UEF//eRepository
DSpace https://erepo.uef.fi
Rinnakkaistallenteet Luonnontieteiden ja metsätieteiden tiedekunta
2018
Near-surface permafrost aggradation in Northern Hemisphere peatlands shows regional and global trends during the past 6000 years
Treat, Claire C
SAGE Publications
article
info:eu-repo/semantics/acceptedVersion
© Authors
All rights reserved
https://doi.org/10.1177/0959683617752858
https://erepo.uef.fi/handle/123456789/5813
Downloaded from University of Eastern Finland's eRepository
Near-surface permafrost aggradation in Northern Hemisphere peatlands shows regional 1
and global trends during the past 6000 years 2
3
Claire C. Treat1*, Miriam C. Jones2 4
1University of Alaska Fairbanks, Fairbanks, AK, USA 5
2U.S. Geological Survey, Reston, VA, USA 6
* Now at Department of Environmental and Biological Sciences, University of Eastern Finland, 7
Kuopio, Finland 8
Correspondence to: Claire Treat (Claire.treat@uef.fi) 9
10
Accepted at The Holocene on 9 December 2017 11
12
Abstract 13
The history of permafrost aggradation and thaw in northern peatlands can serve as an indicator of 14
regional climatic history in regions where records are sparse. We infer regional trends in the 15
timing of permafrost aggradation and thaw in North American and Eurasian peatland ecosystems 16
based on plant macrofossils and peat properties using existing peat core records from more than 17
250 sites. Permafrost was continuously present in peatlands during the last 6000 years in some 18
present-day continuous permafrost zones and formed after 6000 BP in peatlands in the isolated to 19
discontinuous permafrost regions. Rates of permafrost aggradation in peatlands generally 20
increased after 3000 BP and were greatest between 750 and 0 BP, corresponding with neoglacial 21
cooling and the Little Ice Age (LIA), respectively. Peak periods of permafrost thaw occurred 22
after 250 BP, when permafrost aggradation in peatlands reached its maximum extent and as 23
temperatures began warming after the LIA, suggesting that permafrost thaw is likely to continue 24
in the future. The broader correlation of permafrost aggradation in peatlands with known climatic 25
trends and other proxies such as pollen records suggests that this record can be a valuable 26
addition to regional climate reconstructions.
27
1 Introduction 28
Widespread permafrost aggradation and degradation in high latitude peatlands highlights 29
the importance of long-term permafrost dynamics in northern peatlands. The recent permafrost 30
degradation and thaw that has been observed in peatlands in Alaska (Jorgenson et al., 2006;
31
Wickland et al., 2006; Jones et al., 2016a), Canada (Vitt et al., 2000; Payette et al., 2004; Sannel 32
and Brown, 2010), and Scandinavia (Johansson et al., 2006; Hodgkins et al., 2014) is predicted 33
to continue due to climatic warming (Lawrence et al., 2012). The timing of near-surface 34
permafrost aggradation and degradation within the northern high-latitude permafrost zone, where 35
peat formation initiated often millennia after the Last Glacial Maximum (LGM) (MacDonald et 36
al., 2006), remains unevaluated at a broad, regional scales. Similarly, whether widespread 37
permafrost thaw is a new phenomenon in peatlands or has occurred previously during the 38
Holocene in response to warmer temperatures remains largely unknown. Hence, an important 39
gap remains in understanding the timing and rates of permafrost aggradation and degradation in 40
peatlands across northern high latitudes, which has important implications for carbon (C) storage 41
in permafrost peatlands as climate warms and permafrost thaws (Jones et al., 2016b; O'Donnell 42
et al., 2012; Schneider von Deimling et al., 2012).
43
Permafrost is ground that remains frozen for more than two consecutive years and occurs 44
both where mean annual air temperatures (MAAT) are less than 0º C, although it remains in 45
some regions where the present-day MAAT are greater than 0ºC and the permafrost is insulated 46
by vegetation or peat (Halsey et al., 1995; Shur and Jorgenson, 2007). Climate plays a role in 47
permafrost aggradation because air temperatures must be cold enough to result in perennially 48
frozen ground (Shur and Jorgenson, 2007). However, mean annual air temperatures alone are 49
not enough to determine whether permafrost will aggrade or thaw. Permafrost aggradation and 50
thaw have been linked to a range of local and regional factors including the colonization of 51
peatland surfaces by Sphagnum species (spp.), microtopography, tree and shrub cover, snow 52
thickness, snow distribution, and disturbance (Allard and Seguin, 1987; Camill, 2000; Camill, 53
2005; Camill and Clark, 1998; Zoltai and Tarnocai, 1975; Payette et al., 2004; Johansson et al., 54
2013). These local factors cause changes in the soil thermal regime and can result in decreased 55
thermal conductivity during the summer or increased exposure to cold winter temperatures 56
(Halsey et al., 1995; Seppälä, 1994; Seppälä, 2011; Zoltai, 1993; Zoltai and Tarnocai, 1975;
57
Oksanen et al., 2003; Zoltai, 1995), which can ultimately result in permafrost formation given 58
sufficiently cold temperatures. Permafrost thaw in peatlands can be associated with climate 59
warming (Halsey et al., 1995) and local factors such as hydrologic changes, disturbance 60
(including wildfire), and increased snow cover (Camill, 2005; Johansson et al., 2006; Payette et 61
al., 2004; Zoltai, 1993). The combination of site factors and climate factors are key to 62
determining both the history of permafrost dynamics and future responses to climate change with 63
warming temperatures (e.g. Seppälä, 2011).
64
Northern peatlands have experienced both warmer and cooler temperatures during the 65
Holocene. Whether permafrost aggradation or degradation occurred broadly in peatlands as a 66
result of these differing climatic conditions is unknown. Temperatures were warmer than present 67
during the time-transgressive Holocene Thermal Maximum (HTM) (Kaufman et al., 2004).
68
Peatland initiation, accumulation, and expansion rose sharply on the landscape during the HTM 69
(Smith et al., 2004; MacDonald et al., 2006; Jones and Yu, 2010). In nearly all regions, the HTM 70
preceded the period of study; the latest occurrence of the HTM in North America was in the mid- 71
Holocene in Eastern Canada (Kaufman et al., 2004). Whether permafrost aggradation and/or 72
permafrost thaw in peatlands occurred under the warmer conditions during the HTM is unknown 73
but is highly relevant given future climatic warming. Subsequent neoglacial cooling may have 74
set the stage for permafrost aggradation in peatlands. Decreasing summer insolation across the 75
northern high latitudes contributed to cooler-than-present temperatures following the HTM 76
(Berger and Loutre, 1991), in particular during neoglaciation (Sharp, 1960) and the Little Ice 77
Age (LIA) (Alley, 2000; Marcott et al., 2013). Colder temperatures resulted in permafrost 78
aggradation during the Holocene across northern high latitudes (Mann et al., 2010; Mann et al., 79
2002). For example, detailed plant macrofossil analyses from peat cores in Canada show 80
permafrost aggradation occurred in numerous peatland sites coinciding with cooling at the end of 81
the HTM around 4000 BP as evidenced by vegetation changes from wet fen vegetation to drier, 82
forested bog species (Zoltai 1995).
83
The timing of near-surface permafrost aggradation in peatlands (hereafter “permafrost 84
aggradation”) is an important controlling factor for determining the potential for C loss with 85
permafrost thaw (Jones et al., 2016b; Treat et al., 2014). If permafrost aggradation results in the 86
incorporation of relatively undecomposed material into permafrost (i.e. syngenetic permafrost), 87
then the potential C losses found in soil incubations are similar to C losses from surface soils, 88
whereas thawing of highly decomposed material in peat deposits with epigenetic permafrost 89
results in small C losses (Treat et al., 2014; Lee et al., 2012). Given that in the present-day 90
permafrost zone, permafrost peatlands (histels) comprise ~ 19% of the area but contain 40% of 91
the soil organic C (top 3m of soil; Hugelius et al., 2014; Tarnocai et al., 2009), understanding 92
how permafrost aggradation and thaw impacts biogeochemical cycling of C is important for 93
understanding feedbacks to warming. The timing of permafrost aggradation and thaw in 94
peatlands can be inferred from a combination of detailed plant macrofossil analysis, 95
physicochemical peat properties, and detailed chronologies (Camill et al., 2009; Oksanen et al., 96
2003; Treat et al., 2016).
97
Here, we use a dataset of peat properties and peatland vegetation community 98
reconstructions to identify the timing of peatland initiation, permafrost aggradation, and 99
permafrost thaw in 266 cores from across the northern hemisphere for the last 6000 years (Treat 100
et al., 2016). Specifically, we ask whether there are coherent, regional trends in the timing of 101
permafrost aggradation and thaw in northern hemisphere peatlands and how they relate to 102
paleoclimate.
103
2 Methods 104
2.1 Dataset development: ecosystem classification, age-depth models & synthesis 105
Permafrost aggradation can result in changes in vegetation that can be preserved in 106
organic soil horizons, which generally contain both plant macrofossil records and means to 107
obtain good chronologic constraints (Zoltai and Tarnocai, 1971; Zoltai and Tarnocai, 1975).
108
Therefore, this study focused on the aggradation of permafrost in peatlands. We compiled 109
records of plant macrofossils, radiocarbon dates, lithologies, and peat properties from cores from 110
441 peatland cores within the boreal and tundra regions of North America, Europe, and Asia 111
using methods described in detail in a previous study (Treat et al., 2016). We selected cores that 112
were well-described using the criteria below. We were interested in the development of 113
permafrost during the Holocene, and given the climatic variability during the Holocene, we 114
selected peat cores from the regions that contained permafrost during the LGM (Vandenberghe 115
et al., 2014). While there were numerous records and even permafrost aggradation prior to 6000 116
BP (Figure 1; Table S1), we focused on the period after 6000 BP in this analysis in order to have 117
a higher data density to conduct an analysis of regional trends in permafrost aggradation.
118
We selected cores from the larger dataset that met the following criteria: 1) organic soils 119
> 30 cm thick that contained a minimum of 65% organic matter); 2) available plant macrofossil 120
assemblages to classify the peatland environmental type, including the presence/absence of 121
permafrost; 3) chronologic control of one or more dates for every 2000 years; 4) location within 122
the present-day permafrost zone (Brown et al., 1998, revised 2001) or within the zone of 123
permafrost at the Last Permafrost Maximum (LPM), 18-21 ka BP (Figure 1) (Vandenberghe et 124
al., 2014; Lindgren et al., 2015). This resulted in the inclusion of 266 cores from 214 sites across 125
the northern hemisphere (Figure 1; Table S1). These data are summarized in Table S1; the 126
complete dataset including site information, peat properties, plant macrofossils and chronologic 127
information is available through PANGAEA (https://doi.org/10.1594/PANGAEA.863697).
128
Sections of peat cores were classified into wetland classes including permafrost-free fens, 129
permafrost-free bogs, permafrost peatlands (including peat plateaus, palsas, bogs and fens with 130
permafrost, polygonal peat plateaus, high- and low-center polygons, and tundra with > 30 cm of 131
organic soils in present-day), thawed permafrost (including collapse-scar fens, bogs, and thaw 132
ponds), and other (peatland pools, marshes, swamps, and ponds, lakes, and upland forests that 133
later develop peatlands) using the classification scheme described by Treat et al. (2016) and in 134
more detail below regarding the delineation of permafrost dynamics. Briefly, plant macrofossil 135
assemblages, detailed descriptions of lithology, and peat properties were used to classify peat 136
core sections into wetland classes based on the Canadian Wetland Classification system (Group, 137
1988; Treat et al., 2016). This approach relied heavily on the original authors interpretation of 138
the plant macrofossil data because there is no single indicator species of permafrost formation 139
(see discussion below; Oksanen and Väliranta, 2006).
140
In this study, we derived new age-depth models for each core based on the reported 141
chronology using BACON (Blaauw and Christen, 2011) and IntCal13 (Reimer et al., 2013). We 142
assumed that peat core surfaces were from the year of sampling, unless specified otherwise in the 143
original dataset. The wetland classification of each peat core was converted from the depth-scale 144
to a time scale using the age-depth model for each core and subsequently binned into 250-year 145
age bins for analysis and plotted using bin midpoints. We use calibrated 14C ages throughout the 146
text and abbreviate “cal yr BP” as simply “BP”.
147
2.2 Determination of potential permafrost aggradation 148
The presence of permafrost in a peatland site at the time of sampling considerably 149
simplifies the identification of permafrost aggradation. In boreal regions, surface uplift from ice 150
expansion within the permafrost results in a vegetation shift to species indicative of dry 151
conditions (Seppälä, 2011; Zoltai and Tarnocai, 1975; Zoltai et al., 1988). Plant macrofossil 152
analysis can be used to identify the vegetation shifts to dry, forested communities associated with 153
permafrost aggradation from relatively wet vegetation communities associated with fens and 154
bogs (Zoltai and Tarnocai, 1975). While this is generally evident in peat cores from a transition 155
from wet, sedge-dominated peat to forest peat or Sphagnum-woody peat (Camill et al., 2009;
156
Jones et al., 2013; Kuhry, 2008; Zoltai and Tarnocai, 1975), more detailed analysis is required 157
due to the similarities between dry bogs and permafrost peatland species (Camill et al., 2009;
158
Jones et al., 2013; Oksanen, 2006). Moss species associated with permafrost aggradation include 159
Polytrichum spp., Pleurozium spp., Tomenthypnum nitens, Dicranum elongatum, and some 160
hummock-forming Sphagnum mosses, generally Sphagnum sect. Acutifolia (Kuhry, 2008;
161
Camill et al., 2009; Jones et al., 2013; Oksanen, 2006; Zoltai, 1993; Sannel and Kuhry, 2009). In 162
near surface peat, an increase in lichen abundance (Cladina spp. and Cladona spp.) and fungal 163
sclerotia is also commonly associated with permafrost aggradation (Kuhry, 2008; Oksanen, 164
2006; Zoltai and Tarnocai, 1975; Zoltai et al., 1988; Camill et al., 2009). The alternation of 165
Sphagnum fuscum peat and rootlet layers has also been used to identify permafrost aggradation 166
and persistence in both Western Canada and European Russia (Sannel and Kuhry, 2008; Sannel 167
and Kuhry, 2009; Oksanen et al., 2003).
168
In continuous permafrost zones and tundra sites, indicators of permafrost aggradation 169
take somewhat different forms. In many places, permafrost may have formed during or prior to 170
the LGM (e.g. Kanevskiy et al., 2014; Vandenberghe et al., 2014) and may be present in the 171
mineral soil underlying at the time of peat inception (Zoltai and Tarnocai, 1975). In this case, a 172
mixing of peat and mineral soils indicative of cryoturbation at the peat-mineral soil interface has 173
been used to indicate permafrost presence at the time of peat formation (Zoltai and Tarnocai, 174
1975). The presence of permafrost in the years to centuries following peat inception was unlikely 175
for peats formed in drained thaw lake basins, a relatively common occurrence in Arctic regions 176
of Alaska and Siberia (de Klerk et al., 2011; Jones et al., 2012; Walter Anthony et al., 2014;
177
Bockheim et al., 2004). In these cases, the timing of permafrost aggradation has been identified 178
using species composition shifts towards drier conditions, including from Carex-dominated 179
tundra fen vegetation to dry tundra vegetation (Jones et al., 2012) or from the development of dry 180
microforms associated with patterned ground permafrost features (Davis, 2001). In low-center 181
polygons formed in drained thaw lake basins, permafrost aggradation can be identified by the 182
development of patterned ground features such as polygon rims or ridges. Some species 183
associated with polygon rims or ridges include mosses such as T. nitens, Hylocomnium splendens, 184
Sphagnum cf. subsecundum, S. teres, S. warnstorfia, and shrubs including Salix spp. (de Klerk et 185
al., 2011). The polygon ridges are easily distinguishable from the wetter, Carex- and 186
Eriophorum- dominated polygon low centers with mosses such as Calliergon giganteum and 187
Drepanocladus revolvens (Zoltai and Tarnocai, 1975; Tarnocai and Zoltai, 1988; de Klerk et al., 188
2011), which also occur in wet, permafrost-free tundra fen sites.
189
The delineation of the timing of permafrost aggradation and thaw also becomes more 190
difficult with a more complex site history, including instances of partial or complete thaw 191
resulting in the absence of permafrost in the present-day. In the discontinuous permafrost region 192
of North America, specifically, in Alaska and Western Canada, the clearest indicator of past 193
permafrost is a sequence indicative of permafrost aggradation followed by thaw, similar to the 194
transitions between species observed on the margins of and lawns of collapse scar features in the 195
present-day (Jones et al., 2013). In these records, species in the macrofossil record frequently 196
indicated dry, treed peat plateaus with Picea mariana, lichens, and fungal sclerotia, followed by 197
a distinct, abrupt transition to wetter conditions with species such as Sphagnum riparium, Carex 198
spp., and Eriophorum spp., and the decrease/disappearance of Picea mariana, all indicating 199
permafrost thaw (e.g. Zoltai, 1993; Jones et al., 2013; Jones et al., 2016b; Kuhry, 2008; Oksanen 200
et al., 2003; Sannel and Kuhry, 2008).
201
Permafrost aggradation can also result in changes in peat properties. Due to the drier 202
conditions associated with surface uplift due to ice expansion, the degree of peat decomposition 203
shifts to a more or highly humified peat (e.g. Camill et al., 2009; Sannel and Kuhry, 2008;
204
Sannel and Kuhry, 2009; Kuhry, 2008; Oksanen et al., 2003) and apparent C accumulation rates 205
generally slow (Zoltai et al., 1988; Jones et al., 2012; Treat et al., 2016; Oksanen, 2006; Sannel 206
and Kuhry, 2009). Using a large synthesis dataset from the permafrost region, Treat et al. (2016) 207
found higher C/N ratios in peats that were likely deposited after permafrost aggradation, 208
suggesting that C/N ratios could provide an additional evidence for the timing of permafrost 209
aggradation. More advanced chemical analysis has also been used to characterize the timing of 210
permafrost aggradation with varying degrees of success, but these newly developed approaches 211
have not yet been applied widely (Ronkainen et al., 2015; Routh et al., 2014).
212
It is difficult, if not impossible, to determine the “exact” rather than the “potential” timing 213
of permafrost aggradation. Using detailed vegetation surveys in adjacent areas with and without 214
permafrost, previous studies have shown that there is no single indicator species of permafrost 215
formation (Oksanen and Väliranta, 2006). For example, in the boreal region, the species 216
composition of boreal dry bogs and permafrost peat plateaus is similar. Both are generally 217
distinguished by the transition to drier species, often making a determination of the timing of 218
permafrost aggradation imprecise (Camill et al., 2009). Slow peat accumulation rates or erosion 219
of surface peat due to windscour or other factors can remove important parts of the peat record 220
(Ronkainen et al., 2015; Peteet et al., 1998), introducing difficulties for determining the timing of 221
permafrost aggradation or other environmental changes precisely. Still, a careful multi-proxy 222
analysis of plant macrofossils, peat accumulation rates, degree of decomposition, and peat 223
chemistry (including carbon, nitrogen, and hydrogen content) offers the best chance at 224
identifying this important aspect of ecosystem history.
225
2.3 Regional trends 226
We aggregated the peat cores regionally in order to analyze trends in permafrost 227
aggradation and thaw. The regional analysis was based on administrative unit boundaries and 228
geographic positions where the cores were located (Figure 1). Regions in North America 229
included Eastern North America, Central North America (Eastern Canadian Rockies, Hudson 230
Bay Lowlands, the Great Lakes, and Ontario), Alaska and British Columbia, and the Canadian 231
Arctic (Yukon Territory, Northwest Territories, and Nunavut). Regions in Eurasia included the 232
Fennoscandia (including the Kola Peninsula in Russia), European Russia, West and Central 233
Siberia (including the Lake Baikal Region, n=1), and Eastern and Far Eastern Siberia.
234
The number of cores in each wetland classification were summed within each region for 235
each time bin in order to understand regional trends in fen-to-bog transitions (an increase in the 236
number of bog cores and a decrease in fen cores), permafrost aggradation (an increase in the 237
number of cores with permafrost), and permafrost thaw (a decrease in the number of cores with 238
permafrost). When discussing the regional trends in rates of permafrost aggradation and thaw, 239
we use the normalized rate of transition. The normalized rate of transition (dNa/dt) for a given 240
peatland type a is calculated using Equation 1, where N is the number of peatland cores.
241
!"#
!" =!"!!!! !"!!
!! [1]
242
This approach reduces bias introduced by varying data density (number of cores) within each 243
region, and specifically looks at changes occurring relative to the existing peat cores. The mean 244
normalized rates and standard errors of rates from Equation 1 are calculated for the time periods 245
of interest. Both approaches assume that an increase in the number of cores or percentage of 246
cores correlates with an increasing peatland area and that cores in this study are representative of 247
peatland dynamics as a whole. However, without accurate maps and areal estimates of each type 248
of peatland area regionally, we cannot evaluate whether the number of samples from each 249
peatland type is truly representative.
250
For comparison between the records of permafrost aggradation and thaw in peatlands to 251
regional and hemispheric climatic trends (e.g. Figure 2), we used existing paleoclimatology 252
records. We used the modeled summer solar insolation at 65°N (Berger and Loutre, 1991) 253
because of the correlation between radiation and peatland C accumulation at shorter time scales 254
(Charman et al., 2013). We used percent melt data from the Agassiz Ice Cap in the Canadian 255
Arctic, which is indicative of summer temperatures routinely increasing to above freezing 256
(Fisher et al., 1995) and correlated well with sea ice records from this region (Vare et al., 2009).
257
While it is likely that regional temperature trends diverged from the local trends in the Canadian 258
Arctic, given its high-latitude location, continuous nature, proximity to sites in North America, 259
and lack of other continuous regional records, it is included in this analysis.
260
3 Results 261
3.1 Northern high latitudes 262
Over the past 6000 years, the number of peatland sites increased by 63% from 164 cores 263
in 6000 BP to 266 cores in 0 BP (Figure 2g). During this time, the percentage of permafrost-free 264
fens decreased from 62% to 23%, despite occurring in ~100 sites until 750 BP (Figure 2c). After 265
750 BP, the number of permafrost-free fens decreased rapidly by 40%, from 96 cores to 62 cores 266
at 0 BP. The number of permafrost-free bogs increased from 20% of cores (Figure 2d, n=35) to 267
30% of cores (n= 79) over this same period. Permafrost aggradation in peatlands increased from 268
less than 10% of peatland cores (n=15) in 6000 BP to 40% of peatland cores at 0 BP (Figure 2e, 269
n=111). Overall, rates of permafrost aggradation were greatest between 750 BP and 0 BP, 270
developing in 55 new cores during this period, or at a rate of 2.2 ± 0.4 percent of cores per 271
century. From 6000 BP to 1000 BP, permafrost aggraded at a rate of 0.4 ± 0.1 percent of cores 272
per century although rates were slightly higher between 2750 and 2250 BP. Permafrost thaw was 273
greatest between 250 BP and present, occurring in 10 of 266 cores, or less than 4% of cores.
274
Rates of permafrost thaw were low, averaging 0.07 ± 0.05 percent of cores per century.
275
The location of sites was important for determining the timing of permafrost aggradation.
276
Permafrost aggraded earliest in sites in the continuous permafrost zone (3500 ± 500 BP, where 277
error is standard error among sites), and on average 1730 ± 470 years after peatland inception. In 278
discontinuous permafrost, permafrost aggraded 1500 years later at 2000 ± 250 BP in sites that 279
were older (4320 ± 320 years). Permafrost aggraded more frequently in bog cores than fen cores 280
in the boreal region (42/62 cores) and discontinuous permafrost zones (40/70 cores), but more 281
frequently in fens than bogs in the tundra region (28/33 cores) and continuous permafrost zone 282
(20/28 cores).
283
3.2 North America 284
In North America, the number of peatland sites increased during the past 6000 years 285
(Figure 3). The majority of cores in North America were from Eastern and Central North 286
America (more than 50 cores each, Figure 3a, 3b), whereas Alaska and British Columbia 287
combined had the fewest records (28 sites, Figure 3d) followed by the Canadian Arctic (29 cores, 288
Figure 3c).
289
In Eastern North America, the number of fens peaked around 3125 BP (n= 21 cores) and 290
decreased to a minimum of less than 5% of cores in the present from a maximum of 65% at 6000 291
BP (Figure 3a). During this period, the percentage of bogs increased from 20% at 6000 BP (n=6 292
cores) to 50% in the present (n=26 cores). The first appearance of permafrost occurred between 293
5500 – 5250 BP (Payette, 1988) and increased to 43% of cores in the present day (n=22). Peak 294
periods of permafrost aggradation in Eastern North America occurred between 2250 – 2000 BP 295
and 1250 – 1000 BP (Figure 3a).
296
In Central North America, the fraction of fens decreased from 50% in 6000 BP to 20% in 297
present day, while the fraction of bogs reached a maximum between 3000 – 1750 BP. Bogs 298
occurred in ~50% of cores before decreasing to 37% of sites at 0 BP (Figure 3b). Permafrost 299
aggradation in Central North America was small prior to 750 BP, when rates increased from 0.2 300
± 0.1 percent of cores per century to 3.3 ± 0.8 percent of cores per century between 750-500 BP 301
(Figure 3b). The occurrence of permafrost in Central North America peaked at 0 BP (n=25, 42%, 302
Figure 3b). Some permafrost thaw also occurred in Central North America, with a maximum 303
occurring between 1500 BP and 500 BP in two to three cores (Figure 3b)(Kuhry, 2008; Zoltai, 304
1993).
305
In many sites in both the Canadian Arctic and Northern Alaska, permafrost aggradation 306
occurred in peats prior to 5000 BP (Figure 1, Figure 3c, 3d). In the Canadian Arctic, permafrost 307
formed in 30% of cores prior to 5000 BP (n=6). Most peatland initiation occurred in this region 308
before 5000 BP and from 4000 BP to 1250 BP. Permafrost aggradation increased between 4000 309
BP and 3000 BP and again from 1250 BP to 0 BP, resulting in 90% of cores developing 310
permafrost by 0 BP (n=26, Figure 3c).
311
In Alaska and British Columbia, the fraction of fens and bogs remained relatively 312
constant until 250 BP, when the number of fens decreased and the number of bogs and 313
permafrost cores increased (Figure 3d). Between 750 BP and present, permafrost thaw occurred 314
and reached a maximum of eight cores in present day (2015 CE; Figure 3d).
315
3.3 Eurasia 316
Trends in peatland transitions differed between North America and Eurasia and were 317
generally more stable in Eurasia (Figure 3, Figure 4). In Fennoscandia, fens were the most 318
common peatland type in cores at 6000 BP (73%), but were found in less than 50% of records 319
after 1250 BP as the number of bogs increased (Figure 4b). Permafrost aggradation in 320
Fennoscandian peatlands occurred after 1250 BP in very few sites (Figure 4b, Table S1)(Kokfelt 321
et al., 2010; Van der Knaap, 1989). In continental Europe, the proportion of fens and bogs were 322
relatively stable during the past 6000 years at 57% fen (n=7) and 36% bog, (n=6) respectively;
323
no permafrost formed in this region.
324
In the Russian and European Arctic, permafrost consistently aggraded in peatlands after 325
1000 BP and led to a sharp decrease in the number of permafrost-free fens (Figure 4a). The 326
occurrence of permafrost increased from 31% between 1000 – 750 BP (n= 5 cores) to 59% in 327
the present day (n=10, Figure 4a). Several cores in the Russian and European Arctic experienced 328
permafrost thaw and re-aggradation in the past 6000 years (Table S1; Oksanen et al., 2001;
329
Oksanen et al., 2003).
330
In West and Central Siberia, fraction of permafrost-free fens decreased slightly from 63%
331
to 44% of cores and the fraction of permafrost-free bogs increased from 11% to 28% between 332
6000 BP and present. Permafrost-free fens were the dominant peatland type in these records 333
between 6000 BP and 4750 BP and again 3750 BP and 750 BP (Figure 4c). From 4750 BP to 334
3750 BP and again between 750 BP to present, permafrost-free bogs and permafrost assemblages 335
increased (Figure 4c). Present-day permafrost extent was reached in West and Central Siberia 336
before 500 BP and occurred in 24% of cores (Figure 4c).
337
In East and Far East Siberia, permafrost-free fens were more common than permafrost- 338
free bogs, which were relatively rare in our records (Figure 4d). In Eastern and Far East Siberia, 339
permafrost aggradation accelerated after 1000 BP and peaked after 250 BP, occurring in 58% of 340
cores (n=14, Figure 4d). Many peatlands formed in thermokarst lake basins in this region, 341
resulting in a relatively high number of permafrost thaw records in these cores (Figure 4d, Table 342
S1; Walter Anthony et al., 2014).
343
4 Discussion 344
4.1 Regional trends in permafrost aggradation 345
One advantage to the reconstructions of permafrost history using peat cores is the broad 346
spatial coverage of this dataset (Figure 1), which allows global and regional comparisons to other 347
climate proxies, as well as information from relatively data-sparse regions, such as much of 348
Siberia (Mann et al., 2009; Marcott et al., 2013; Kaufman et al., 2009). The increased 349
aggradation of permafrost in peatlands over the past 6000 years is in general agreement with 350
other regional proxies of these northern regions showing a general cooling trend (Andreev et al., 351
2011; Kaufman et al., 2004; Kaufman et al., 2016; MacDonald et al., 2000; Marcott et al., 2013;
352
Fisher et al., 1995). Permafrost aggradation in peatlands increased since 6000 BP, with the most 353
widespread evidence for permafrost aggradation occurring in the mid- to late-Holocene after 354
4000 BP and more strongly after 1000 BP (Figure 2e), which generally followed the decrease in 355
summer solar insolation (Figure 2a) and decreasing melt recorded on the Greenland icesheet 356
(Figure 2b; Fischer et al., 1995). The increased rate of permafrost aggradation across the 357
northern high latitudes after 4000 BP suggests a climatic response to neoglacial cooling. Both 358
decreased Greenland melt layers (Figure 2b) and evidence for increased Arctic sea ice (Vare et 359
al., 2009) corroborate the trend of a cooling climate. Today, more than 40% of the peat cores 360
studied are underlain by permafrost. We observe regionally coherent patterns in permafrost 361
aggradation in both North America and Eurasia, potentially indicating climatic cooling such that 362
mean annual air temperatures dropped below 0º C during the time periods of high permafrost 363
aggradation.
364
Prior to neoglaciation (> 4000 BP) 365
Our data show that less than 10% of the northern high latitude peatlands studied 366
contained permafrost prior to 6000 BP (14 of 153 cores), owing to warmer-than-present 367
temperatures during the HTM (Kaufman et al., 2004; Fisher et al., 1995). This study shows early 368
(prior to 6000 BP) permafrost aggradation in peatlands of northern Alaska, similar to early 369
permafrost aggradation in Arctic Canada and parts of Siberia (Figure 1, Table S1). At 6000 BP, 370
scattered permafrost existed in peatlands within the continuous permafrost zone of Arctic Canada 371
but did not exist in peatlands in other parts of Canada (Zoltai, 1995), indicating warmer 372
temperatures than present-day in much of Canada (Kaufman et al., 2004).
373
Peat records in Alaska and British Colombia remain nearly unchanged throughout the 374
past 6000 years (Figure 3d), which suggests that these peatland ecosystems are either resilient to 375
environmental change, or that there has been relatively little environmental change over the last 376
6000 years that affects temperature or effective moisture, or both. The largest temperature 377
changes in Alaska and Yukon occurred prior to 6000 years ago, when cold deglacial 378
temperatures followed by warmer-than-present temperatures driven by insolation-driven changes 379
in the early Holocene (Kaufman et al., 2004). Summer temperature reconstructions from two 380
Alaskan lakes are stable for the past 6000 years (Kurek et al., 2009). Sea ice extent in the 381
Canadian Arctic showed only small increases from 6000-4000 BP and more abrupt increases 382
after 4000 BP (Vare et al., 2009), indicating stable conditions during the former period. A recent 383
study by Kaufman et al. (2016) showed high spatial variability in climate in Alaska and the 384
Yukon territories over the Holocene, likely driven by orographic complexity in the region, 385
masking trends in temperature and moisture beyond the identification of a mid-Holocene thermal 386
maximum from 7000-5000 BP and neoglacial cooling after 4000 BP. Alternatively, the 387
resiliency and stability of permafrost peatlands in the Alaska/Western Canada region (Figure 3d) 388
could be a reflection of the age of the peatlands and subsequent peat thickness (Table S1). These 389
peatlands largely developed in the early Holocene (Jones and Yu, 2010), several millennia before 390
the other peatland regions in this study (Gorham et al., 2007), and accumulated peat rapidly in 391
their initial stages (Jones et al., 2009; Jones et al., 2014).
392
In several Arctic and taiga regions, the peatland permafrost aggradation followed a 393
previous cycle of thaw of surface permafrost (yedoma) that formed on this unglaciated landscape, 394
forming thermokarst lakes, which later drained to create flat basins ideal for peatland 395
accumulation (Kanevskiy et al., 2014; Jorgenson et al., 2013; Jones et al., 2012). These processes 396
are common drivers of peatland initiation in Western Alaska, including the Seward Peninsula, 397
Koyukuk National Wildlife Refuge, and Innoko National Wildlife Refuge (Jones et al., 2012;
398
Jones et al., 2016b; O'Donnell et al., 2012; Kanevskiy et al., 2014) as well as in Eastern Siberia, 399
where peat also accumulated in drained lake basins (Walter Anthony et al., 2014; de Klerk et al., 400
2011). Therefore, Holocene permafrost aggradation in these regions were not unique, first-time 401
events, but instead part of a more complex evolution of deglacial climate and geomorphology.
402
The formation of peatlands in response to permafrost thaw occurred in several other sites, but the 403
initial cause of permafrost thaw is not described (Table S1).
404
In Siberia, permafrost aggradation occurs prior to 6000 BP in many sites, indicating 405
continued cold temperatures throughout much of the Holocene (Figure 1). Much like Alaska, 406
most of Siberia remained unglaciated and peat initiation was driven by warming and increased 407
moisture availability at the end of the last glacial maximum (Smith et al., 2004; MacDonald et al., 408
2006; Alexandrov et al., 2016). A relatively stable number of permafrost peatlands existed in 409
western and central Siberia, consistent with the stability of the unglaciated Alaska sites (Figure 410
4c, Figure 3d). These sites also have in common with Alaska an early HTM, followed by a 411
cooler and more stable mid- to late- Holocene climate (Kaufman et al., 2004). Permafrost also 412
aggraded prior to 5500 BP in other Arctic sites in eastern European Russia (Väliranta et al., 413
2003; Ronkainen et al., 2015; Hugelius et al., 2012).
414
Neoglaciation (4000 – 1500 BP) 415
The records of permafrost aggradation in peatlands (Figure 2e) show general agreement 416
with the climate trends of neoglacial cooling beginning 5000 - 3500 BP (Kaufman et al., 2004;
417
Payette and Lavoie, 1994; Alley, 2000; Fisher et al., 1995). The widespread presence of 418
permafrost in peatlands in both Eurasia and North America began around 4000 BP and 419
accelerated after 1000 BP (Figure 2- 4). Regionally, permafrost in peatlands steadily increased in 420
Arctic Canada, Eastern North America, and Arctic and European Russia after ~3000 BP (Figure 421
3a, 3c, 4a). Records of glacial expansions from both Alaska (Barclay et al., 2009) and Europe 422
(Holzhauser et al., 2005) during neoglaciation also show increased prevalence of glacial advance 423
after 3000 BP, with multiple glacier expansion events (3000 BP, 1500 BP, 250 BP; Barclay et al., 424
2009), roughly coinciding with the timing of increases in permafrost occurrence (Figure 2e). A 425
remarkable increase in sea ice in the Canadian Arctic Archipelago occurs after 3000 BP (Vare et 426
al., 2009; Belt et al., 2010), suggesting that insolation-driven neoglacial cooling increased 427
persistence of sea ice, which would have resulted in cooler and drier conditions on land. This is 428
consistent with the small number of new peat sites (one) in the Canadian Arctic between 3000 429
BP and 2250 BP (Table S1; Ellis and Rochefort, 2006) but is not reflected in the timing of the 430
aggradation of permafrost, which mainly aggraded either prior to 3000 BP or after 1000 BP 431
(Figure 1).
432
The regional trends in permafrost aggradation show spatial and temporal differences that 433
suggest regional differences in timing of post-HTM neoglacial cooling across North America.
434
Permafrost aggradation increased steadily after 3000 BP in eastern North America (Figure 3a), 435
which agrees with other records showing a steady cooling in eastern North America (Allard and 436
Seguin, 1987). The cooling, driven by decreasing summer insolation (Berger and Loutre, 1991), 437
shortened growing seasons and resulted in shifts in vegetation, as evidenced by pollen records 438
(Viau et al., 2006) and tree-line limits retreated in eastern Canada (Payette and Gagnon, 1985).
439
However, in Arctic Canada, permafrost occurrence increased earlier, beginning around 4000 BP 440
(Figure 3c; Zoltai, 1995) which coincides with the onset of cooler summer temperatures (Figure 441
2b; Fisher et al., 1995).
442
However, the most rapid rates of permafrost aggradation occurred even later in Central 443
North America than in Eastern North America or Arctic Canada, beginning after 1000 BP 444
(Figure 3b). The regional coherence of permafrost aggradation in Central North America (Figure 445
3b) suggests a strong climatic driver of permafrost aggradation that interacted with site-factors 446
such as peat thickness, Sphagnum presence, or snow cover to control whether permafrost 447
aggraded in individual sites. The contrast in the timing of permafrost aggradation between the 448
more continental climate of Central North America (later, Figure 3b) and more humid, maritime 449
Eastern North America (earlier, Figure 3a) suggests that, in addition to neoglacial cooling, other 450
climatic factors may have affected the timing of permafrost aggradation in Central North 451
America, such as changes in winter precipitation or snow cover.
452
Across Eurasia, the strongest trend in permafrost aggradation occurred in the Russian and 453
European Arctic after 3250 BP (Figure 4a). In the Laptev Sea region, Betula nana pollen 454
disappeared after 3300 BP, indicating that conditions became colder and more similar to today 455
after that time (Andreev et al., 2011), and tree lines retreated southward (MacDonald et al., 2000;
456
Oksanen et al., 2001). Permafrost aggradation in all Eurasian peat cores increased between 457
~3000-1000 BP, with the increase occurring earlier in Arctic Europe, European Russian, and 458
Russian Coastal Plain (Figure 4a)(Oksanen, 2006; Väliranta et al., 2003) and transgressing 459
eastward across central Siberia (Figure 4c) and eastern and far eastern Siberia (Figure 4d), where 460
permafrost increased markedly after 1000 BP. A transgressive west to east cooling is also 461
suggested by pollen-derived temperature reconstructions that show similar lags between western 462
and eastern Siberia (Anderson et al., 2002). Evidence for permafrost aggradation from peatland 463
sites in the eastern part of Eurasia that were not included in this analysis both agree and disagree 464
with this interpretation. Permafrost aggradation was indicated around 2550 BP at Vaisjeäggi 465
palsa mire in northern Finland (Oksanen, 2006) and around 3100 BP in one of two sub-Arctic 466
cores from Ortino in far northeastern European Russia (Väliranta et al., 2003). However, 467
permafrost aggraded in the other core from Ortino significantly earlier, around 5550 BP 468
(Väliranta et al., 2003), again suggesting that local factors also play a significant role in the 469
timing of permafrost aggradation.
470
Generally, Eurasian neoglacial cooling results in a decrease of permafrost-free fens in our 471
records and increase in permafrost in the cores after 2000 BP (Figure 4a, b, c), suggesting a 472
colder and potentially drier shifted fens to drier, ombotrophic bogs and permafrost peatlands.
473
Pollen and macrofossil data from a peatland in the Pur Taz region of Siberia indicate dry 474
conditions with abundant charcoal, and either little peat accumulation or oxidation of peat 475
following deposition after ~4500 BP, or both (Peteet et al., 1998).
476
Late Holocene (1500 – 750 BP) 477
Several Arctic regions showed an increase in permafrost aggradation beginning around 478
1500 BP. Other regions continued to aggrade permafrost (Eastern North America) or showed 479
little change in permafrost conditions. The large increase in permafrost presence after ~1500 BP 480
in eastern Siberia (Figure 4d) agrees with the timing of permafrost aggradation in Fennoscandia, 481
Arctic Europe, and the Russian Coastal Plain (Figure 4a,b,d) as well as Arctic Canada (Figure 482
3c), suggesting much colder temperatures in the Arctic at that time. Tree growth rings also 483
indicated cool conditions in this region during this period (Naurzbaev and Vaganov, 2000) and 484
reconstructed winter sea surface temperatures also decline after 1500 BP (de Vernal et al., 2013;
485
Voronina et al., 2001).
486
From 1250 - 1000 BP, the number of cores with fen vegetation increased in Central North 487
America (Figure 3b), as does the number of cores with thawed permafrost peat (Figure 3b). This 488
pattern suggests slightly warmer temperatures, which coincides with the Roman Warm Period in 489
North America (Viau et al., 2006), or that permafrost temperatures remained close to 0ºC and 490
may have increased above the freezing point interannually, increasing their susceptibility to thaw.
491
Medieval Climate Anomaly and Little Ice Age (after 1000 BP) 492
The peak in permafrost aggradation observed in this study occurs between 750-0 BP 493
(Figure 2e), the beginning of which broadly corresponds with the end of the Medieval Climate 494
Anomaly (MCA) and encompasses the LIA (Mann et al., 2009). This trend was spatially 495
widespread and occurred throughout the study region after 750 BP (Figure 1, 3, 4)(Oksanen, 496
2006; Kultti et al., 2004), with the exception of Western and Central Siberia, where the rate of 497
permafrost aggradation increased earlier (Figure 1, 4c). Our data show that the maximum 498
modern permafrost extent was reached ~125 BP (Figure 2e), corresponding with the end of the 499
Little Ice Age, which slightly lags the coolest LIA temperatures that occurred ~200 years BP 500
(Kaufman et al., 2004; Marcott et al., 2013).
501
In North America, the pattern of greater permafrost aggradation earlier in Western North 502
America (Arctic Canada and Alaska, British Columbia) (Figure 1, 3c, 3d) than in Eastern North 503
America (Figure 1, 3a, 3b) agrees broadly with the spatial patterns of cool and warm anomalies 504
(relative to 1961-1990 CE) during the MCA. During the MCA, warm anomalies persisted over 505
the North Atlantic and eastern Canada and cool anomalies persisted over western Canada and 506
Alaska (Mann et al., 2009). This likely resulted in climatic conditions that were not favorable to 507
permafrost aggradation in Eastern North America, and thus, permafrost aggradation commenced 508
when conditions cooled after 1000 BP.
509
At the end of the MCA, permafrost aggradation increased sharply in Central North 510
America after 750 BP at rate of more than 3% of cores per century, especially compared with 511
previous rate in the Holocene of 0.2% of cores per century. The increased permafrost 512
aggradation in Central North America, especially in the present-day discontinuous and sporadic 513
permafrost zone of Manitoba and Alberta (Figure 1, 3b), culminated in the greatest number of 514
permafrost sites after 250 BP (Figure 3b). This suggests that MAAT decreased below 0° C in this 515
region, and agrees with previous studies that found much of the permafrost in the discontinuous 516
permafrost zone of Canada formed during the Little Ice Age (Halsey et al., 1995; Zoltai, 1995).
517
Eastern North America was the only region experiencing an increasing fraction of 518
permafrost-free bogs relative to other peatland classes during the LIA as indicated by the 519
presence of ombrotrophic bog vegetation. This suggests either autogenic shifts toward 520
ombrotrophy as aging peatlands accumulated peat above the water table or drier conditions 521
resulting in a decrease in water level, or both, that would have resulted in ombrotrophic 522
conditions and vegetation changes (Figure 3a)(e.g. Magnan and Garneau, 2014). A recent late 523
Holocene hydroclimate synthesis of multi-proxy records found drying in many records from 524
Quebec and Nova Scotia during this time (Rodysill et al., 2017). Using lake level 525
reconstructions, Payette and Delwaide (2004) found increased tree mortality at the onset of the 526
LIA (350 BP to 200 BP) in Northern Quebec due to a combination of peatland permafrost 527
aggradation and a drier climate. Taken together, peatland succession and permafrost aggradation 528
does not appear to be the result of autogenic peatland processes alone, but due in part to climatic 529
drivers, including both temperature and moisture.
530
4.2 Permafrost thaw 531
Unlike permafrost aggradation, permafrost thaw is rarely recorded in the peat cores 532
studied despite warmer late Holocene climatic periods associated with the Roman Warm Period 533
and the Medieval Climate Anomaly. The highest rates of permafrost thaw in this study occurred 534
as a result of both widespread permafrost extent in peatlands and warmer conditions since the 535
end of the LIA (Figure 2b, 2f). Given the relatively widespread occurrence of permafrost in 536
peatlands today (Figure 2e) combined with warming temperatures, it seems likely that the 537
thawing trend will continue, especially in regions where permafrost has thawed before.
538
Understanding the resilience of permafrost under a warming climate is important given the 539
potential for losses of peatland carbon stocks with permafrost thaw is significant (Jones et al., 540
2016b; O'Donnell et al., 2012). The long-term response of peatland C stocks to permafrost thaw 541
is important for predicting carbon cycle feedbacks to future warming.
542
Permafrost thaw requires that permafrost was established at the site previously. This 543
study examines permafrost aggradation since 6000 BP, as permafrost was only present in a few 544
cores prior to this time (Figure 2-4). Many peatlands formed in formerly glaciated regions 545
(Gorham et al., 2007), where in lieu of permafrost, ice sheets covered much of the present-day 546
permafrost peatland region (Vandenberghe et al., 2014). Rapid warming following the LGM was 547
likely unfavorable to permafrost aggradation, as HTM occurred early in the Holocene (Kaufman 548
et al., 2004; Fischer et al., 1995). However, where permafrost was present prior to 6000 BP, 549
evidence for permafrost thaw prior to peat formation (6000-5000 BP) occurred in both Arctic 550
Canada, Alaska, and Eastern Siberia (Table S1, Figure 4d), as peatlands developed in drained 551
thaw lake basins following permafrost thaw (Walter Anthony et al., 2014; Geurts, 1985; Jones et 552
al., 2012; de Klerk et al., 2011). While permafrost aggradation generally increases over the 6000 553
year period (Figure 2e) in agreement with decreasing solar insolation (Figure 2a) and evidence 554
for lower Greenland melt (Figure 2b), greater thaw is recorded western and central Siberia, 4500- 555
2000 BP (Figure 4c; n=1), and later still in Arctic Europe and European Russia (1750 – 0 BP;
556
n=3). Sea ice records from the western Arctic Ocean suggest millennial-scale sea ice minima 557
occurred at 6000, 4000, and 2000 BP (de Vernal et al., 2005), which could have resulted in a 558
warmer and/or wetter climate on land, as land-locked sea ice shifts moisture from ice-free oceans 559
farther off shore and vice versa. In regions such as Central North America, Alaska and British 560
Columbia, and to a greater extent in Arctic Europe, permafrost thaw accompanies an increase in 561
the prevalence of permafrost, suggesting that the recently aggraded permafrost was relatively 562
warm (~0ºC) and therefore more susceptible to thaw (Figure 3b, 3d, 4a).
563
Regardless, evidence for permafrost thaw is relatively rare in most plant macrofossil 564
records (10 of 266 records), especially compared with evidence of permafrost aggradation. It is 565
possible that permafrost thaw is uncommon, or this could result from a sampling bias against 566
thaw features because of the difficulties associated with sampling wet sites. Lack of evidence of 567
permafrost thaw could also result from the rapid loss of peat following decomposition in thawed 568
areas (e.g. O'Donnell et al., 2012; Jones et al., 2016b) that makes distinguishing the timing of 569
permafrost thaw very difficult (Fillion et al., 2014; Hunt et al., 2013). Notably, the repeated 570
cycles of permafrost aggradation and degradation in peatlands that were observed in 571
discontinuous permafrost regions of West-Central Canada and European Russia (Sannel and 572
Kuhry, 2008; Sannel and Kuhry, 2009; Oksanen et al., 2003; Zoltai, 1993; Oksanen et al., 2001) 573
were relatively rare outside of those regions during the past 6000 years, with the exception of 574
peat formation in drained thermokarst lake basins (Jones et al., 2012; Walter Anthony et al., 575
2014).
576
4.3 Regional, local, and site-level drivers of permafrost aggradation and thaw and broader 577
implications 578
Regional climatic factors that affect the distribution of permafrost in the present day, such 579
as the influence of maritime climate and the Gulf Stream, were also likely responsible for 580
regional differences in permafrost history. For example, the absence of permafrost in peatlands 581
of the UK or Europe can be attributed to the maritime climate (Peel et al., 2007) despite 582
occurring at similar latitudes in both North America and Asia (Figure 1). Similarly, there was 583
generally no evidence of permafrost aggradation in coastal sites within Alaska and British 584
Columbia, whereas permafrost aggradation occurred more frequently in sites located further 585
inland, away from the maritime influence (Figure 1; Peel et al., 2007). However, recent 586
discovery of rapidly degrading permafrost on the Kenai Peninsula in Alaska indicates that 587
isolated or sporadic patches of permafrost have persisted from the Holocene in coastal regions in 588
Alaska that experience semi-continental climates, but more research is needed to determine the 589
age of permafrost on the Kenai Peninsula (Jones et al., 2016a). Furthermore, differences in 590
humidity and continentality may be partially responsible for differences in the timing of 591
permafrost among regions such as Central and Eastern North America.
592
Interactions between vegetation and microclimate can result in permafrost aggradation 593
and maintain permafrost in regions where warmer air temperatures would otherwise degrade 594
permafrost (Halsey et al., 1995; Shur and Jorgenson, 2007). Sphagnum mosses are frequently 595
cited as a driver in permafrost aggradation in peatland sites due to their low thermal conductivity 596
when dry (Kujala et al., 2008; O'Donnell et al., 2009). In this study, Sphagnum mosses occurred 597
more frequently prior to permafrost aggradation (28 sites) than the presence of woody peat (12 598
sites) or tree and shrub macrofossils (11 sites). This suggests that permafrost aggradation may 599
result from shading, snow cover, or snow redistribution resulting from the presence of trees and 600
shrubs less often than some insulating property arising from the presence of Sphagnum.
601
Many sites in the isolated, sporadic, and discontinuous permafrost zones are likely to 602
experience permafrost thaw and thermokarst in the next century (Zhang et al., 2008; Olefeldt et 603
al., 2016). Within a site, the timing of permafrost aggradation relative to peat deposition has 604
important implications for the response of peatland C stocks to permafrost thaw. In sites where 605
permafrost aggraded concurrently with peat deposition (syngenetic permafrost), organic matter 606
in permafrost is less decomposed and can be lost from the soil relatively rapidly following 607
permafrost thaw (O'Donnell et al., 2012; Treat et al., 2014; Biasi et al., 2014; Sannel and Kuhry, 608
2009), whereas organic matter in permafrost that formed significantly after peat deposition 609
(epigenetic permafrost) can be more decomposed and therefore less likely to release large 610
amounts of carbon (Lee et al., 2012; Diáková et al., 2016; Treat et al., 2014). The warmer 611
permafrost sites in the discontinuous zone often have permafrost that formed recently (in the past 612
1000 years; Figure 1) when peatlands were older and is likely epigenetic in origin. Therefore, 613
these sites in the discontinuous zone may have limited carbon losses due the prevalence of highly 614
decomposed material assimilated into permafrost. Permafrost aggraded before 1000 BP in 615
southern parts of the West Siberian lowlands, the Nenets Republic, and Southern Siberia of 616
Russia, as it did in sites in far Northern Quebec, the Seward Peninsula, and Koyukuk National 617
Wildlife Refuge in Alaska (Figure 1, Table S1). The location of these sites that contain older 618
permafrost within the discontinuous permafrost zones suggests a relatively high vulnerability to 619
permafrost thaw due to warmer present-day conditions and the potential for the greater loss of 620
less decomposed carbon from the peat in permafrost. These regions might be a priority for future 621
research focused on detecting changes in C stocks and biogeochemical cycling due to permafrost 622
thaw.
623
5 Conclusions 624
This dataset illustrates the utility of reconstructions of permafrost history in the pan-Arctic using 625
records of permafrost aggradation and thaw inferred from peat cores. The broad and regional 626
agreement between the permafrost aggradation history and other climatic proxies show that 627
peatland records of permafrost history can be useful for understanding regional temperature 628
trends (or continued cold temperatures) for the last 6000 years. Because of local, site-level 629
factors, it is crucial to consider multiple sites when reconstructing permafrost history in order to 630
elucidate a regional climatic signal. The regional synthesis shows three periods of increased 631
permafrost aggradation in peatlands: after 3000 BP, around 1500 BP, and 250 BP, aligning 632
roughly with neoglaciation and the LIA, suggesting periods where temperatures within the 633
northern permafrost zone decreased significantly. Periods of increased permafrost aggradation in 634
peatlands broadly agree with sea ice records and glacial expansion in the late Holocene.
635
Permafrost history is important not only as an additional proxy of Holocene climate, but also for 636
determining the magnitude of potential carbon loss from peatlands with permafrost thaw. A 637
combination of regional permafrost history with predictions for permafrost persistence is key to 638
predicting soil carbon loss with permafrost thaw and identifying key regions for monitoring of 639
changes in carbon emissions at a landscape scale.
640
Acknowledgements 641
We thank two anonymous reviewers for thorough and constructive comments. CT was supported 642
by the P2C2 Program of the National Science Foundation granted to MJ (ARC-1304823), the 643
Max Planck Institute for Meteorology, and The Academy of Finland CAPTURE Project. MJ is 644
funded by the USGS Climate and Land-use Change Research and Development Program. Any 645
use of trade, product, or firm names is for descriptive purposes only and does not imply 646
endorsement by the U.S. government.
647 648
References 649
Alexandrov GA, Brovkin VA and Kleinen T. (2016) The influence of climate on peatland 650
extent in Western Siberia since the Last Glacial Maximum. Scientific Reports 6:
651
24784.
652
Allard M and Seguin MK. (1987) The Holocene evolution of permafrost near the tree line, 653
on the eastern coast of Hudson Bay (northern Quebec). Canadian Journal of Earth 654
Sciences 24: 2206-2222.
655
Alley RB. (2000) The Younger Dryas cold interval as viewed from central Greenland.
656
Quaternary Science Reviews 19: 213-226.
657
Anderson PM, Lozhkin AV and Brubaker LB. (2002) Implications of a 24,000-Yr 658
Palynological Record for a Younger Dryas Cooling and for Boreal Forest 659
Development in Northeastern Siberia. Quaternary Research 57: 325-333.
660
Andreev AA, Schirrmeister L, Tarasov PE, et al. (2011) Vegetation and climate history in the 661
Laptev Sea region (Arctic Siberia) during Late Quaternary inferred from pollen 662
records. Quaternary Science Reviews 30: 2182-2199.
663
Barclay DJ, Wiles GC and Calkin PE. (2009) Holocene glacier fluctuations in Alaska.
664
Quaternary Science Reviews 28: 2034-2048.
665
Belt ST, Vare LL, Massé G, et al. (2010) Striking similarities in temporal changes to spring 666 sea ice occurrence across the central Canadian Arctic Archipelago over the last 7000 667
years. Quaternary Science Reviews 29: 3489-3504.
668
Berger A and Loutre MF. (1991) Insolation values for the climate of the last 10 million 669
years. Quaternary Science Reviews 10: 297-317.
670
Biasi C, Jokinen S, Marushchak M, et al. (2014) Microbial Respiration in Arctic Upland and 671
Peat Soils as a Source of Atmospheric Carbon Dioxide. Ecosystems 17: 112-126.
672
Blaauw M and Christen JA. (2011) Flexible paleoclimate age-depth models using an 673
autoregressive gamma process. Bayesian Analysis 6: 457-474.
674
Bockheim JG, Hinkel KM, Eisner WR, et al. (2004) Carbon pools and accumulation rates in 675
an age-series of soils in drained thaw-lake basins, Arctic Alaska. Soil Science Society 676 of America Journal 68: 697-704.
677
Brown J, O.J. Ferrians Jr., J.A. Heginbottom, et al. (1998, revised 2001) Circum-Arctic map of 678
permafrost and ground-ice conditions. Boulder, CO: National Snow and Ice Data 679
Center/World Data Center for Glaciology, Digital Media.
680
Camill P. (2000) How much do local factors matter for predicting transient ecosystem 681
dynamics? Suggestions from permafrost formation in boreal peatlands. Global 682
Change Biology 6: 169-182.
683
Camill P. (2005) Permafrost thaw accelerates in boreal peatlands during late-20th century 684
climate warming. Climatic Change 68: 135-152.
685
Camill P, Barry A, Williams E, et al. (2009) Climate-vegetation-fire interactions and their 686
impact on long-term carbon dynamics in a boreal peatland landscape in northern 687
Manitoba, Canada. J. Geophys. Res. 114.
688
Camill P and Clark JS. (1998) Climate change disequilibrium of boreal permafrost peatlands 689
caused by local processes. American Naturalist 151: 207-222.
690 Charman DJ, Beilman DW, Blaauw M, et al. (2013) Climate-related changes in peatland 691
carbon accumulation during the last millennium. Biogeosciences 10: 929-944.
692
Davis N. (2001) Permafrost: A guide to frozen ground in transition, Fairbanks, AK:
693
University of Alaska Press.
694
de Klerk P, Donner N, Karpov NS, et al. (2011) Short-term dynamics of a low-centred ice- 695
wedge polygon near Chokurdakh (NE Yakutia, NE Siberia) and climate change 696
during the last ca 1250 years. Quaternary Science Reviews 30: 3013-3031.
697
de Vernal A, Hillaire-Marcel C and Darby DA. (2005) Variability of sea ice cover in the 698
Chukchi Sea (western Arctic Ocean) during the Holocene. Paleoceanography 20:
699
PA4018.
700
de Vernal A, Hillaire-Marcel C, Rochon A, et al. (2013) Dinocyst-based reconstructions of 701
sea ice cover concentration during the Holocene in the Arctic Ocean, the northern 702
North Atlantic Ocean and its adjacent seas. Quaternary Science Reviews 79: 111-121.
703
Diáková K, Čapek P, Kohoutová I, et al. (2016) Heterogeneity of carbon loss and its 704
temperature sensitivity in East-European subarctic tundra soils. Fems Microbiology 705
Ecology 92.
706
Ellis CJ and Rochefort L. (2006) Long-term sensitivity of a High Arctic wetland to Holocene 707
climate change. Journal of Ecology 94: 441-454.
708
Fillion M-E, Bhiry N and Touazi M. (2014) Differential Development of Two Palsa Fields in a 709
Peatland Located near Whapmagoostui-Kuujjuarapik, Northern Quebec, Canada.
710
Arctic Antarctic and Alpine Research 46: 40-54.
711
Fisher DA, Koerner RM and Reeh N. (1995) Holocene climatic records from Agassiz ice cap, 712
Ellesmere Island, NWT, Canada. The Holocene 5: 19-24.
713
Geurts M-A. (1985) Le paysage végétal Holocène dans la région d’Escape Rapids, 714
Territoires du nord-ouest. Geographie Physique Et Quaternaire 39: 215-220.
715
Gorham E, Lehman C, Dyke A, et al. (2007) Temporal and spatial aspects of peatland 716
initiation following deglaciation in North America. Quaternary Science Reviews 26:
717
300-311.
718
Group NWW. (1988) Wetlands of Canada. Ecological land classification series, no. 24.
719 Sustainable Development Branch, Environment Canada, Ottawa, Ontario, and 720
Polyscience Publications Inc., Montreal, Quebec 452.
721
Halsey LA, Vitt DH and Zoltai SC. (1995) Disequilibrium Response of Permafrost in Boreal 722
Continental Western Canada to Climate-Change. Climatic Change 30: 57-73.
723
Hodgkins SB, Tfaily MM, McCalley CK, et al. (2014) Changes in peat chemistry associated 724
with permafrost thaw increase greenhouse gas production. Proceedings of the 725
National Academy of Sciences 111: 5819-5824.
726
Holzhauser H, Magny M and Zumbuühl HJ. (2005) Glacier and lake-level variations in west- 727
central Europe over the last 3500 years. The Holocene 15: 789-801.
728