation is the result of autogenic drivers, allogenic drivers, and complex internal ecohydrological feedbacks, and has the potential to vary substantially among peatlands (Belyea and Baird, 2006).

MAAT, GDD₀, AND PAR₀ AS DRIVERS OF POST-2 KA VERTICAL PEAT ACCUMULATION

In our analysis, differing scales showed different correlations between post—2 ka apparent rate of C accumulation, 2 ka depth, and four climate variables. In the eight new sites, MAP correlated significantly with both post—2 ka apparent rate of C accumulation and 2 ka depth. This correlation was also significant and positive in our review data set, but was low relative to MAAT, GDD₀, and PAR₀. MAP may exercise some local control over peat accumulation in the Boreal Shield and Hudson Plains of Ontario, but temperature, seasonality, and photosynthetically active radiation are likely the major drivers of vertical peat accumulation in all peatlands bordering the Hudson and James Bays. This supports the conclusions of a recent global synthesis of peatland C accumulation rates. Water table depths need to be persistently high for peat to form; however, if they are present, they do not explain any of the variance of C accumulation rates (Charman et al., 2013). It is also possible that small sample size in the new sites (n = 8) may lower the data set's descriptive power.

In the overall synthesis, there was a significant and positive correlation between 2 ka and MAP (p < 0.05; Table 6); however, we did not observe the inverse relationship with precipitation similar to that reported by a broad survey of North American peatlands (Gorham et al., 2003). This may be due to the topographic controls in southern sites (Gorham et al., 2003) that are not present in the more topographically homogeneous Hudson Plains, which we heavily sampled. This was also not a robust relationship because the correlation was heavily influenced by EL1 site data (MAP = 208 mm, 2 ka depth = 8 cm). If that point is deleted, the r² value lowers to 0.087, and the statistical significance is eliminated.

Our study showed that MAAT had a significant effect on vertical peat accumulation since 2 ka in northeastern Canada (Fig. 5; Table 6). This is analogous to the relationship found between MAAT and 2 ka depth in WSL; however, there are key differences between the two areas. In WSL, data fits an exponential regression that explains more variation (r² = 0.82; Beilman et al., 2009) than does our data's exponential regression (r² = 0.49; Table 6). In WSL there was less variability in 2 ka depth for permafrost sites, as well as deeper 2 ka depths in southern non-permafrost bogs (Beilman et al., 2009).

Our hypotheses, that GDD₀, and PAR₀ are major drivers of peat accumulation, were supported by significant and positive correlations between these variables and 2 ka depths. PAR₀ had the highest correlation (p < 0.0001, r² = 0.62; Table 6). In a previous study that analyzed global Sphagnum productivity, a linear regression between PAR₀ and Sphagnum productivity explained a significant but comparatively small amount of variance (r² = 0.23; Loisel et al., 2012). PAR₀ integrates photosynthetically active radiation over the growing season length and can potentially influence two different aspects of peat formation. Photosynthetically active radiation and growing season length may both drive plant growth, and thus litter input (Clymo et al., 1998). Growing season length also affects the length of time each year that the ground is unfrozen and biomass can pass from the acrotelm to the catotelm (Clymo et al., 1998; Gorham et al., 2003). Our results also support a global study of peatland apparent rate of C accumulation over the last 1000 years, in which GDD₀ and PAR₀ were found to correlate significantly with post-1 ka apparent rate of C accumulation (Charman et al., 2013). However, the significance of radiation as a driver of Sphagnum growth has been debated because of the high degree of shading in boreal peat bogs, and because of photoinhibition in Sphagnum photosynthetic tissues (Harley et al., 1989; Murray et al., 1993).

The relationship of growing season length and photosynthetically active radiation on vertical peat accumulation is important because northern latitudes will be disproportionately affected by global warming. Models of future anthropogenic climate forcing predict disproportionate seasonality changes in northern latitudes, and increases in the length of the growing season (Pachauri and Reisinger, 2007). Between 1982 and 1999 the effective start of the growing season advanced 5.4 days in the northern hemisphere (Jeong et al., 2011). Remote sensing images indicate that the plant growth in the Canadian Low Arctic has increased 0.46–0.67% yr¹ from 1982 to 2006 (Jia et al., 2009). However, there will likely be complications modeling PAR₀ as a predictive bioclimatic variable because of the uncertainty in model prediction of future cloud cover (Charman et al., 2013).

THE EFFECT OF PERMAFROST ON PEAT DEPTH

Significantly shallower basal depths and 2 ka depths in permafrost peatlands compared to peatlands that were permafrost-free supported our hypothesis that permafrost occurrence inhibits peat accumulation and that permafrost occurrence significantly decreased post-2 ka vertical accumulation. This was supported by significantly shallower 2 ka depths in permafrost peat, compared to non-permafrost peat (p < 0.0001). Previous studies indicated that permafrost likely formed in western Canada during the Little Ice Age (Vitt et al., 2000b) during the late Holocene. However, the timing of permafrost formation in eastern Canada is not as well understood. Overall shallower permafrost peat may be explained either by permafrost establishment earlier than 2 ka, or by general lower productivity due to constantly lower MAAT, GDD₀, and PAR₀ without the direct influence of permafrost.

There was consistent vertical growth in peatlands post-2 ka. Only two cores, JBL5 and EL1, contained near-surface macrofossils that dated pre-modern. JBL5 had dates of 204 and 207 yr BP at 8 and 12 cm, respectively (Table 4), and EL1 had a 13 cm depth dating near 3.0 ka (Appendix Table A1). Besides these two sites, there is little evidence for the recent shutdown by permafrost of 2 ka C accumulation described in the WSL (Beilman et al., 2009). This study showed reduced but ongoing vertical accumulation, similar to the described activity in a survey of subarctic palsas (Olefeldt et al., 2012).

Controls by permafrost of late Holocene peat growth show that productivity, rather than decay, currently drives the significant climatically related differences in peat growth observed here. However, the resulting changes in hydrology and CH₄ emissions from melting permafrost will likely result in a greater uncertainty in peatland C cycle feedbacks (Pachauri and Reisinger, 2007). Past studies of northern and southern regions in the Hudson Plains show that CH₄ emissions were lower than previously estimated (Roulet et al., 1994). If non-permafrost peatlands store C faster than permafrost peatlands, then peatlands could act as a negative feedback to future warming (Turetsky et al., 2007; Beilman et al., 2009) or at least as transient reactions to warming. The effect of permafrost melt on hydrology is also controversial. Some studies project that permafrost melts will increase drainage (Jorgenson and Osterkamp, 2005; Riordan et al., 2006), while others claim that it will increase peatland extent (Payette et al., 2004).