Peatlands Are a Key Component of the Global Carbon Cycle

Northern peatland initiation lagged abrupt increases in deglacial atmospheric CH4

Alberto V. Reyes1,2,3 and Colin A. Cooke1,2,4

Department of Earth and Atmospheric Sciences, University of Alberta, Edmonton, AB, Canada T6G 2E3

Edited by James P. Kennett, University of California, Santa Barbara, CA 93106, and approved January 14, 2011 (received for review September 13, 2010)

14 Peatlands are a key component of the global carbon cycle. Chron- Compilations of peatland basal C dates make the implicit ologies of peatland initiation are typically based on compiled basal assumption that the compiled dates faithfully reflect the timing peat radiocarbon (14C) dates and frequency histograms of binned of peatland initiation, but peat can be difficult to date reliably 14 calibrated age ranges. However, such compilations are problematic (e.g., 14) and many C dates used in such compilations are because poor quality 14C dates are commonly included and because decades old and have large analytical uncertainty terms. More frequency histograms of binned age ranges introduce chronologi- critically, radiocarbon dates must be calibrated to calendar years 14 cal artefacts that bias the record of peatland initiation. Using a because production of C, and its relative distribution in the published compilation of 274 basal 14C dates from Alaska as a case oceans, atmosphere, and terrestrial biosphere, is not constant study, we show that nearly half the 14C dates are inappropriate through time (Fig. 2A and ref. 15). The calibration process yields for reconstructing peatland initiation, and that the temporal potentially large, non-Gaussian, distributions of likely calendar 14 structure of peatland initiation is sensitive to sampling biases and ages that present interpretive challenges for C-dated compila- 14 tions of peatland initiation. treatment of calibrated C dates. We present revised chronologies 14 of peatland initiation for Alaska and the circumpolar Arctic based Here, we use the recent compilation of basal peat C dates 14 from Alaska (9) as a case study to examine how different treat- on summed probability distributions of calibrated C dates. These 14 revised chronologies reveal that northern peatland initiation ment of C dates affects interpretation of temporal trends for peatland initiation, before demonstrating the broad applicability lagged abrupt increases in atmospheric CH4 concentration at the 14 start of the Bølling–Allerød interstadial (Termination 1A) and the of our results for global peatland C datasets and associated C end of the Younger Dryas chronozone (Termination 1B), suggesting cycle implications. Our revised chronologies reveal that northern that northern peatlands were not the primary drivers of the rapid peatland initiation lagged abrupt deglacial and early Holocene increases in AMC, and thus could not have been the principal increases in atmospheric CH4. Our results demonstrate that subtle methodological changes in the synthesis of basal 14C ages lead to driver for the AMC increases. substantially different interpretations of temporal trends in peat- Results and Discussion land initiation, with direct implications for the role of peatlands in Of the 274 14C dates presented by Jones and Yu (9) (hereafter the global carbon cycle. 10 termed JY all) we rejected 115 dates (Fig. 1, Dataset S1, and SI Text) based on poor suitability of the dated material, lack radiocarbon dating ∣ peatland carbon ∣ ice core methane ∣ paleoclimate of supporting stratigraphic context, poorly constrained stratigra- phy in the original source, or because the most parsimonious eatlands play an important role in millennial-scale climate stratigraphic interpretation does not involve peatland initiation. Pchange and the carbon (C) cycle because they are important For example, we deemed inappropriate the inclusion of 14C dates C sinks (1, 2) as well as a substantial methane (CH4) source (3, 4). on peat interbedded within fluvial sediments or thin peat strin- Atmospheric CH4 concentrations (AMC) rose quickly during gers overlying till (e.g., I-10469 and CAMS-41410 in Dataset S1); deglaciation and following the Younger Dryas chronozone, but such dates record, at best, the timing of geomorphic processes then declined steadily until ∼5 ka (ka ¼ 103 cal yr BP; cal yr BP ¼ or glacial recession, and are unrelated to peatland initiation. 14 10 14 calendar years before A.D. 1950) before rising again during the The vetted compilation of C dates (JY vet) contains 159 C late Holocene (5). However, controversy surrounds the temporal dates. Of these, 70 were sampled from sites within last glacial pattern of peatland initiation and expansion, the role of peatlands maximum (LGM) ice limits and only 76 were sampled from areas in deglacial and early Holocene atmospheric CH4 fluctuations, with at least 5% areal coverage by peat (Fig. 1 and Table S1). and what, if any, role peatlands played in the middle Holocene The histogram method (HIST; see Methods), employed in 14 reversal of AMC (6–9). most previous compilations of basal peat C dates (6, 7, 9, 10, Compilations of radiocarbon (14C) dates on basal peat deposits 12), uses the summed number of calibrated age ranges within provide the foundation for efforts aimed at assessing the long- a given year, or range of years, as a proxy time series for peatland term role of northern peatlands in the climate system and the initiation; cumulative initiation curves are then used to infer C cycle (6–12). In a landmark study, MacDonald et al. (7) com- changes in total peatland area over time (e.g., 9, 16). Using piled 1,516 14C dates on basal peat from across middle- and high-

latitude Asia, Europe, and North America. They found that Author contributions: A.V.R. and C.A.C. designed research, performed research, analyzed boreal and northern peatlands expanded rapidly between 12 data, and wrote the paper. and 8 ka, and proposed a direct link between northern peatland The authors declare no conflict of interest. expansion and the early Holocene rise in AMC. This article is a PNAS Direct Submission. Using similar compilations of 14C dates, others have noted 1A.V.R. and C.A.C. contributed equally to this work. the relation between deglaciation and peatland initiation (13), 2To whom correspondence may be addressed. E-mail: [email protected] or colin.cooke@ and the implications of lateral peatland expansion for late Holo- yale.edu. cene terrestrial C cycling (8). Most recently, Jones and Yu (9) 3Present address: Department of Geoscience, University of Wisconsin, Madison, WI 53706. compiled new and previously published basal 14C dates from 4Present address: Department of Geology and Geophysics, Yale University, New Haven, Alaskan peatlands (Fig. 1), and suggested that they were an CT 06520-8109. important CH4 source during the initial abrupt rise in AMC at This article contains supporting information online at www.pnas.org/lookup/suppl/ Termination 1B. doi:10.1073/pnas.1013270108/-/DCSupplemental.

4748–4753 ∣ PNAS ∣ March 22, 2011 ∣ vol. 108 ∣ no. 12 www.pnas.org/cgi/doi/10.1073/pnas.1013270108 Downloaded by guest on September 23, 2021 Fig. 1. Map of Alaska showing locations of sampling sites for basal peat 14C dates in Alaska (9). Last glacial maximum ice limits (17) show that large areas of interior Alaska remained un- covered by ice. Organic-rich soils of the Histosol order and Histel suborder (18) are mapped as an approxima- tion of peatland extent in Alaska. Shaded relief bathy- metry and elevation maps from ETOPO2 and the Alas- ka Geospatial Data Clearing- house, respectively.

10 10 14 HIST,both JY all and JY vet exhibit similar temporal structure, initiation events, of different ages, based on one C date. Sec- with a pronounced mode of peatland initiation between ∼10 ond, by weighing equally the 50-yr bins at the extremes of the and 12.5 ka (Fig. 2B). The magnitude of this mode is muted calibrated age range, the HIST method assumes that the tails of 10 in JY vet, though cumulative peatland development for both the probability density function for a given calibrated age range 10 10 JY all and JY vet is similar. Both datasets show peatland are just as likely to represent the true calendar age of the basal initiation commencing ∼18–16 ka and increasing steadily until peat sample as the mode or modes of the probability density ∼11 ka, with short-lived drops in peatland initiation ∼11.2 ka. function (e.g., Fig. S2). Finally, because large calibrated age The rate of peatland initiation slows precipitously in both records ranges generate more bins (i.e., more peatland initiation events)

after ∼11–10 ka. By 9 ka, the cumulative curves associated than small calibrated age ranges, the HIST method biases the GEOLOGY with the HIST method for both datasets show almost 70% of 14C record of peatland initiation toward the late Pleistocene present peatland area was established, and only ∼15% of Alaska and early Holocene. This is because older 14C dates tend to peatland area formed after 5 ka. have larger associated laboratory uncertainties (Table S1 and The HIST method generates n ¼ r · b−1 peatland initiation Fig. S3), in turn yielding larger calibrated age ranges and more events for each calibrated age range of an individual radiocarbon numerous bin entries, particularly for calibrated dates affected date, where r and b are the length in years of the calibrated age by 14C plateaus. range and the histogram bins, respectively; thus, the number of As an alternative to the HIST method of summarizing the peatland initiation events generated by the HIST method is basal peat 14C dataset for Alaska, we calculated the sum of the SCIENCES much higher than the number of basal peat 14C dates used in the probability density functions associated with each calibrated age ENVIRONMENTAL 10 analysis (e.g., Fig. S1). For example, JY all yields 5099 peatland range (hereafter termed the PROB method). In this approach, initiation events (i.e., 50-yr calibrated age range bins) for only the shape of the summed probability curve represents relative 14 10 274 basal C dates. Similarly, using the HIST method, JY vet changes in the frequency of peatland initiation (see Methods). gives 3,165 peatland initiation events for 159 basal peat 14C dates. This alternative reconstruction of peatland initiation in Alaska We suggest that the HIST method for generating proxy records is strikingly different from the HIST reconstruction (Fig. 2C). of peatland initiation from large compilations of 14C dates is Peatland initiation does not begin in earnest until ∼14 ka. There flawed. First, by counting multiple bins within a single calibrated is a pronounced (∼1;000 yr) pause in the rate of peatland initiation age range, the HIST method effectively creates multiple peatland during the Younger Dryas, when the HIST record shows only a

Reyes and Cooke PNAS ∣ March 22, 2011 ∣ vol. 108 ∣ no. 12 ∣ 4749 Downloaded by guest on September 23, 2021 short pause, followed by a rapid increase, in the expansion of peatlands. Both records exhibit a sharp decrease in peatland initiation after ∼10 ka, but the PROB record reveals substantial late Holocene peatland initiation after 5 ka that is not present in the HIST record (Fig. 2C). In terms of cumulative peatland devel- 10 opment, the PROB method for JY vet shows 55% of Alaska peat- ∼25% A lands were developed by 9 ka, and initiated after 5 ka. Limitations of the Summed Probability Approach. Although many B other studies have employed summed calibrated probability distributions to summarize 14C date compilations (e.g., 19–21), we are mindful that there are limitations and pitfalls to this approach (22–24). For example, 14C dates with large laboratory uncertainty terms typically result in wide calibrated age ranges that can blur peaks and troughs in summed probability curves. Indeed, more 14 10 10 than half the C dates in JY vet and JY all have laboratory uncertainties >100 yr (Table S1). However, blurred probability distributions are a more critical issue for identifying short-duration events than multicentennial or millennial-scale temporal trends (e.g., 19). In addition, summed probability distributions C must be interpreted with caution, particularly when the number of 14C dates in a particular time interval is low. Differences in the slope of the 14C calibration curve can also alter relative peak heights within summed calibrated probability distributions. However, the direct stratigraphic association between basal 14C dates and peatland initiation simplifies our analysis, particularly in this carefully vetted dataset. Moreover, the use of 100-yr means (or larger) is thought to minimize chronological artefacts caused by fluctuations in the 14C calibration curve (25). We emphasize that there is much room for improvement in the quality of 14C dates in large basal peat compilations, irrespec- D tive of the method used to infer trends in peatland initiation. For example, only a minority of 14C dates from Alaska were measured by accelerator mass spectrometry (Table S1), which provides high-precision age determinations and, importantly, allows dating of discrete terrestrial plant macrofossils. Numerous 14C dates in JY10, and other compilations, were obtained from bulk peat samples that may be contaminated with older C due to hard-water effects (26), the incorporation of finely comminu- ted Cenozoic coals (27), or potential assimilation of recycled 14 CH4 in submerged Sphagnum (28). Contamination of bulk peat through penetration by rootlets or humic acids (e.g., 14) is also E of particular concern because appropriate pretreatments are rarely documented in the original published sources (Table S1). Finally, we acknowledge that basal peat 14C dates often reflect minimum ages for peatland initiation (i.e., the true initiation date may be older than suggested by a given 14C age), though this is balanced to some extent by some of the contamination issues discussed above.

Peatland Initiation, Deglaciation, and Sampling Bias. While peatlands within glacial limits are suitable for evaluating the long-term role of peatlands in the C cycle, they may not be ideal targets when attempting to identify extrinsic controls on peatland initiation (9). Fig. 2. Comparison of latest Pleistocene/early Holocene radiocarbon-based This is because it may be difficult to disentangle climatic and/or trends of peatland initiation in Alaska and the circum-Arctic, and thermo- hydrological drivers of peatland initiation from deglacial land- karst lakes initiation. Rapid changes in climate, including Terminations 1A scape succession (13). There also exists a sampling bias toward 14 (T1A) and 1B (T1B), the Younger Dryas (YD) and Bølling–Allerød (B–A), are sites within ice limits because many basal peat C dates were also indicated. (A) IntCal09 14C calibration and Δ14C curves; periods of steep compiled from early studies that were focused on establishing declines in Δ14C highlight so-called 14C “plateaus.” (B) Frequency and cumu- chronologies for regional deglaciation (Fig. 1 and Dataset S1). lative percent of Alaska peatland initiation events based on the HIST method This sampling bias is particularly severe in Alaska, where the 10 10 applied to JY all and JY vet (see Methods). (C) Relative frequency and majority of mapped peatlands are located outside LGM ice limits, cumulative percent of Alaska peatland initiation events based on PROB 14 10 but about half of the sampling sites for basal peat C dates are method applied to JY vet.(D) Normalized HIST and PROB reconstructions of peatland initiation based on the MacD06 dataset (7), and cumulative within LGM ice limits (Fig. 1 and Table S1). peatland initiation derived from the nonnormalized records. (E) Normalized HIST and PROB reconstruction methods of thermokarst lake initiation Reassessing the Initiation Chronologies of Circum-Arctic Peatlands events applied to Walter et al. (29). Note that the HIST method in ref. 29 uses and Thermokarst Lakes. Our results from Alaska reveal differences midpoints of the calibrated age range (see Methods). in the trajectory of peatland initiation as determined by the

4750 ∣ www.pnas.org/cgi/doi/10.1073/pnas.1013270108 Reyes and Cooke Downloaded by guest on September 23, 2021 PROB and HIST methods. While vetting of the JY10 14C database changed the relative magnitude of peaks and troughs in the HIST-based record of peatland initiation, it did not demon- A strably alter its temporal structure. In contrast, the PROB-based record of Alaska peatland initiation differs in key respects from the HISTrecord. Toassess the broader implications of our results, we applied the PROB method to the MacDonald et al. (7) circum-Arctic compilation of 1,516 basal peat 14C dates (hereafter: MacD06), which was originally reported using the HIST method with 1-yr bins of 1σ calibrated age ranges, and to the B Walter et al. (29) compilation of 69 basal 14C dates from thermokarst lakes, which was originally reported using a modified HIST method (see Methods). As is the case with the Alaska dataset, the PROB method yields important differences in the timing and cumulative pace of circum-Arctic peatland initiation (Fig. 2D). The HIST reconstruction overestimates circum-Arctic peatland initiation between ∼13 and 9 ka, and there are millennial-length periods of underestimation after 8 ka (Fig. 2D). As observed in Alaska, circum-Arctic peatland initiation increases nearly monotonically C through the Younger Dryas in the HIST reconstructions, but the PROB method shows that peatland initiation slackened dur- D ing that interval. Interpretation of trends in peatland initiation also diverges during the late Holocene, particularly after ∼3 ka, when the PROB method suggests renewed, though relatively low magnitude, northern peatland initiation that it is not present in the HISTreconstruction. Cumulative curves of peatland initiation based on the HIST method applied to MacD06 reveal that ∼40% of peatlands were initiated by 9 ka and ∼25% of peatlands formed after 5 ka (Fig. 2D). In contrast, when MacD06 is reas- sessed using the PROB method, ∼30% of peatlands were initiated by 9 ka, and another ∼30% postdate 5 ka. The HIST and PROB-based reconstructions of thermokarst E lake initiation are broadly similar (Fig. 2E) because the modified HIST method used by Walter et al. (29) is based on midpoints of calibrated age ranges. This eliminates systematic biases in the frequency histogram that are the result of different lengths of calibrated age ranges. In general, and as suggested by Walter et al. (29), the most rapid increase in thermokarst lake initiation occurs after the Younger Dryas, with the most prominent increase occurring between ∼10.5 and 9.5 ka. Cumulative curves for thermokarst lake initiation indicate that nearly half (45%) of thermokarst lake initiation since 20 ka occurred between 9 and 11.5 ka, suggesting rapid thermokarst lake development after Termination 1B (Fig. 2E). High-magnitude, short-duration peaks in the thermokarst lake PROB curve, particularly after ∼8 ka, are Fig. 3. Profiles of deglacial and Holocene climate change, the global C probably overemphasized due to the small sample size (n ¼ 66 cycle as indicated by ice-core geochemistry, and peatland and thermokarst 14C dates) and therefore should be interpreted cautiously. lake initiation. Rapid changes in climate, including Terminations 1A (T1A) and 1B (T1B), the Younger Dryas (YD) and Bølling–Allerød (B–A), are also indicated. (A) Summer insolation for 60º N (dashed line) (37) and temperature Implications for Deglacial and Holocene Changes in Atmospheric CH4

reconstruction from the Greenland GISP2 ice core (solid line) (38). (B) Recon- GEOLOGY Concentration. Atmospheric CH4 concentration increased gradu- structed atmospheric CH4 concentrations from GISP2 (5), showing rapid ally between 19 and 15 ka, before increasing rapidly at the start shifts in CH4 concentrations during the Bølling-Allerød interstadial and – of the Bølling Allerød interstadial (Termination 1A) (Fig. 3 A the Younger Dryas chronozone, and the ∼100 ppb increase in CH4 concentra- and B). AMC remained stable around 680 ppb, then decreased tion between ∼5 ka and the start of the industrial revolution. Also shown is abruptly at the start of the Younger Dryas. At the end of the the interhemispheric CH4 gradient, which is compiled from refs. 5 (filled Younger Dryas (Termination 1B), AMC abruptly increased once triangles), 35 (white circles), and 36 (black squares). Higher interhemispheric CH4 gradient values indicate increased contribution from northern hemi- again, this time to over 700 ppb. Changes in the strength of CH4 δ13 sources are usually advanced as the main drivers of deglacial and sphere sources. (C) Profile of ice-core CH4 through the Holocene; data compiled from refs. 40 (filled squares) and 39 (open squares). (D) PROB-based SCIENCES early Holocene AMC fluctuations (30), though some modeling

10 ENVIRONMENTAL reconstructions of peatland initiation applied to both JY vet and MacD06. efforts suggest that variations in CH4 sinks may have been impor- (E) PROB-based reconstruction of thermokarst lake initiation events applied tant as well (e.g., 31). Various sources have been proposed for the to Walter et al. (29). abrupt high-magnitude increases in AMC during Terminations 1A and 1B. One hypothesis proposes release of large quantities as potential drivers of deglacial AMC, based on modeled inter- of CH4 from marine gas hydrates during deglaciation (32). How- hemispheric gradients in CH4 sources (5, 35, 36) and HIST-based 14 ever, isotopic analyses of ice-core CH4 (Fig. 3C) demonstrate that compilations of basal peat C dates (6, 7, 9, 10, 12, 16). CH4 clathrates were not responsible for the increases in AMC Alternatively, Walter et al. (29) suggested that CH4 ebullition during the Pleistocene–Holocene transition (33, 34). Methane from newly formed thermokarst lakes drove much of the in- emissions from expanding northern peatlands have been invoked creases in AMC during the deglacial and early Holocene interval.

Reyes and Cooke PNAS ∣ March 22, 2011 ∣ vol. 108 ∣ no. 12 ∣ 4751 Downloaded by guest on September 23, 2021 Our PROB-based reanalyses of basal peatland 14C dates reveal (16) recently compiled basal 14C dates from tropical and southern a previously unidentified lag between the abrupt rises in deglacial hemisphere peatlands, and suggested that these regions could AMC and the timing of northern peatland initiation. Peatland have been important source areas for atmospheric CH4 prior initiation in Alaska did not increase rapidly until ∼14 ka (Fig. 3D), to widespread initiation of northern peatlands. However, the well after the rapid initial increase in AMC during Termination tropical peatland 14C database in (16) includes relatively few 1A (Fig. 3B). A similar lag exists in the larger circumpolar dataset dates, hindering definitive evaluation of southern peatlands and of MacD06, which exhibits little peatland development from tropical wetlands as the source for the initial rises of AMC during 18–14 ka, and only a gradual increase between 13 and 14 ka. Terminations 1A and 1B. Nevertheless, the lag between the AMC The initiation of new thermokarst lakes also clearly lags pro- rise that precedes rapid increases in northern peatland and nounced warming in Greenland and the abrupt rise in AMC thermokarst lake initiation, as well as isotopic evidence against (Fig. 3E). Thus, we conclude that the initial rapid increase in the clathrate source hypothesis, suggests that temporal trends AMC at Termination 1A was not driven by an expansion of north- of tropical and southern hemisphere peatland initiation warrant ern peatlands or thermokarst lakes. further study. During the Younger Dryas, AMC declined by ∼200 ppb Atmospheric CH4 concentration declined between ∼10 and (Fig. 3B). Fischer et al. (40) report a reduced boreal CH4 source 8 ka, then remained relatively constant until ∼5 ka, when 13 during this period, based on δ CH4 values from the EPICA AMC began a gradual rise, increasing by ∼100 ppb over the late Dronning Maud Land ice core that are ∼2‰ higher than during Holocene (Fig. 3B). While the interhemispheric CH4 gradient the Bølling–Allerød (Fig. 3C). Furthermore, reconstructions of suggests that northern peatlands remained an important CH4 ∼2–3 the interhemispheric CH4 gradient are at their lowest during source until ka, rates of northern peatland initiation are the Younger Dryas (5, 35, 36) (Fig. 3B), also suggesting a reduc- generally thought to have decreased after ∼10 ka (6, 7, 9–12). tion in boreal sources of CH4. In contrast, previous HIST-based The perceived lack of substantial late Holocene peatland initia- compilations of basal 14C dates have reported little to no decline tion has opened the door for alternative drivers [e.g., early agri- in the rate of expansion for both Alaskan (9) and northern peat- cultural activities (42)] of late Holocene atmospheric CH4. lands during the Younger Dryas (6, 7). Our PROB-based reana- However, lateral expansion of northern peatlands increased lysis reconciles these differences by revealing a near-millennial- through the late Holocene (8), and the margins of expanding length reduction in the rate of northern peatland initiation during peatlands are typically minerotrophic fens, which are hot spots for CH4 emissions (43). In addition, our PROB record of basal the Younger Dryas (Fig. 3D). We acknowledge, however, that it 14 can be difficult to obtain reliable peatland 14C chronologies dur- C dates in Alaska also shows a pronounced rise in peatland ing the Younger Dryas because of a plateau in the 14C calibration initiation after ∼4.5 ka (Fig. 3D). Thus, our reanalysis, together curve (Fig. 2A). with reconstructions that incorporate lateral peatland expansion Atmospheric CH4 concentrations rose by ∼200 ppb in (8), implies that northern peatlands may be a potentially under- ∼200 years (Fig. 3B) at the end of the Younger Dryas ∼11.6 ka appreciated CH4 source that could have contributed to the late (Termination 1B), in concert with an abrupt increase in Green- Holocene rise in AMC. land temperatures (Fig. 3A). Earlier HIST-based analyses of 14C Radiocarbon dating is the foundation for efforts aiming to un- compilations proposed that Alaskan peatlands (9) and/or ther- derstand millennial patterns of peatland initiation. Our reanalysis of two basal peat 14C-date compilations shows that subtle mokarst lakes (29) drove much of this rapid increase in AMC. 14 As with the rise in AMC at Termination 1A, our reanalysis reveals differences in the selection and treatment of basal peat C dates that the initiation of new peatlands across the circum-Arctic lags can have important implications for interpretation of trends in the Termination 1B rise in AMC by 500–1000 yr (Fig. 3D). The peatland initiation. These trends have been incorporated into thermokarst lake PROB profile reveals a similar lag between C cycle models and estimates of global C sequestration in peat- – the expansion of thermokarst lakes and the rise in AMC (Fig. 3E). lands (16, 44 46). For example, the most current estimate of the We also note that refinements to the 14C calibration curve (41) global pool of C sequestered in peatlands is based on the product and inclusion of additional high-precision accelerator mass spec- of peat C accumulation rates and cumulative peatland area 14 σ through time (16). In turn, the estimates of cumulative peatland trometry C dates should yield smaller 2 calibrated age ranges, 14 which in turn will increase the temporal lag between peatland/ area are based on frequency histograms of binned C dates of thermokarst lake initiation and the abrupt rises in AMC. peatland initiation, an approach that we have shown is liable to introduce chronological artefacts. Future efforts will benefit Our reanalyses suggest northern peatlands and thermokarst 14 lakes did not initiate the abrupt increases in AMC during Termi- from more critical screening of existing C dates before they nations 1A and 1B. Nonetheless, northern peatlands were clearly are included in large compilations of basal peat dates, and should apply a more nuanced approach to the incorporation of cali- important CH4 sources during the early Holocene. The initiation 14 of new Alaskan peatlands reaches a maximum between 10 brated C dates in chronologies of peatland development. When and 11.5 ka, whereas the expansion of northern peatlands peaks such strategies are adopted, it is possible to discern important between 8 and 10 ka (Fig. 3D). In Alaska, this brief early Holo- temporal relations between peatland initiation and atmospheric cene interval contains ∼20% of the cumulative probability in CH4 concentrations that were previously obscured by systematic peatland basal dates (Fig. 2C), roughly equivalent to the cumu- methodological bias. lative probability contained within the preceding 7.5 ka. A similar Methods result is revealed by the circum-Arctic dataset, in which ∼20% of 14 14 We chose the JY10 dataset as a case study to evaluate the effects of C date the cumulative probability of calibrated C dates is contained vetting and different data synthesis methods because we are familiar with between 8 and 10 ka. Likewise, formation of new thermokarst the geographic setting and Quaternary history of Alaska, and because details lakes exhibits a bimodal peak between 9 and 10.5 ka. These early for 14C dates presented outside the refereed journal literature are widely Holocene increases in peatland and thermokarst lake formation available through the Alaska Division of Geological and Geophysical Surveys 14 are mirrored by an increase in the interhemispheric CH4 gradient website (http://www.dggs.dnr.state.ak.us/). In contrast, many C dates in (Fig. 3B), pointing to the increased importance of northern MacD06 are not readily available. We used a corrected dataset of 274 14C dates from JY10 for further ana- sources for the global CH4 budget after Termination 1B. 10 lysis (JY all; Dataset S1; http://www.pnas.org/content/107/16/7347/suppl/ If the initial abrupt increases in AMC during Terminations DCSupplemental). We then compiled information about each of these 274 1A and 1B were not caused by either a destabilization of CH4 dates from original sources, including: the material dated; the pretreatment hydrates or a rapid expansion of northern peatlands and thermo- method employed (if any); the counting method (i.e., radiometric or accel- karst lakes, then other CH4 sources must be advanced. Yu et al. erator mass spectrometry); stratigraphic context; and the original interpreta-

4752 ∣ www.pnas.org/cgi/doi/10.1073/pnas.1013270108 Reyes and Cooke Downloaded by guest on September 23, 2021 tion of the 14C date, if provided. We vetted the 14C dates and excluded those Dataset S2. To facilitate comparison of the two methods for compilation that were unreliable (SI Text, Table S1, and Dataset S1), resulting in a dataset of 14C dates, we normalized the HIST and PROB datasets to zero mean and 14 10 of 159 C dates (JY vet). unit variance (z-score), then rescaled the normalized scores so that the lowest 14 We calibrated C dates using CALIB v6.0 (47) and the IntCal09 calibration score had a value of zero. curve (48), and compiled the 2σ age ranges (Dataset S1) using two methods. 10 We examined spatial patterns of basal peat sampling sites in JY all and First, following many earlier studies of peatland initiation (6, 7, 9, 10, 12, 16), JY10 by mapping their distribution in relation to last glacial maximum ice we created a histogram of all the calibrated age ranges for JY10 ,JY10 , vet all vet limits and organic-rich Histosol and Histel soils (18) using ArcMap GIS (Fig. 1). and MacD06 by rounding the extremes of each 2σ age range to the nearest In the US Department of Agriculture soil taxonomy, Histosols have at least 50 yr, compiling each calibrated age range into 50-yr bins, and summing the >12% number of calibrated ranges for all 14C dates within each 50-yr bin (HIST 40 cm of organic material ( organic C) in the uppermost 80 cm of method; e.g., Fig. S1). We note that our HIST record of peatland initiation the soil profile, whereas Histels are similar to Histosols but are perennially for MacD06 differs slightly from (7) because they used 1-yr bins, as opposed frozen within 100 cm of the soil surface. to our 50-yr bins. We followed the methods used in (29) for the HIST-based 14 compilation of basal C ages from thermokarst lakes; we counted the num- ACKNOWLEDGMENTS. Our research would not have been possible without ber of midpoints from each 2σ age range within 1,000-yr bins centered on support from D. G. Froese and A. P. Wolfe (Department of Earth and successive 100-yr intervals. Second, using a routine within CALIB v6.0, we Atmospheric Sciences, University of Alberta), both of whom provided generated a summed probability curve for each dataset by summing the funding, helpful discussions, and feedback on an early draft of the manu- probability density functions associated with each calibrated age and then script. J. J. Clague and W. O. Hobbs also provided valuable comments on a calculating 100-yr means of the summed probabilities (PROB method). We draft manuscript. We are indebted to two anonymous reviewers and the repeated the PROB analysis on MacD06 using 50- and 500-yr means of the editor, who made important suggestions that led to substantial improve- summed probabilities, and note that the broad PROB-based trends of ments in the manuscript. We thank J. P. Briner, M. E. Edwards, M. C. Jones, northern peatland initiation appear relatively insensitive to the length of and Z. Yu for helpful discussions. The Office of the Vice-President (Research) time used for calculating mean probabilities (Fig. S4). Summed probability at the University of Alberta provided generous assistance with publication 10 data for JY vet, MacD06, and thermokarst lakes (29) are available in charges.

1. Gorham E (1991) Northern peatlands: Role in the carbon cycle and probable responses 24. Kennett DJ, Stafford TW, Southon J (2008) Standards of evidence and Paleoindian to climatic warming. Ecol Appl 1:182–195. demographics. Proc Natl Acad Sci USA 105:E107. 2. Turunen J, Tomppo E, Tolonen K, Reinikainen A (2002) Estimating carbon accumula- 25. Macklin MG, Jones AF, Lewin J (2010) River response to rapid Holocene environ- tion rates of undrained mires in Finland-application to boreal and subarctic regions. mental change: evidence and explanation in British catchments. Quaternary Sci Rev Holocene 12:69–80. 29:1555–1576. 3. Harriss RC, Gorham E, Sebacher DI, Bartlett KB, Flebbe PA (1985) Methane flux from 26. MacDonald G, Beukens RP, Kieser WE (1991) Radiocarbon dating of limnic sediments: northern peatlands. Nature 315:652–654. A comparative analysis and discussion. Ecology 72:1150–1155. 4. Mikaloff Fletcher SE, Tans PP, Bruhwiler LM, Miller JB, Heimann M (2004) CH4 sources 27. Reger RD, Sturmann AG, Berg EE, Burns PAC (2007) A Guide to the Late Quaternary 13 12 estimated from atmospheric observations of CH4 and its C∕ C isotopic ratios: 1. History of the Northern and Western Kenai Peninsula, Alaska. (Alaska Division of Geo- Inverse modeling of source processes. Global Biogeochem Cy 18:GB4004. logical and Geophysical Surveys, Fairbanks, AK). 5. Brook EJ, Harder S, Severinghaus J, Steig EJ, Sucher CM (2000) On the origin and timing 28. Kip N, et al. (2010) Global prevalence of methane oxidation by symbiotic bacteria in of rapid changes in atmospheric methane during the last glacial period. Global Bio- peat-moss ecosystems. Nat Geosci 3:617–621. – geochem Cy 14:559 572. 29. Walter KM, Edwards ME, Grosse G, Zimov SA, Chapin FS, III (2007) Thermokarst lakes as 6. Smith LC, et al. (2004) Siberian peatlands a net carbon sink and global methane source a source of atmospheric CH4 during the last deglaciation. Science 318:633–636. – since the early Holocene. Science 303:353 356. 30. Martinerie P, Brasseur GP, Granier C (1995) The chemical composition of ancient atmo- 7. MacDonald GM, et al. (2006) Rapid early development of circumarctic peatlands and spheres: A model study constrained by ice core data. J Geophys Res 100:14291–14304. – atmospheric CH4 and CO2 variations. Science 314:285 288. 31. Kaplan JO, Folberth G, Hauglustaine DA (2006) Role of methane and biogenic volatile 8. Korhola A, et al. (2010) The importance of northern peatland expansion to the organic compound sources in late glacial and Holocene fluctuations of atmospheric – late-Holocene rise of atmospheric methane. Quaternary Sci Rev 29:611 617. methane concentrations. Global Biogeochem Cy 20:GB2016. 9. Jones MC, Yu Z (2010) Rapid deglacial and early Holocene expansion of peatlands in 32. Kennett JP, Cannariato KG, Hendy IL, Gehl RJ (2003) Methane Hydrates in Quaternary Alaska. Proc Natl Acad Sci USA 107:7437–7352. Climate Change (American Geophysical Union, Washinton, DC). 10. Campbell ID, Campbell C, Yu Z, Vitt DH, Apps MJ (2000) Millennial-scale rhythms in 33. Sowers T (2006) Late Quaternary atmospheric CH4 isotope record suggests marine peatlands in the western interior of Canada and in the global carbon cycle. Quaternary clathrates are stable. Science 311:838–840. Res 54:155–158. 13 34. Schaefer H, et al. (2006) Ice record of δ C for atmospheric CH4 across the Younger 11. Kremenetski KV, et al. (2003) Peatlands of the Western Siberian lowlands: Current Dryas–Preboreal transition. Science 313:1109–1112. knowledge on zonation, carbon content and Late Quaternary history. Quaternary 35. Chappellaz J, et al. (1997) Changes in the atmospheric CH4 gradient between Sci Rev 22:703–723. Greenland and Antarctica during the Holocene. J Geophys Res 102:15987–15997. 12. Yu Z, Beilman DW, Jones MC (2009) Carbon Cycling in Northern Peatlands, eds 36. Dällenbach A, et al. (2000) Changes in the atmospheric CH4 gradient between AJ Baird, LR Belyea, X Comas, AS Reeve, and LD Slater (American Geophysical Union, Greenland and Antarctica during the Last Glacial and the transition to the Holocene. Washington, DC), 184, pp 55–69. Geophys Res Lett 27:1005–1008. 13. Gorham E, Lehman C, Dyke A, Janssens J, Dyke L (2007) Temporal and spatial aspects of 37. Berger A, Loutre MF (1991) Insolation values for the climate of the last 10 million years. peatland initiation following deglaciation in North America. Quaternary Sci Rev Quaternary Sci Rev 10:297–317. 26:300–311. 38. Alley RB (2000) The Younger Dryas cold interval as viewed from central Greenland. 14. Turetsky MR, Manning SW, Wieder RK (2004) Dating recent peat deposits. Wetlands – Quaternary Sci Rev 19:213–226.

24:324 356. GEOLOGY 15. Stuiver M, Suess HE (1966) On the relationship between radiocarbon dates and true 39. Sowers T (2010) Atmospheric methane isotope records covering the Holocene period. – sample ages. Radiocarbon 8:534–540. Quaternary Sci Rev 29:213 221. 16. Yu Z, Loisel J, Brosseau DP, Beilman DW, Hunt SJ (2010) Global peatland dynamics 40. Fischer H, et al. (2008) Changing boreal methane sources and constant biomass – since the last glacial maximum. Geophys Res Lett 37:L13402. burning during the last termination. Nature 452:864 867. 14 17. Kaufman DS, Young NE, Briner JP, Manley WF (2011) Quaternary Glaciations—Extent 41. Muscheler R, et al. (2008) Tree rings and ice cores reveal C calibration uncertainties – and Chronology, Part IV: A Closer Look, eds J Ehlers, PL Gibbard, and PD Hughes (Else- during the Younger Dryas. Nat Geosci 1:263 267. vier, Amsterdam), (in press). 42. Ruddiman WF (2003) The anthropogenic greenhouse era began thousands of years – 18. Tarnocai C, et al. (2002) Northern Circumpolar Soils 1:10,000,000 scale map (Research ago. Climatic Change 61:261 293. Branch, Agriculture and Agri-Food Canada, Distributed by the National Snow and Ice 43. Rydin H, Heglum JK (2006) The Biology of Peatlands (Oxford University Press, Oxford).

Data Center/World Data Center for Glaciology, Boulder, CO). 44. Frolking S, Roulet NT (2007) Holocene radiative forcing impact of northern peatland SCIENCES 19. Buchanan B, Collard M, Edinborough K (2008) Paleoindian demography and the carbon accumulation and methane emissions. Glob Change Biol 13:1079–1088. ENVIRONMENTAL extraterrestrial impact hypothesis. Proc Natl Acad Sci USA 105:11651–11654. 45. Wang Y, Roulet NT, Frolking S, Mysak LA (2009) The importance of Northern Peatlands 20. Pierce JL, Meyer GA, Jull AJT (2004) Fire-induced erosion and millennial-scale climate in global carbon systems during the Holocene. Clim Past 5:683–693. change in northern ponderosa pine forests. Nature 432:87–90. 46. Kleinen T, Brovkin V, von Bloh W, Archer D, Munhoven G (2010) Holocene carbon cycle 21. Hunt TL, Lipo CP (2006) Late colonization of Easter Island. Science 311:1603–1606. dynamics. Geophys Res Lett 37:L02705. 22. Blackwell PG, Buck CE (2003) The Late Glacial human reoccupation of north-western 47. Stuiver M, Reimer PJ (1993) Extended 14C database and revised CALIB radiocarbon Europe: New approaches to space-time modelling. Antiquity 77:232–240. calibration program. Radiocarbon 35:215–230. 23. Surovell TA, Brantingham PJ (2007) A note on the use of temporal frequency distribu- 48. Reimer PJ, et al. (2009) IntCal09 and Marine09 radiocarbon age calibration curves, tions in studies of prehistoric demography. J Archaeol Sci 34:1868–1877. 0–50,000 years cal BP. Radiocarbon 51:1111–1150.

Reyes and Cooke PNAS ∣ March 22, 2011 ∣ vol. 108 ∣ no. 12 ∣ 4753 Downloaded by guest on September 23, 2021