Climate-related changes in peatland carbon accumulation during the last millennium

Biogeosciences, 10, 929–944, 2013 www.biogeosciences.net/10/929/2013/ doi:10.5194/bg-10-929-2013 © Author(s) 2013. CC Attribution 3.0 License. EGU Journal Logos (RGB) Advances in Geosciences Open Access Natural Hazards and Earth System Sciences Open Access Annales Geophysicae Open Access Nonlinear Processes in Geophysics Open Access Atmospheric Chemistry and Physics Open Access Atmospheric Chemistry and Physics Open Access Discussions Atmospheric Measurement Techniques Open Access Atmospheric Measurement Techniques Open Access Discussions Biogeosciences Open Access Open Access Biogeosciences Discussions Climate of the Past Open Access Open Access Climate of the Past Discussions Earth System Dynamics Open Access Open Access Earth System Dynamics Discussions Geoscientific Instrumentation Methods and Data Systems Open Access Geoscientific Instrumentation Methods and Data Systems Open Access Discussions Geoscientific Model Development Open Access Open Access Geoscientific Model Development Discussions Hydrology and Earth System Sciences Open Access Hydrology and Earth System Sciences Open Access Discussions Ocean Science Open Access Open Access Ocean Science Discussions Solid Earth Open Access Open Access Solid Earth Discussions The Cryosphere Open Access Open Access The Cryosphere Discussions Natural Hazards and Earth System Sciences Open Access Discussions Climate-related changes in peatland carbon accumulation during the last millennium D. J. Charman1, D. W. Beilman2, M. Blaauw3, R. K. Booth4, S. Brewer5, F. M. Chambers6, J. A. Christen7, A. Gallego-Sala8,9,10, S. P. Harrison9,11, P. D. M. Hughes12, S. T. Jackson13, A. Korhola14, D. Mauquoy15, F. J. G. Mitchell16, I. C. Prentice11,17, M. van der Linden18, F. De Vleeschouwer19, Z. C. Yu4, J. Alm20, I. E. Bauer21, Y. M. C. Corish16, M. Garneau22, V. Hohl1, Y. Huang23, E. Karofeld24, G. Le Roux19, J. Loisel4, R. Moschen25, J. E. Nichols26, T. M. Nieminen27, G. M. MacDonald28, N. R. Phadtare29, N. Rausch30,¨ U. Sillasoo31, G. T. Swindles32, E.-S. Tuittila14, L. Ukonmaanaho27, M. V¨ aliranta14, S. van Bellen15, B. van Geel33, D. H. Vitt34, and Y. Zhao35 1Department of Geography, College of Life and Environmental Sciences, University of Exeter, EX4 4RJ, UK 2Department of Geography, University of Hawai‘i M¯ anoa, Honolulu, HI 96822, USA 3School of Geography, Archaeology and Palaeoecology, Queen’s University Belfast, Belfast BT7 1NN, UK 4Department of Earth and Environmental Sciences, Lehigh University, Bethlehem, PA 18015, USA 5Department of Geography, University of Utah, Salt Lake City, UT 84112, USA 6Centre for Environmental Change and Quaternary Research, SNSS, University of Gloucestershire, Cheltenham, GL50 4AZ, UK 7Centro de Investigaci´ on en Matem´ aticas, A.P. 402, 36000 Guanajuato, Gto., Mexico 8QUEST, Department of Earth Sciences, University of Bristol, Bristol, BS8 1RJ, UK 9School of Geographical Sciences, University of Bristol, Bristol, BS8 1SS, UK 10Department of Earth and Ecosystem Sciences, Lund Universitet, S¨ olvegatan 12, 223 62 Lund, Sweden 11Department of Biological Sciences, Macquarie University, North Ryde, NSW 2109, Australia 12Geography and Environment, University of Southampton, Highfield, Southampton, Hants, SO17 1BJ, UK 13Department of Botany, University of Wyoming, Laramie, WY 82071, USA 14Department of Environmental Sciences, P.O. Box 65, University of Helsinki, 00014 Helsinki, Finland 15School of Geosciences, University of Aberdeen, Elphinstone Road, Aberdeen AB24 3UF, UK 16Botany Department, Trinity College Dublin, Dublin 2, Ireland 17Grantham Institute for Climate Change and Department of Life Sciences, Imperial College, Silwood Park, Ascot, SL5 7PY, UK 18BIAX Consult, Hogendijk 134, 1506 AL Zaandam, The Netherlands 19CNRS and Universit´ e de Toulouse, INP, UPS, EcoLab, ENSAT, Avenue de l’Agrobiopole, 31326 Castanet-Tolosan, France 20School of Forest Sciences, University of Eastern Finland, P.O. Box 68, 80101 Joensuu, Finland 21Sir Wilfred Grenfell College, Memorial University of Newfoundland, Corner Brook, Newfoundland A2H 6P9, Canada 22D´ epartement de G´ eographie and GEOTOP, Universit´ e du Qu´ ebec ` a Montr´ eal, Montr´ eal, Quebec, H3C 3P8, Canada 23Department of Geological Sciences, Brown University, Providence, RI 02912, USA 24University of Tartu, Institute of Ecology and Earth Sciences, Lai 40, Tartu 51005, Estonia 25Institute of Bioand Geosciences 3: Agrosphere, Research Centre Juelich (Forschungszentrum J¨ ulich), J¨ ulich 52428, Germany 26NASA Goddard Institute for Space Studies, 2880 Broadway, New York, NY 10025, USA, and Lamont-Doherty Earth Observatory at Columbia University, Palisades, NY 10964, USA 27The Finnish Forest Research Institute, P.O. Box 18, 01301 Vantaa, Finland 28Institute of the Environment and Sustainability, and Department of Geography, UCLA, Los Angeles, CA 90095, USA 29Wadia Institute of Himalayan Geology, Dehra Dun, India 30Institute of Environmental Geochemistry, University of Heidelberg, Im Neuenheimer Feld 236, 69120 Heidelberg, Germany 31Institute of Ecology, Tallinn University, Uus-Sadama 5, Tallinn 10120, Estonia Published by Copernicus Publications on behalf of the European Geosciences Union. 930 D. J. Charman et al.: Climate-related changes in peatland carbon accumulation 32School of Geography, University of Leeds, Leeds, LS2 9JT, UK 33Institute for Biodiversity and Ecosystem Dynamics, P.O. Box 94248 1090 GE Amsterdam, The Netherlands 34Department of Plant Biology, Southern Illinois University, Carbondale, IL 62901, USA 35Institute of Geographic Science and Natural Resources Research, Chinese Academy of Sciences, Beijing 100101, China Correspondence to: D. J. Charman ([email protected]) Received: 2 October 2012 – Published in Biogeosciences Discuss.: 17 October 2012 Revised: 11 January 2013 – Accepted: 11 January 2013 – Published: 8 February 2013 Abstract. Peatlands are a major terrestrial carbon store and a persistent natural carbon sink during the Holocene, but there is considerable uncertainty over the fate of peatland carbon in a changing climate. It is generally assumed that higher temperatures will increase peat decay, causing a positive feedback to climate warming and contributing to the global positive carbon cycle feedback. Here we use a new extensive database of peat profiles across northern high latitudes to examine spatial and temporal patterns of carbon accumulation over the past millennium. Opposite to expectations, our results indicate a small negative carbon cycle feedback from past changes in the long-term accumulation rates of northern peatlands. Total carbon accumulated over the last 1000yr is linearly related to contemporary growing season length and photosynthetically active radiation, suggesting that variability in net primary productivity is more important than decomposition in determining long-term carbon accumulation. Furthermore, northern peatland carbon sequestration rate declined over the climate transition from the Medieval Climate Anomaly (MCA) to the Little Ice Age (LIA), probably because of lower LIA temperatures combined with increased cloudiness suppressing net primary productivity. Other factors including changing moisture status, peatland distribution, fire, nitrogen deposition, permafrost thaw and methane emissions will also influence future peatland carbon cycle feedbacks, but our data suggest that the carbon sequestration rate could increase over many areas of northern peatlands in a warmer future. 1 Introduction Peatlands contain around 600gigatonnes of carbon (GtC) that has accumulated since the last glacial maximum in northern mid–high latitudes, tropical regions and temperate areas of the Southern Hemisphere, and the steady accumulation of carbon has been a small but persistent sink for atmospheric CO2throughout the Holocene (Yu, 2011). The relationship between climate change and the rate of carbon sequestration is important for understanding the past and future global carbon cycle, and it has generally been assumed that because temperature drives increasing decay (Ise et al., 2008; Dorrepaal et al., 2009), peatlands could be part of the positive feedback from the global carbon cycle (Friedlingstein et al., 2006). A key objective in improving understanding of the global carbon cycle in climate models is to be able to simulate past observed atmospheric CO2changes. There is growing interest in the last millennium as a climate-modelling target, and especially in the assessment of the sensitivity of the global carbon cycle to climate warming (Abe-Ouchi and Harrison, 2009; Jungclaus et al., 2010). In the Northern Hemisphere, the transition from the generally warmer Medieval Climate Anomaly (MCA) to the cooler Little Ice Age (LIA) (Mann et al., 2008, 2009; Jansen et al., 2007) was associated with a ca. 7–10ppmv decline in atmospheric CO2concentration (Ahn et al., 2012). This pattern supports the existence of a positive global climate–carbon cycle feedback, as suggested by coupled climate–carbon cycle models (Friedlingstein et al., 2006; Denman et al., 2007). However, estimates of the magnitude of the climate sensitivity of the global carbon cycle based on data from the last millennium vary from 1.7–21.4ppmCO2K−1(Frank et al., 2010) to 40–60ppmCO2K−1(Cox and Jones, 2008). Carbon cycle models also vary greatly in their assessment of this feedback (Friedlingstein et al., 2006), although recent estimates (Jungclaus et al., 2010) suggest sensitivity within the lower end of this range (3.2–12ppmCO2K−1). The causes of the reduction in CO2concentrations during the MCA to LIA transition are poorly known, but reduced soil heterotrophic respiration is assumed to be important (Jungclaus et al., 2010; Pongratz et al., 2009). However, the models do not specifically take into account possible climate-related variations in the rate of peatland carbon sequestration. Peatlands have sequestered and exchanged atmospheric carbon over millennia (MacDonald et al., 2006; Frolking and Roulet, 2007), with the largest store in northern extratropical peatlands, an estimated 545GtC (Yu et al., 2010). The annual uptake of CO2by peatlands, previously estimated as 0.076GtCyr−1(Gorham, 1991) or 0.088GtCyr−1(Yu, 2011), without considering long-term decay (see below), is a small but temporally persistent component of land carbon uptake. This is equivalent to 36ppm atmospheric CO2over 1000yr, based on a simple conversion from change in carbon pool to atmospheric CO2of 1GtC=2.123ppm. However, Biogeosciences, 10, 929–944, 2013 www.biogeosciences.net/10/929/2013/ D. J. Charman et al.: Climate-related changes in peatland carbon accumulation 931 0 10 31 50 82 C mass per unit area (kg m-2) a 6 5 4 3 2 1 0 0 1,000 2,000 3,000 4,000 5,000 6,000 Precipitation / equilibrium evapotranspiration Growing degree days above 0°C b Fig. 1. Distribution of sites in geographic and climate space, overlain on soil carbon storage. Points show high-resolution peat records (large blue circles) used for calculating temporal variation in carbon accumulation rates and low-resolution peat records (small white circles) used for estimates of total millennial carbon. (a) The soil carbon density in the top 1m mapped from 10-min IGBP soil data; (b) climate space (defined by growing degree days and moisture balance) of soil carbon density in the same classes as in panel a, at 0.5×0.5◦grid cells for all land north of 40◦N. Peatlands generally occur in the areas of >31kgCm−2(brown and black). over millennial timescales, carbon uptake of this magnitude would be compensated by ocean outgassing processes as a result of reduced CO2in the atmosphere and reduction in air–sea CO2partial pressure, so that the actual effect on the atmosphere is only 20–35% of this total over periods of 200– 2000yr (Archer et al., 2009), or 7–12ppm atmospheric CO2 over a 1000yr period. Variations in the size of the peatland sink could therefore have a significant cumulative effect on global atmospheric CO2concentrations over the last millennium, of the same order of magnitude as the observed changes. In this study, we compiled peat core data from northern peatlands to estimate changes in carbon accumulation over the last millennium and to explore the spatial relationship between climate and the total size of the carbon sink accumulated over this period. Further analysis on a subset of well-dated cores allowed an analysis of temporal variation in carbon accumulation in relations to the MCA–LIA climate changes estimated from palaeoclimate records. We use these data to help understand the relationships between climate and peatland carbon accumulation and to assess the direction and strength of the peatland carbon cycle feedback. 2 Methods 2.1 Site selection and carbon measurement A list of Northern Hemisphere, extratropical peatland profiles with published and unpublished carbon accumulation data was compiled (Tables 1 and 2) for sites that met the following criteria: a. at least 3 evenly spaced dates (including 210Pb, tephra, spheroidal carbonaceous particles, pollen markers or 14C (preor postdating the period of nuclear bomb testing), and the uncut peat surface) and spanning the last ca. 1000yr. Most of the sites had more than 5 dates (Table 2), but we also rejected some sites where a satisfactory age–depth model could not be produced, because of age reversals or other problems; and b. contiguous bulk density measurements at <5cm resolution. Application of these criteria resulted in the selection of 24 sites that were used in subsequent analyses of the temporal changes in carbon accumulation through the last millennium (Table 2). A second tier of sites (Table 1) was used for the millennium carbon inventory analysis against climate indices. These sites did not meet criteria (a) and (b), but did meet the following criteria: c. a basic age–depth model for the last millennium; and d. contiguous bulk density measurements but not necessarily at high resolution. A total of 90 sites met these less stringent criteria and were used in the climate–carbon inventory analyses (Tables 1 and 2). The sites in both data sets are widely distributed geographically and broadly representative of the climate space occupied by northern peatlands (Fig. 1). www.biogeosciences.net/10/929/2013/ Biogeosciences, 10, 929–944, 2013 932 D. J. Charman et al.: Climate-related changes in peatland carbon accumulation Table 1. Characteristics of the low-resolution sites used in the analyses. No. Site name Latitude Longitude Peatland type Contributor Source reference 1 E115 67.8095 75.4346 Ombrotrophic D. Beilman Beilman et al. (2009) 2 E110 66.4698 76.9943 Ombrotrophic D. Beilman Beilman et al. (2009) 3 E113 66.4497 79.3234 Ombrotrophic D. Beilman Beilman et al. (2009) 4 D122 65.5831 73.0058 Ombrotrophic D. Beilman Beilman et al. (2009) 5 E119 65.4998 75.5025 Ombrotrophic D. Beilman Beilman et al. (2009) 6 D127 64.3068 70.2948 Ombrotrophic D. Beilman Beilman et al. (2009) 7 G136 64.1476 75.3611 Ombrotrophic D. Beilman Beilman et al. (2009) 8 G137 63.7504 75.7662 Ombrotrophic D. Beilman Beilman et al. (2009) 9 N015 63.6501 74.2693 Ombrotrophic D. Beilman Beilman et al. (2009) 10 N001 63.1611 74.8233 Ombrotrophic D. Beilman Beilman et al. (2009) 11 S009 62.1229 73.8412 Ombrotrophic D. Beilman Beilman et al. (2009) 12 V034 61.4675 79.4601 Ombrotrophic D. Beilman Beilman et al. (2009) 13 V039 61.0895 79.3806 Ombrotrophic D. Beilman Beilman et al. (2009) 14 SIB02 61.0553 70.0588 Ombrotrophic D. Beilman Beilman et al. (2009) 15 V026 61.0286 76.4686 Ombrotrophic D. Beilman Beilman et al. (2009) 16 S022 60.8401 71.2558 Ombrotrophic D. Beilman Beilman et al. (2009) 17 V038 60.8039 74.5416 Ombrotrophic D. Beilman Beilman et al. (2009) 18 SIB01 59.3601 68.9849 Ombrotrophic D. Beilman Beilman et al. (2009) 19 SIB06 58.4358 83.4343 Ombrotrophic D. Beilman Beilman et al. (2009) 20 SIB05 57.3541 81.1647 Ombrotrophic D. Beilman Beilman et al. (2009) 21 SIB03 56.3552 79.0689 Ombrotrophic D. Beilman Beilman et al. (2009) 22 C site 01 60.167 72.8330 Ombrotrophic Z. C. Yu Yu et al. (2009) 23 C site 02 60.167 72.8330 Ombrotrophic Z. C. Yu Yu et al. (2009) 24 C site 03 56.833 78.4170 Ombrotrophic Z. C. Yu Yu et al. (2009) 25 C site 10 54.15 −130.2500 Ombrotrophic Z. C. Yu Yu et al. (2009) 26 C site 13 55.017 −114.1500 Ombrotrophic Z. C. Yu Yu et al. (2009) 27 C site 17 55.85 −107.6830 Ombrotrophic Z. C. Yu Yu et al. (2009) 28 C site 20 59.883 −104.2000 Ombrotrophic Z. C. Yu Yu et al. (2009) 29 C site 21 45.684 −74.0470 Ombrotrophic Z. C. Yu Yu et al. (2009) 30 C site 23 47.933 −64.5000 Ombrotrophic Z. C. Yu Yu et al. (2009) 31 C site 24 45.2 −60.2670 Ombrotrophic Z. C. Yu Yu et al. (2009) 32 C site 25 57.522 −5.1600 Ombrotrophic Z. C. Yu Yu et al. (2009) 33 C site 26 57.56 −5.3770 Ombrotrophic Z. C. Yu Yu et al. (2009) 34 C site 27 57.687 −5.6870 Ombrotrophic Z. C. Yu Yu et al. (2009) 35 C site 28 68.4 23.5500 Ombrotrophic Z. C. Yu Yu et al. (2009) 36 C site 30 60.817 26.9500 Ombrotrophic Z. C. Yu Yu et al. (2009) 37 Kohlh¨ utten Moor 47.9269 8.1844 Ombrotrophic G. Le Roux Le Roux et al. (2005) 38 P131 66.1664 73.9889 Minerotrophic D. Beilman Beilman et al. (2009) 39 C site 04 56.333 84.5830 Minerotrophic Z. C. Yu Yu et al. (2009) 40 C site 05 60.446 −151.2470 Minerotrophic Z. C. Yu Yu et al. (2009) 41 C site 06 60.641 −151.0800 Minerotrophic Z. C. Yu Yu et al. (2009) 42 C site 07 60.416 −150.9020 Minerotrophic Z. C. Yu Yu et al. (2009) 43 C site 08 60.784 −150.8190 Minerotrophic Z. C. Yu Yu et al. (2009) 44 C site 09 64.875 −147.7670 Minerotrophic Z. C. Yu Yu et al. (2009) 45 C site 11 53.583 −118.0170 Minerotrophic Z. C. Yu Yu et al. (2009) 46 C site 12 52.45 −116.2000 Minerotrophic Z. C. Yu Yu et al. (2009) 47 C site 14 61.8 −121.4000 Minerotrophic Z. C. Yu Yu et al. (2009) 48 C site 15 68.288 −133.2500 Minerotrophic Z. C. Yu Yu et al. (2009) 49 C site 16 69.493 −132.6720 Minerotrophic Z. C. Yu Yu et al. (2009) 50 C site 18 64.713 −105.5790 Minerotrophic Z. C. Yu Yu et al. (2009) 51 C site 19 66.451 −104.8350 Minerotrophic Z. C. Yu Yu et al. (2009) 52 C site 22 82.333 −68.2500 Minerotrophic Z. C. Yu Yu et al. (2009) 53 C site 29 68.4 23.5500 Minerotrophic Z. C. Yu Yu et al. (2009) 54 C site 31 65.65 27.3170 Minerotrophic Z. C. Yu Yu et al. (2009) 55 C site 32 65.65 27.3170 Minerotrophic Z. C. Yu Yu et al. (2009) 56 C site 33 65.65 27.3170 Minerotrophic Z. C. Yu Yu et al. (2009) Biogeosciences, 10, 929–944, 2013 www.biogeosciences.net/10/929/2013/ D. J. Charman et al.: Climate-related changes in peatland carbon accumulation 933 Table 1 . Continued. No. Site name Latitude Longitude Peatland type Contributor Source reference 57 Old Black Spruce Fen 53.9983 −105.1153 Minerotrophic I. Bauer Bauer et al. (2009) 58 Sandhill Fen 53.8261 −104.6250 Minerotrophic I. Bauer Bauer et al. (2009) 59 Dhakuri 30.0500 79.9333 Minerotrophic N. R. Phadtare Unpublished data 60 Aeroport 4 54.1041 −72.5167 Minerotrophic M. Garneau Unpublished data 61 Aeroport 5 54.1041 −72.5167 Minerotrophic M. Garneau Unpublished data 62 Lac Le Caron Lateral Core4 52.2945 −75.8408 Ombrotrophic S. van Bellen van Bellen et al. (2011) 63 M179 60.5875 −149.5347 Minerotrophic Z. C. Yu Unpublished data 64 ZB08-S4 33.0954 102.6650 Minerotrophic Y. Zhao Unpublished data 65 OURS4 Peatland 54.0597 −72.4602 Minerotrophic M. Garneau Unpublished data 66 LG1 Peatland 54.0597 −78.4602 Minerotrophic M. Garneau Unpublished data Bulk density was measured on carefully cut fresh or frozen material using freeze drying or oven drying of samples of known volume. Sample sizes varied depending on the sampling method and core size, and sample resolution varied from 0.5 to 5cm3(Table 2). In all cases samples were large enough to accurately measure bulk density and were taken contiguously to enable reliable estimates of dry mass accumulation over time. Carbon density was derived from bulk density multiplied by the carbon content for each sample. Where carbon data were not available, we assumed that 50% of the organic fraction (measured by standard losson-ignition analysis at 500◦C) was organic carbon. A carbon value of approximately 50% is routinely used for peat (Gorham, 1991; Vitt et al., 2000) and is reasonable compared to the mean carbon content of the nine sites for which we have measured values in this study (46.6±0.33%), and other studies in western Canada (51.8%; Yu et al., 2009) and West Siberia (50.7–56.3%; Beilman et al., 2009). To provide an assessment of hydrological differences among the peatlands in our analyses, we classified sites as either bogs or fens. Although differences between these two peatland types are related to the relative influence of different water sources (i.e. groundwater, surface water, precipitation), thresholds used for distinction between the two types are regionally varied. For our site classification, we used a relatively conservative approach, including only Sphagnumdominated systems that lacked vegetative or morphological evidence of minerotrophic conditions in our “ombrotrophic” category. Sites characterised as ombrotrophic included raised bogs, blanket bogs, and the extensive bog systems of western Siberia (Kremenetski et al., 2003). 2.2 Chronology and age modelling All sites were 14C dated using selected aboveground plant remains, except for site 68 where bulk peat was 14C dated. We recalibrated all the dates from the original studies. For modern (post-AD 1950) 14C dates, the NH1 postbomb calibration curve was used (Hua and Barbetti, 2004). Remaining dates were calibrated using IntCal09 (Reimer et al., 2009). Age models for the temporal analysis were based on the program “Bacon”, a flexible Bayesian age–depth modelling approach that uses prior information on plausible accumulation rates and their variability and autocorrelation over time (Blaauw and Christen, 2005, 2011). Peat cores were divided into contiguous 2cm segments, and linear accumulation rates were calculated for all individual segments sequentially down the core. Age models were developed based on several million iterations, followed by thinning to remove any autocorrelation between individual model runs, yielding ca. 5000–8000 iterations for each site (Fig. 2a). 2.3 Spatial analysis of carbon accumulation Total carbon accumulation over the 1000yr was estimated based on a data set of the 90 radiocarbon-dated peat profiles (Tables 1 and 2). The post-1000yr carbon pool is the difference between total carbon additions from photosynthesis and cumulative respirative carbon release over this interval, reflecting carbon sequestration at a site. We analysed the relationship between total carbon accumulation over the last 1000yr and climate parameters using a 0.5◦grid, derived from the CLIMATE 2.2 data (Kaplan et al., 2003). Climate parameters included growing degree days above 0◦C (GDD0), cumulative photosynthetically active radiation during the growing season (PAR0), PAR over the growing season, growing season length (days) and the moisture index P/Eq, where Pis annual precipitation and Eq is annually integrated equilibrium evapotranspiration calculated from daily net radiation and temperature (Prentice et al., 1993). PAR was calculated from latitude and sunshine hours (Prentice et al., 1993; Harrison et al., 2010). 2.4 Temporal variation in carbon accumulation A composite carbon accumulation curve was constructed based on the subset of 24 well-dated, high-resolution sites with continuous records for the past 1000yr (Table 2). The Bayesian age–depth models allowed chronological uncertainty to be included in carbon accumulation curves (Fig. 2b). www.biogeosciences.net/10/929/2013/ Biogeosciences, 10, 929–944, 2013 934 D. J. Charman et al.: Climate-related changes in peatland carbon accumulation Table 2 . Characteristics of the high-resolution sites used in the analyses. BDs – bulk density sample size; BDi – bulk density increment depth; dates (other): Pb – lead 210, S – spheroidal carbonaceous particles, T – tephra, A – Ambrosia pollen rise; scaling for carbon estimates based on C (carbon analyses), LOI (loss on ignition), or 50% of dry mass. No. Site Lat. Long. Alt. Peatland BDs BDi 14C Dates C Main Site Site name (m) type (cm3) (cm) dates (other) scaling investigator code reference 67 Ballyduff 53.0861 −7.9930 60 Ombrotrophic 45–117 1 12 S C F. J. G. Mitchell ir ballyd Unpublished data 68 Upper Pinto Fen 53.5833 −118.0167 1310 Minerotrophic 0.8 1 4 LOI Z. C. Yu ca uppin Yu et al. (2003) 69 M¨ annikj¨ arve Bog 58.8753 26.2556 79 Ombrotrophic 200–250 5 10 Pb 50% E. Karofeld es mannik Sillasoo et al. (2007) 70 Misten 50.5631 6.1642 620 Ombrotrophic 7–10 1 11 Pb LOI F. De Vleeschouwer be misten De Vleeschouwer et al. (2010, 2012) 71 Słowi´ nskie Błota 54.3646 16.4902 30 Ombrotrophic 7 1 7 Pb LOI F. De Vleeschouwer po slowbl De Vleeschouwer et al. (2009) 72 Walton Moss 54.9833 −2.6670 109 Ombrotrophic 2 1 30 C D. Mauquoy en walton Mauquoy et al. (2002) 73 D¨ urres Maar 50.1228 6.8722 455 Ombrotrophic 3–4 2 5 50% R. Moschen de durrma Moschen et al. (2009) 74 Irwin Smith Bog 45.0319 −83.6181 223 Ombrotrophic 3 1 3 A 50% R. Booth us irwsmi Unpublished data 75 Sidnaw Bog 46.5618 −88.7824 400 Minerotrophic 3 1 5 Pb, A 50% R. Booth us sidnaw Unpublished data 76 Pinhook Bog 41.6152 −86.8468 245 Minerotrophic 3 1 5 A 50% R. Booth us pinhoo Unpublished data 77 Colesdalen 78.0925 14.9787 8 Minerotrophic 1.2 0.5 9 50% D. Beilman no colesd Unpublished data 78 Fenvyestet¨ o 47.6670 24.0330 1340 Ombrotrophic 45 1 16 50% D. Charman ro fenvye Unpublished data 79 Saxn¨ as Mosse 56.8558 13.4610 161 Ombrotrophic 10.5 1 25 C M. van der Linden se saxnas van der Linden (2006) 80 Dead Island Bog 54.8875 −6.5491 41 Ombrotrophic 20 1 0 S, T(5) 50% G. Swindles ni deasis Swindles et al. (2010) 81 Sib-04 56.8039 78.7369 131 Ombrotrophic 2–3 1 5 LOI D. Beilman SIB04 Beilman et al. (2009) 82 The Great Heath 44.7013 −67.8092 75 Ombrotrophic 3.5 1 8 50% J. Nichols us greath Unpublished data 83 Kontolanrahka 60.7830 22.7830 87 Ombrotrophic 125–320 5 18 C M. V¨ aliranta fi kontol V¨ aliranta et al. (2007) 84 Lappmyran 64.1647 19.5830 295 Ombrotrophic 10.5 1 20 C M. van der Linden se lappmy van der Linden et al. (2008) 85 Outokumpu 62.6667 28.8500 108 Ombrotrophic 3 1 4 Pb C N. Rausch fi outoku Rausch et al. (2005), Ukonmaanaho et al. (2006) 86 Lac Le Caron Central 52.2897 −75.4331 254 Ombrotrophic 1 1 3 LOI S. van Bellen ca llcen van Bellen et al. (2011) 87 Hietaj¨ arvi 63.1500 30.6667 168 Ombrotrophic 3 1 3 Pb C N. Rausch fi hietaj Rausch et al. (2005), Ukonmaanaho et al. (2006) 88 Mariana Lake 56.0167 −111.9333 700 Ombrotrophic 0.8 1 6 LOI Z. C. Yu ca marian Unpublished data 89 Harjavalta 61.3500 22.1833 40 Ombrotrophic 3 1 4 Pb C N. Rausch fi harjav Rausch et al. (2005), Ukonmaanaho et al. (2006) 90 Nordan’s Pond bog 49.1500 −53.5830 60 Ombrotrophic 5 2 2 Pb 50% P. Hughes ca nordan Hughes et al. (2006) Biogeosciences, 10, 929–944, 2013 www.biogeosciences.net/10/929/2013/ D. J. Charman et al.: Climate-related changes in peatland carbon accumulation 935 b a Fig. 2. Age–depth and carbon accumulation estimates for individual profiles. The example shown here is Misten Bog, Belgium. (a) An age– depth plot showing the calibrated ages (blue shapes) and age–depth model (grey-scale graph) (Blaauw and Christen, 2005, 2011). (b) Carbon accumulation derived from age–depth models, and bulk density and C measurements. Curves are fitted to each of 10000 possible age models based on Bayesian analysis. Points represent individual samples on different age models, and grey lines are fitted curves for individual models. Only curves fitted incorporating long-term decay and ecosystem maturity (Yu et al., 2003) are shown here. All age depth models were converted to carbon accumulation using bulk density and carbon or LOI measurements. We derived different estimates of variability in carbon accumulation rates based on different assumptions about autogenic processes of long-term decay (Clymo, 1984) and ecosystem maturity (Yu et al., 2003). Carbon accumulation rates calculated from our age–depth models and carbon density do not take account of autogenic peat accumulation processes, most importantly the effect of long-term decay. Dead plant material decays rapidly in the surface layers, as the most labile organic matter is broken down quickly by microbial activity. Decomposition rates are much slower (though not zero) in the permanently saturated zone, which contains more recalcitrant organic matter (Clymo, 1984; Belyea and Baird, 2006). If productivity and decay are constant, measured apparent accumulation rates will be higher for more recent peat, and the long-term carbon storage will appear to increase. We accounted for this ecological process by fitting decay curves to each profile (Clymo, 1984). We also tested the effect of “ecosystem maturity”, that is the slowing of peat growth under stable conditions because of autogenic limits on the height of the peat surface (Yu et al., 2003). We excluded carbon accumulation changes in the uppermost peat (conservatively approximated here as peat formed after 1850) where relatively rapid aerobic decay is still taking place. We used AD 1850 for this because this is likely outside of the aerobic decay zone for all cores. The changes in accumulation rates for each site were expressed as differences between observed accumulation and those derived from three models: (1) linear decay model (i.e. no autogenic processes); (2) the Clymo model, which includes long-term decay only (Clymo, 1984): M=Pc ac(1−e−act), (1) where Mis the accumulated carbon, Pcis the peat added to the catotelm each year (gCcm−2),acis the catotelm decay constant and tis time; and (3) the extended peat accumulation rate (ExtPAR) model (Yu et al., 2003), which includes long-term decay and ecosystem maturity: M=Pc ac−bc(e−bct−e−act), (2) where the parameters are the same as those listed above, with the addition of bc, a coefficient that allows the accumulation rate to be modified. Each curve fitting exercise produces estimated values for Pc,acand bc. There are several other more complex models that could be applied to account for longterm decay, but it is often difficult or impossible to determine the most appropriate one, given the subtle variations in the carbon accumulation curves (Belyea and Baird, 2006). Our intention here is to test whether observed variations in the raw carbon accumulation data could be explained by longterm decay and ecosystem maturity. Decay models were fit individually to each carbon accumulation curve derived from the Bacon routine. For the linear model, the fitting was carried out using ordinary leastsquares. For the Clymo and ExtPAR models, optimization was carried out using an iterative orthogonal sampling technique that samples the entire parameter space, then uses a least squares fit to obtain a subset of the parameter space. This subset is then sampled in the next iteration to produce www.biogeosciences.net/10/929/2013/ Biogeosciences, 10, 929–944, 2013 936 D. J. Charman et al.: Climate-related changes in peatland carbon accumulation Fig. 3. Steps involved in deriving a non-autogenic accumulation curve from a single age profile after fitting of the peat accumulation model with long-term decay and ecosystem maturity (Yu et al., 2003). See text for details. an increasingly well defined parameter space. All following analyses were applied to results from all three versions (Fig. 3). The number of age models varied by over an order of magnitude between sites (from ca. 2300 to ca. 44000). To avoid biases towards sites with a higher number of age models, a single age model was randomly sampled from each site. The accumulation rates were interpolated onto a regular time step by taking the median value in a moving window (halfwindow of 25yr) with a step of 10yr to avoid bias toward sites with higher sampling resolution. Finally, a time series of accumulation rates was calculated as the median of the 24 interpolated accumulation rates (one per site). This was repeated 10000 times to provide a fair sampling of the available age models, and gave a matrix of median time series on a regular time step. Finally, this matrix was used to calculate the median and percentile values of the accumulation rates for each time. The effect of this Monte Carlo resampling of the possible age models (and associated accumulation rate curves) is to give greater weight to the sites with the best-constrained chronologies. In each iteration of the resampling, we took one age model and set of accumulation rates from each site. Sites that are well-constrained will provide age models that are similar in each iteration, and poorly constrained sites will provide age models that are widely different. The end result of this is that well-constrained sites will effectively have a greater weight in the overall composite. To avoid any bias toward sites with generally very high accumulation rates, a second composite was made, based on transformed values. This followed the methodology used previously for charcoal data (Marlon et al., 2008): (1) minimax transformation of the original accumulation rate time series; (2) Box–Cox transformation to normalise the time series; and (3) z-score calculation. The composite z-scores were estimated using the same procedure as for compilation of the untransformed values described above. The final composite curves are shown in Fig. 4. 3 Results and discussion 3.1 Spatial relationships between carbon accumulation and climate Warming would be expected to increase net primary productivity (NPP) in high-latitude ecosystems because of increased growing season length. The growing season for northern peatlands is appropriately defined as the period of the year with air temperatures above freezing, because bryophytes begin photosynthesis at this threshold, and are the dominant peat-former in most of our sites. PAR, determined by latitude and cloudiness, is the driver of photosynthetic carbon fixation and may also be an important control on NPP. However, higher temperatures could also increase peat decomposition rates through accelerated microbial activity (Ise et al., 2008; Dorrepaal et al., 2009). Linear regression of total carbon accumulated over the last 1000yr (C) against PAR0 yielded the strongest relationship: C=0.0055 PAR0−3.82,(3) with an R2of 0.33 (Fig. 5a). In single-predictor regressions, C showed a weaker relationship with GDD0 (R2=0.13, Fig. 5b) and no significant relationship with P/Eq (P=0.19, Fig. 5c). Residuals from Eq. (3) showed no systematic relation to either GDD0 or P/Eq and inclusion of these additional predictors in a multiple linear regression yielded nonsignificant regression coefficients. The correlation between PAR0 and GDD0 is high (0.83), owing to the growing season length that is shared by both variables. We checked the influence of two apparent outliers with higher PAR0 values on our conclusions. These are the two southernmost sites from Dhakuri (India) and Pinhook (USA). Removing these two sites does not affect the significance of the relationship between PAR0 and 1kaC (P < 0.0001) but changes the R2 values from 0.33 to 0.24 and slightly changes the slope from 0.0055 to 0.0049. Thus, it still explains more of the variation than GDD0. The influence of these two sites is not insignificant, but removing them does not impact our main conclusions concerning PAR. Without the two “outliers” total C still shows a positive significant relationship (P < 0.001) with GDD0 but with a change in R2from 0.18 to 0.13 and a Biogeosciences, 10, 929–944, 2013 www.biogeosciences.net/10/929/2013/ D. J. Charman et al.: Climate-related changes in peatland carbon accumulation 937 05−95% 10−90% 15−85% 20−80% 25−75% 30−70% 35−65% 40−60% 45−55% 1000 1200 1400 1600 1800 2000 Years g m−2 yr−1 15 10 5 0 Non−autogenic C Accumulation Rate (Linear model) a 05−95% 10−90% 15−85% 20−80% 25−75% 30−70% 35−65% 40−60% 45−55% 1000 1200 1400 1600 1800 2000 Years z–score 0.4 0.2 0 -0.2 -0.4 Non−autogenic C Accumulation Rate Z−scores (Linear model) b 05−95% 10−90% 15−85% 20−80% 25−75% 30−70% 35−65% 40−60% 45−55% 1000 1200 1400 1600 1800 2000 Years g m−2 yr−1 15 10 5 0 Non−autogenic C Accumulation Rate (Clymo model) 05−95% 10−90% 15−85% 20−80% 25−75% 30−70% 35−65% 40−60% 45−55% 1000 1200 1400 1600 1800 2000 Years z–score 0.4 0.2 0 -0.2 -0.4 Non−autogenic C Accumulation Rate Z−scores (Clymo model) c 05−95% 10−90% 15−85% 20−80% 25−75% 30−70% 35−65% 40−60% 45−55% 1000 1200 1400 1600 1800 2000 Years g m−2 yr−1 15 10 5 0 Non−autogenic C Accumulation Rate (ExtPAR model) 05−95% 10−90% 15−85% 20−80% 25−75% 30−70% 35−65% 40−60% 45−55% 1000 1200 1400 1600 1800 2000 Years z–score 0.4 0.2 0 -0.2 -0.4 Non−autogenic C Accumulation Rate Z−scores (ExtPAR model) Fig. 4. Composite carbon accumulation curves for the last millennium based on different assumptions concerning autogenic processes. Left panels show untransformed data, right panels show z-scores. (a) Linear accumulation without considering autogenic processes, (b) with long-term decay rates, and (c) including long-term decay and ecosystem maturity (left panel also shown in Fig. 6). The data after AD 1850 are shown only in outline because the apparent upturn in carbon accumulation is due to incomplete decay of recently accumulated organic materials. b Growing degree days above 0°C 0 10 20 30 40 50 60 70 0 1,000 2,000 3,000 4,000 0.22 0.12 0.15 c Precipitation / equilibrium evapotranspiration 0 10 20 30 40 50 60 70 0 1 2 43 5 6 C accumulated (kg m-2) a Photosynthetically active radiation days above 0°C (mol photons m-2 season-1) 0 10 20 30 40 50 60 70 2,000 4,000 6,000 8,000 10,000 0.48 0.33 0.25 Fig. 5. Relationships between climate variables and peat carbon accumulation. The total carbon accumulated over the last 1000yr (1ka) at each site compared to PAR0 (a), GDD0 (b) and the ratio of precipitation to equilibrium evapotranspiration (c). Bog (ombrotrophic) and fen (minerotrophic) sites (see Tables 1 and 2) are shown in blue and green, respectively, and separate regressions (R2values are shown) have been calculated for each site type, with a black regression line for the merged data sets. Vertical error bars represent chronological uncertainties (2σ) in estimating AD 1000 in each profile (n=90). www.biogeosciences.net/10/929/2013/ Biogeosciences, 10, 929–944, 2013 944 D. J. Charman et al.: Climate-related changes in peatland carbon accumulation van Bellen, S., Dallaire, P.-L., Garneau, M. and Bergeron, Y.: Quantifying spatial and temporal Holocene carbon accumulation in ombrotrophic peatlands of the Eastmain region, Quebec, Canada, Global Biogeochem. Cy., 25, GB2016, doi:10.1029/2010GB003877, 2011. van der Linden, M. and van Geel, B.: Late Holocene climate change and human impact recorded in a South Swedish ombrotrophic peat bog, Palaeogeogr. Palaeoecol., 240, 649–667, 2006. van der Linden, M., Barke, J., Vickery, E., Charman, D. J., and van Geel, B.: Late Holocene human impact and climate change recorded in a north Swedish peat deposit, Palaeogeogr. Palaeoecol., 258, 1–27, 2008. Vitt, D. H., Halsey, L. A., Bauer, I. E., and Campbell, C.: Spatial and temporal trends in carbon storage of peatlands of continental western Canada through the Holocene, Can. J. Earth Sci., 37, 683–693, 2000. Wania, R., Ross, I., and Prentice, I. C.: Integrating peatlands and permafrost into a dynamic global vegetation model: 1. Evaluation and sensitivity of physical land surface processes, Global Biogeochem. Cy., 23, GB3014, doi:10.1029/2008GB003412, 2009. Young, G. H. F., McCarroll, D., Loader, N. J., and Kirchhefer, A. J.: A 500-year record of summer near-ground solar radiation from tree-ring stable carbon isotopes, Holocene, 20, 315–324, 2010. Yu, Z. C.: Holocene carbon flux histories of the world’s peatlands: Global carbon-cycle implications, Holocene, 21, 761–774, 2011. Yu, Z. C., Vitt, D. H., Campbell, I. D., and Apps, M. J.: Understanding Holocene peat accumulation pattern of continental fens in western Canada, Can. J. Bot., 81, 267–282, 2003. Yu, Z., Beilman, D. W., and Jones, M. C.: Sensitivity of northern peatland carbon dynamics to Holocene climate change, in: Carbon cycling in northern peatlands, edited by: Baird, A. J., Belyea, L. R., Comas, X., Reeve, A., and Slater, L., American Geophysical Union, Washington DC, Geoph. Monog. Series, 184, 55–69, doi:10.1029/2008GM000822, 2009. Yu, Z. C., Loisel, J., Brosseau, D. P., Beilman, D. W., and Hunt, S. J.: Global peatland dynamics since the Last Glacial Maximum, Geophys. Res. Lett., 37, L13402, doi:10.1029/2010GL043584, 2010. Biogeosciences, 10, 929–944, 2013 www.biogeosciences.net/10/929/2013/