Steve Frolking,1* Nigel T. Roulet,2,3 Tim R. Moore,2,3
Pierre J. H. Richard,3,4 Martin Lavoie,4 and Serge D. Muller3,4
1Institute for the Study of Earth, Oceans, and Space, University of New Hampshire, Durham, New Hampshire 03824, USA;2Department of Geography, McGill University, 805 Sherbrooke Street West, Montreal, Quebec, H3A 2K6, Canada; 3Centre forClimate and Global Change Research, McGill University, 805 Sherbrooke Street West, Montreal, Quebec, H3A 2K6, Canada;
4Department de Geographie, Universite de Montreal, C.P. 6128, Montreal, Quebec, H3C 3J7, Canada
ABSTRACTTo test the hypothesis that long-term peat accumula-tion is related to contemporary carbon flux dynamics,we present the Peat Decomposition Model (PDM), anew model of long-term peat accumulation. Decom-position rates of the deeper peat are directly related toobservable decomposition rates of fresh vegetationlitter. Plant root effects (subsurface oxygenation andfresh litter inputs) are included. PDM considers twovegetation types, vascular and nonvascular, with dif-ferent decomposition rates and aboveground and be-lowground litter input rates. We used PDM to inves-tigate the sensitivities of peat accumulation in bogsand fens to productivity, root:shoot ratio, tissue de-composability, root and water table depths, and cli-mate. Warmer and wetter conditions are more con-ducive to peat accumulation. Bogs are more sensitivethan fens to climate conditions. Cooler and drier con-ditions lead to the lowest peat accumulation whenproductivity is more temperature sensitive than de-composition rates. We also compare peat age–depthprofiles to field data. With a very general parameter-
ization, PDM fen and bog age–depth profiles weresimilar to data from the the most recent 5000 years atthree bog cores and a fen core in eastern Canada, butthey overestimated accumulation at three other bogcores in that region. The model cannot reliably predictthe amount of fen peat remaining from the first fewmillennia of a peatland’s development. This discrep-ancy may relate to nonanalogue, early postglacial cli-matic and nutrient conditions for rich-fen peat accu-mulation and to the fate of this fen peat material,which is overlain by a bog as the peatland evolves, acommon hydroseral succession in northern peatlands.Because PDM sensitivity tests point to these possiblefactors, we conclude that the static model represents aframework that shows a consistent relationship be-tween contemporary productivity and fresh-tissue de-composition rates and observed long-term peat accu-mulation.
Peatland ecosystems accumulate carbon becauseannual net primary productivity (NPP) of the peat-land vegetation generally exceeds the annual de-
composition of litter and peat. Relative to otherecosystems, northern peatlands have low rates ofNPP (Thormann and Bayley 1997), decomposition(Brinson and others 1981; Bartsch and Moore1985; Johnson and Damman 1993; Belyea 1996;T. R. Moore unpublished), and net carbon dioxide(CO2) exchange (Frolking and others 1998), butover millennia NPP has been greater than decom-position. Hence, northern peatlands have been a
Received 19 June 2000; accepted 24 January 2001.Current address for M. Lavoie: Centre d’Etudes Nordiques, Universite Laval,Pavillon Abitibi-Price, Sainte-Foy, Quebec, G1K 7P4, Canada*Corresponding author; e-mail: [email protected]
persistent sink for CO2, averaging 0.02 to 0.03 kgCO2-C m22 y21 over the past 5000–10,000 years(Gorham 1995; Tolonen and others 1992). This hasresulted in 200–450 Pg C being sequestered inabout 3.5 million km2 of northern peatlands(Gorham 1991). Slow decomposition rates innorthern peatlands result from the combined effectsof limited oxygen diffusion into saturated peat lead-ing to anoxic conditions for a large portion of thepeat profile (Clymo 1992); the inherent resistanceto decomposition of some peatland vegetation tis-sues, particularly Sphagna (see, for example, John-son and Damman 1993; Hogg 1993); and generallycool temperatures of peat (Puranen and others1999). Clymo (1984) hypothesized that the accu-mulation of peat has a theoretical limit becauseeven very low decomposition rates applied to anever-increasing mass of peat will eventually ap-proach and equal NPP. The rate of decomposition inthe anoxic zone of the peatland and the long-termNPP ultimately control the time it takes for a peat-land to grow and the final mass of carbon stored ina peatland.
In Clymo’s peat models (Clymo 1984, 1992), aconstant water table depth divides the peat profileinto a surface oxic zone (or acrotelm), where fasteraerobic decomposition pathways dominate, and adeeper anoxic zone (or catotelm), where slower,anaerobic decomposition occurs. These peat accu-mulation models do not explicitly simulate the massbalance of the surface, oxic peat; instead they aremodels of catotelm peat accumulation and assumethat the acrotelm is a constant mass that “floats” ontop of the accumulating catotelm. Litter mass lossduring passage through the oxic zone is typically80%–90% (Clymo 1992); therefore, the input ofmass to the anoxic zone would be 10%–20% oftotal vegetation litter production. The anoxic zoneor catotelm peat decomposes very slowly, atroughly 0.1% of the oxic zone rate (Belyea andClymo 1999). In one formulation (Clymo 1984),the input and decomposition rates are assumed tobe constant, and the accumulation of peat in theanoxic zone is modeled as
dM
dt5 P* 2 kM (1)
where M is the peat mass, P* is the input to theanoxic zone, and k is the anoxic zone decomposi-tion rate. Peat continues to accumulate at an ever-declining rate until, barring major disturbances, iteventually reaches steady state:
M~t! 5P*
k~1–e2kt!O¡
t3 `
P*
k(2)
The mass, M, is the catotelm mass, which is usuallya large majority of the total peat. Values for the twoparameters P* and k are not directly derived fromnor comparable to any measurements of peatlandvegetation productivity or fresh-tissue decomposi-tion. Instead they are empirically determined byfitting this equation to age–depth profile data ofcatotelm peat (Clymo 1984, 1992; Clymo and oth-ers 1998).
The Peat Decomposition Model (PDM) was de-veloped with the primary objective of testing thehypothesis that long-term peat accumulation isconsistent with observed rates of vegetation pro-ductivity and fresh-tissue decomposition. Like themodel of Clymo (1984), PDM is a static modelsimulating long-term peat accumulation and age–depth profiles under constant conditions (NPP andwater table depth). We extend the work of Clymoby modeling the complete acrotelm/catotelm peatprofile. Decomposition rates down the peat profileare directly linked to observable initial mass-lossrates of fresh peat litter tissue, which have beenmeasured for numerous peatland plant tissues (forexample, see Belyea 1996; Johnson and Damman1993; Hogg 1993; Bartsch and Moore 1985). Massinput rates for the model are equal to observablelitter production of peatland vegetation. We alsodirectly simulate the effects of plant roots, whichdeliver both fresh litter and oxygen some distancedown the peat profile. Finally, PDM considers twovegetation types, vascular and nonvascular, withdifferent initial decomposition rates and differentaboveground and belowground litter input ratesthat can thus be used to explore the influence ofvegetation type on peat accumulation. The inclu-sion of both roots and two vegetation types allowsthe model to simulate both bogs (with fewer rootinputs generally above the mean water table and,typically, a higher fraction of moss inputs) and fens(with more root litter inputs going deeper than themean water table and, typically, a lower fraction ofmoss inputs). Our objective in this initial modeldevelopment is to explore how well such a simplestatic model can produce reasonable peat age–depthprofiles without curve fitting, using parameters rep-resenting mean peatland characteristics, based onfield and laboratory measurements. We use sensi-tivity analyses and comparisons with a selection offield-based age–depth profiles with known individ-ual paleoecological histories.
480 Frolking and others
MODEL
Most models of soil organic matter dynamics aggre-gate the accumulated organic matter into one orseveral pools. The pools can have characteristic andconstant decay rates, or they can have a time-vary-ing decay rate that is a function either of time orsome characteristic of the organic matter pool, suchas its lignin–nitrogen ratio (for example, see reviewby Paustian and others 1997). Agren and Bosatta(1996b) have developed a theoretical approach inwhich the dynamics of litter quality (related to itssusceptibility to decomposition) and soil carbon arelinked by coupled differential equations throughthe activity of the decomposer community. Undercertain simplifying assumptions, their model can besolved analytically for the mean quality of an ag-gregated continuum of litter cohorts of varyingquality, and this quality can be related to chemicalfractionations of the organic matter (Agren and Bo-satta 1996a).
There are two characteristics of peatlands, how-ever, that favor an explicit modeling of individualcohorts. First, accumulating organic matter in apeatland develops a well-defined stratigraphy, withminimal bioturbation, so age–depth profiles aregenerally monotonic, with the oldest peat at thebase of the profile. Second, there is a very steep andrelatively stable gradient in the conditions for de-composition because the deeper peat is continu-ously saturated, while the shallowest peat is oftendrained. In a typical northern bog, the top 0.2–0.4m will be oxic for much of the year, whereas thepeat below about 0.4–0.6 m will be anoxic. There-fore, there is a strong correlation between cohortage (and thus degree of decomposition) and theconditions to which the cohort is exposed. Becauseof this, we chose to develop an annual cohortmodel of decomposition in which the fate of eachyear’s litter input (a cohort) is tracked as the peatcontinues to accumulate above it.
PDM operates at an annual time step, withaboveground litter input deposited as an annuallitter cohort at the top of the existing peat profile.Root litter inputs also occur every year and areadded to upper cohorts, down to the bottom of therooting zone. These amended cohorts now containlitter of multiple ages, mixing older surface litterwith younger root litter, and hereafter are called“peat layers”. The vegetation is assumed to be insteady state, so aboveground litter productionequals aboveground NPP, and root litter productionequals root NPP. Annual peat layers in the rootingzone lose mass through decomposition, but theyalso gain mass from the root litter input. All peat
layers below the rooting zone only lose mass. PDMvariables for each peat layer are mass, bulk density,depth, age, and decomposability. PDM was devel-oped assuming constant NPP. Constant initial de-composition rates were used for each litter type,and these rates decreased as a simple function ofpeat layer mass loss, representing an increasing re-calcitrance or decreasing quality (Agren and Bosatta1996a, b) of the remaining mass of the peat layer asit moves deeper into the peat. Decomposition ratesare subsequently modified to reflect the influenceof the soil climate at the depth of each peat layer.These simplifying assumptions permit an analyticalsolution of the equations of state of the peat layermass and thickness, which can be solved iterativelyfrom the top to the bottom of the peat profile. Inputvariables are aboveground and belowground NPP,water table and rooting depths, and the peatlandage (Table 1), all based on field observables. Modelparameters are the initial decomposability of thelitter types, degree of anaerobicity at depth, andeffect of temperature with depth (Table 1), again allbased on field and laboratory measurements. Themodel has no free parameters to be used to fitage–depth profile data; hence, the model is notcalibrated. PDM output is the mass and decompos-ability of peat for each age peat layer. Using a pre-scribed density profile, PDM can calculate peat layerthickness. The mass and thickness of each peat layercan be summed to obtain total mass and depth ofaccumulated peat or used to generate age–depthprofiles.
Decomposition
A general formulation for decomposition of a massof organic matter, m, is
dm~t!
dt5 2k0m0 Sm~t!
m0Da
(3)
where m0 is the initial mass and k0 is the decompo-sition or mass loss rate at t 5 0. If the parameter aequals 1.0, this simplifies to simple exponential de-cay, as in the second term in Eq. (1). Otherwise, ageneral solution is given by
m~t! 5m0
@1 1 ~a 2 1!k0t#1/~a21! (4)
Clymo and others (1998) have shown that whenfitting peat core data from 310 bogs and fens fromsouthern Finland, they can get roughly equivalentgoodness of fit with a 5 1, 2, or 3. Because there isstrong evidence that the fractional loss rate of adecomposing tissue declines with time (Heal and
Peat Decomposition and Accumulation 481
others 1997), a should be greater than 1. For sim-plicity we have chosen a 5 2, so the model has asolution of the peat layer mass as a function of time:
m~t! 5m0
1 1 k0t(5)
and an effective decomposition rate of
k~t! 5 k0
m~t!
m05
k0
1 1 k0t(6)
Note that after 1 year of decomposition m(1) 5m0/(1 1 k0). In the first few years of decomposi-tion, this function is similar to the simple exponen-tial decay, but the decomposition rate slows as de-composition progresses (that is, it slows with time).If this formulation is combined with a constantinput (as in the first term in Eq. [1]), it would lead
logarithmically to an infinite accumulation of peat,given infinite time. This time-dependent functionalform of the decomposition rate is the same form asthe decline in litter quality with time in one repre-sentation of the more general decomposition modelof Agren and Bosatta (1996b) with a 5 2.
Litterbag decomposition studies provided initialmass loss rates, k0, for various peatland tissue types.Because litterbag studies are typically of only a fewyears duration, the litterbags are exposed to soilclimate conditions representative of near-surfaceconditions. Deeper in the peat, decomposition ratesshould reflect the confounding factors of coolertemperatures and saturation of the peat leading toanoxic conditions that do not occur at the peatsurface. These additional effects are modeled asmultiplicative factors, and actual decomposition ofa peat layer is modeled as
Table 1. Model Parameters and Values for Base Case Scenarios
ms,b Moss surface litter input kg m22 y21 0.2 0.1 11 1 Moore 1989k0,b Moss initial mass loss rate y21 0.05 0.08 2 2 T.R. Moore unpublishedZWT Mean water table depth m 0.3 0.05 22 0 Ingram 1983; Verry and
others 1988; Siegel 1993Zanox Thickness of oxic–anoxic
transition zone belowroot zone
m 0.05 0.0 22 n.a. Lahde 1967
fanox Anaerobic:aerobicdecomposition ratio
— 0.025 0.1 2 22 Scanlon and Moore 2000
DT Reduction in rate ofdecomposition due totemperature profile
— 0.2 0.2 0 0 P. Lafleur unpublished
A Peat age y 8000 8000 1 1 —P0 Bulk density at surface
11, strong positive correlation; 1, moderate positive correlation; 0, weak correlation; 2, weak negative correlation; 22, strong negative correlation; n.a., not applicable*These default values are not meant to represent any specific peatland, but rather to be generally representative of these two broad peatland classes bog and fen.**Sensitivities are for changes in each parameter value of up to 6 30%, with other parameters held constant.
482 Frolking and others
dm~t!
dt5 2k0
m~t!2
m0z f~T! z f~W! (7)
We assumed that the annual effect of the soil cli-mate multipliers f(T) and f(W) were constant fromyear to year but varied with depth, z, with both f(T)5 fT(z) and f(W) 5 fw(z) equal to 1.0 at the surface.The temperature modifier, f(T), is based on an ob-served peatland temperature profile and an as-sumed and simple temperature effect (Q10 5 2) (seeScanlon and Moore 2000). We applied this functionto monthly means of hourly soil temperature mea-surements from Mer Bleue bog near Ottawa, Can-ada (Lafleur and others 2001) to calculate a generalmonthly modifier, then averaged these for the year,and normalized the values to the surface value. Wethen fit (by eye) a simple exponential curve to thesedata to get a decomposition temperature modifierthat was only a function of depth (Figure 1a).
The moisture modifier, f(W), is set to one for thesurface oxic zone and decreases below the watertable (Figure 1b). For bogs, which have a deeperwater table, we assume that there is a rapid lineartransition between oxic and anoxic conditionsjust below the mean long-term water table depth.For fens, which have a shallow water table and adeeper rooting depth, we assume that there is alinear decline in oxic status from the mean watertable depth to the bottom of the rooting zone. Forall of the following discussion, decompositionrates are implicitly assumed to be multiplied byfT(z) and fW(z), and these terms are left out of theequations for simplicity (that is, k z fT(z) z fw(z) 3k). A simple bulk density profile (Figure 1c) wasbased on bulk density data from two bogs insoutheastern Canada (P. J. H. Richard and S. D.Muller unpublished).
Figure 1. (a) Mean annual temperature effect on decom-position, fT(z), as a function of depth. Monthly average
soil temperatures were calculated from hourly averagetemperature profile data recorded continuously at MerBleue bog, near Ottawa, Ontario (Lafleur and others2001) and used to calculate a monthly temperature mul-tiplier (5 2T#/10), which was then normalized so that thesurface value was 1.0. The curve fit is an exponentialdecay with depth, having an asymptotic value of 0.8 andan initial value of 1.0. (b) Modeled effect of peat watertable and fen roots on fanox. This multiplier on decompo-sition rates accounts for the effect of anoxicity. (c) Shapeof model’s peat bulk density profile. Surface and deeppeat values based on references in Table 1; shape basedon unpublished data from P. J. H. Richard.
Peat Decomposition and Accumulation 483
Root Inputs
Most peatland plants (except bryophytes) havelarge root:shoot ratios (see, for example, Wallen1986; Saarinen 1996; Moore and others 2001);thus, some fraction of their annual litter inputinto the peat will be root litter. To develop alitter/peat profile with root inputs, we consideredthe analytical solution to steady-state conditions(as above) but incorporated the effect of roots.We made the following two simplifying assump-tions: (a) Root decomposability equals that ofaboveground tissue of the same plant type, and(b) root density and root litter input rates areuniform to the rooting depth and zero below thatdepth. This simplified rectangular root distribu-tion is a rough approximation to the typical pro-file (for example, see Wallen 1986). The modelrooting depth input variable, Zr, thus representsthe depth to which the bulk of the root turnovertakes place but not the maximum observablerooting depth.
Annual peat layers get a fresh input (root litter)every year until they are buried below the rootingzone. Adding this fresh litter increases the peatlayer’s net decomposability because fresh litterdecomposability is always greater than partiallydecomposed peat layer’s decomposability (see Eq.[6]). Consider a peat layer of mass m* (the massremaining from all previous inputs, includingboth surface litter and shallower roots, with totalprevious inputs equal to M) and decomposabilityk*, receiving a root litter input of mr with decom-posability k0. The new peat layer mass will be m95 m* 1 mr. PDM updates the peat layer’s decom-posability, k9, with the following equation:
k9 5 k0
m* 1 mr
M 1 mr(8)
which gave a better approximation to the exactsolution of tracking all root litter cohorts indepen-dently than did a mass-weighted average (k9 5 (m*k* 1 mrk0)/(m* 1 mr)).
Calculating the Litter/Peat Profile
Because peatland plant tissues decompose at dif-ferent rates, we generalized the solution above totwo tissue types that do not interact—vascular,rooted vegetation (tissue type a with initial de-composition rate k0,a) and bryophyte (rootless)vegetation (tissue type b with initial decomposi-tion rate k0,b). Let ms,a be the surface vascularlitter input, mr,a be its root litter input (NPPa 5m0,a 5 ms,a 1 mr,a), Zr be the rooting depth, ba
(5mr,a/Zr) be the root input per unit depth, ms,b bethe surface litter input for the nonvascular vege-tation (NPPb 5 m0,b 5 ms,b), and zi be the thick-ness and ri the bulk density of the ith peat layer.Then the surface peat layer (i 5 0) can be de-scribed by the following set of equations:
5m0 5 m0,a 1 m0,b 5 ms,a 1 baz0 1 ms,b
z0 5m0
r0
k0,a 5 k0,a
k0,b 5 k0,b
(9)
Combining the equations for m0 and z0, this can besolved as
H z0 5ms,a 1 ms,b
r0S 1
1 2 ba/r0D5 ~ms,a 1 ms,b!S 1
r0 2 baD
m0 5 z0r0 (10)
The 1-year-old peat layer’s mass will equal the massremaining from the surface peat layer after 1 year ofdecomposition plus the fresh root litter input intothe 1-year-old layer. The peat layer can be de-scribed by a similar set of equations, as follows:
5m1 5
m0,a
1 1 k0,a1 baz1 1
m0,b
1 1 k0,b
z1 5m1
r1
k1,a 5 k0,a
m1,a
m0,a 1 baz1
k1,b 5 k0,b
m1,b
m0,b
(11)
These can be solved in a similar manner:
5 z1 5 S m0,a
1 1 k0,a1
m0,b
1 1 k0,bDS 1
r1 2 baD
m1 5 z1r1
m1,b 5m0,b
1 1 k0,b
m1,a 5 m1 2 m1,b
(12)
This methodology can be used to construct the peatprofile down from the top in annual peat layers,and the ith peat layer will be described by
5zi 5 S mi21,a
1 1 ki21,a1
mi21,b
1 1 ki21,bDS 1
ri 2 baD
mi 5 ziri
ki,a 5 k0,a
mi,a
m0,a 1 baOj51
i zj
ki,b 5 k0,b
mi,b
m0,b
(13)
484 Frolking and others
and solved in the same manner. Once below therooting depth, the calculation proceeds as above,but without root inputs (that is, ba 5 0).
We characterized bogs by rooting depth equal tothe mean water table depth (for example, seeBackeus 1990) and fens by rooting depth deeperthan their characteristically shallow mean watertable depth (see, for example, Saarinen 1996). Wemodeled the oxic/anoxic effect for bogs as previ-ously described—that is, a linear drop in decompos-ability to a fixed factor (fanox) in a narrow bandbelow the water table. Because fens have livingroots that function below the water table, we con-sidered the fen root zone to be partially oxic andprescribed a linear decline in decomposition ratefrom the oxic rate at the fen water table to fanox, atthe bottom of the rooting zone. Because deep peatin bogs is more hydrologically isolated (Verry andothers 1988; Siegel 1993) and more acidic (Gorhamand Jannsens 1993) than deep peat in fens, weassigned bogs an fanox value one-fourth the fenvalue.
Aggregating the Annual Peat Layers
The total mass in a collection of peat layers betweenages T1 and T2 is given by
M1,2 5 E1
2
dm 5 ET1
T2
m~t! z dt 5 ET1
T2 m0
1 1 k0tz dt
5m0
k0z lnS1 1 k0T2
1 1 k0T1D (14)
where m is the mass per unit time interval, or thepeat layer mass when the time interval is 1 year.The mass loss rate of this aggregated peat layer canbe approximated by
k1,2 5
ET1
T2
m~t!k~t!dt
ET1
T2
m~t!dt
5 k0
S 1
1 1 k0T22
1
1 1 k0T1D
lnS1 1 k0T2
1 1 k0T1D
(15)
Because this integration over time (age) is also anintegration over some depth interval, this approxi-mation depends on the temperature and moisturefactors implicit in k being approximately constantover the depth interval. We used this aggregatedpeat layer approximation for deep peat (generallyolder than 500–1000 years) so the model did notneed to keep track of each annual peat layer for all
5000–12,000 years of the peat profile. For depthsfrom about 1–2 m, we aggregated the peat into10-year peat layers, typically about 1–2 cm thick.Below this, we aggregated the peat into 150-yearpeat layers, typically 1–10 cm thick. The mean de-composition rate (Eq. [15]) was not used to calcu-late the peat profile, but only to characterize thepeat layers’ decomposability.
RESULTS
Baseline Bog and Fen Scenarios
We developed very general parameterizations for abog and a fen from typical values reported in theliterature for peatland productivity, decompositionrates of various tissues, root biomass and depth, andtypical values for long-term mean water table depth(Table 1). Annual NPP for the bog scenario was 0.6kg m22 y21, with one-third from moss, one-thirdfrom aboveground vascular productivity, and one-third from belowground vascular productivity(Thormann and Bayley 1997; Backeus 1990). An-nual NPP for the fen scenario was 0.7 kg m22 y21,with 14% from moss, 43% from aboveground vas-cular productivity, and 43% from belowgroundvascular productivity (Thormann and Bayley 1997;Saarinen 1996). Throughout this paper, all massvalues are given as biomass, not carbon.
PDM generated depth profiles of peat mass, mossand vascular tissue fractions, decomposability, andage for the bog and fen baseline scenarios. The fenmaterial had a faster initial decomposition rate;thus, fen peat layer mass fell more rapidly withdepth (age) than bog peat layer mass (Figure 2a).However, root inputs for the fen were larger thanthose for the bog. Thus, below about 0.2 m, the fenpeat layer mass increased above that of the bog. Atthe transition to the fully anoxic zone (ZWT 1 Zanox
5 0.35 m for the bog, Zr 5 0.3 m for the fen), thebog peat layer mass was 0.059 kg m22 (10% ofannual surface plus root litter inputs), and the fenpeat layer mass was 0.17 kg m22 (24% of annualsurface plus root litter inputs). Below this, in thefully anoxic zone, bog peat layers lost mass moreslowly than fen peat layers. The effective decom-posability of each peat layer, calculated as the mass-weighted mean of the ki value for each tissue type,declined with depth (Figure 2b). For the bog peatlayers, there was a rapid drop in decomposability toanaerobic rates in the 0.05 m below the mean watertable depth. Decomposability continued to declinedown the catotelm profile due to continuing massloss (see Eq. [6]), and effective decomposability atthe base of the peat, approximately 0.0001 y21
Peat Decomposition and Accumulation 485
(Figure 2b) was around three orders of magnitudelower than the surface rate.
After 8000 years, fen peat thickness was 2.1 m(180 kg m22); whereas after 8000 years, bog peat
thickness was 3.2 m (290 kg m22) (Figure 3). Thesefigures imply mean accumulation rates of 0.022 kgm22 y21 for the fen and 0.036 kg m22 y21 for thebog. The model assumed constant conditions, sothat each additional year effectively added an addi-tional layer to the bottom of the profile, leading tocurrent accumulation rates of 0.006 kg m22 y21
(fen) and 0.02 kg m22 y21 (bog). Because of higherNPP rates and deeper roots, the fen peat age at adepth of 1 m was about 700 years, whereas the bogpeat age at 1 m was about 1100 years. Because fendecomposition rates were faster at depth, at depthsbelow 1.5 m bog peat was younger than fen peat(Figure 3).
The moss fraction of each peat layer was higherfor the bog than the fen (Figure 4), due to a higherpercentage of moss inputs, both at the surface andoverall (Table 1). Because moss initial decomposi-tion rates were slower than vascular tissue rates forboth bog and fen, the moss fraction increased withdepth (and peat layer age). However, from a depth
Figure 2. (a) Peat layer mass versus depth for the bog (x)and fen (o) base case scenarios. The inset panel highlightsthe top 0.6 m of the profile. Above 0.3 m (the bog watertable and the fen rooting depth), each symbol representsa single peat layer. (b) Peat layer effective decomposabil-ity vs depth (or age) for the bog (x) and fen (o) base casescenarios. The inset panel highlights the top 0.6 m of theprofile. Above 0.3 m (the bog water table and the fenrooting depth), each symbol represents a single peatlayer. The rapid shift in bog decomposability between0.3 m and 0.35 m is the transition from the oxic acrotelmto the anoxic catotelm. The fen scenario has a moregradual transition from the water table down to the bot-tom of the rooting zone.
Figure 3. Age depth profiles for bog and fen base scenar-ios and for a rich-fen scenario. Model parameters for bogand fen are in Table 1. Changes from the fen to rich-fenscenario were ms,a 5 0.35 kg m22 y21; mr,a 5 1.0 kg m22
y21, k0,a 5 0.38 y21, ms,b 5 0.15 kg m22 y21; and k0,b 50.07 y21. This rich-fen scenario was used for the deeperpeat in some of the model comparisons with core data.Although the bog and rich-fen scenarios accumulateabout the same amount of peat in 8000 years, the currentrate of accumulation is given by the slope of the curve at8000 years and is much greater for the bog than for eitherof the fen scenarios.
486 Frolking and others
of about 0.05–0.2 m for the bog and 0.1–0.3 m forthe fen, the moss fraction declined because of vas-cular root inputs in this zone.
Model SensitivitiesSensitivities to individual parameters. The sensitiv-
ity of peat accumulation to model parameters wasexplored by varying each parameter by 6 15%and 6 30% from its baseline value (Table 1). Pa-rameter sensitivities were approximately linear,ranging from strongly positive (a 30% change inthe parameter value caused more than a 25%change of the same sign in peat accumulation) toweak (a 30% change in a parameter value causedless than a 10% change in accumulated peat) tostrongly negative (a 30% change in the parametervalue caused more than a 25% change of oppositesign in peat accumulation). Both bog and fenshowed moderate positive sensitivity to changes invascular plant productivity (ms,a, mr,a). Sensitivity tomoss productivity (ms,b) was higher for bog than forfen because moss productivity was a greater frac-tion of total productivity and bog moss decompos-ability was slow (k0,b). As expected, the bog scenarioshowed moderate negative sensitivity to changes indecomposition rates (k0,a, k0,b, kanox); slower decom-position rates led to more peat accumulation. Fensensitivity to kanox was greater than bog sensitivity
because the base value for the fen was four timesgreater than that for the bog, so a fixed percentagechange was larger and had a greater effect.
Decreasing the surface peat bulk density (r0) in-creased the thickness of each and thus also buriedthe peat below the water table more quickly, lead-ing to increased accumulation. Bog sensitivity wasabout three times that of fen sensitivity, as the ratioof anoxic to oxic decomposition rates was smallerfor the bog than the fen scenario. Changing Dr, theincrease in bulk density with depth in the peat hadvery little effect on mass accumulation for eitherbog or fen because most of the change took placebelow 0.5 m, where conditions were relatively con-stant (see Figure 2). Changing Dr did modify thetotal depth of peat accumulated in a very predict-able way.
The bog scenario was very sensitive to the watertable depth (ZWT). A shallower water table led tosignificantly greater peat accumulation becausepeat layers reached the anoxic zone with more massintact. The fen scenario had little sensitivity to wa-ter table depth; as the change in decomposition ratefor the fen was much less abrupt at the water table,and the anoxic zone decomposition rate was gen-erally higher for the fen than for the bog scenario(see Figure 2b). Increasing rooting depth in the fendecreased accumulation as it extended the partiallyoxic zone deeper into the peat. Decreasing rootingdepth in the bog decreased accumulation becauseless fresh litter was input deeper in the profile,closer to the transition to anoxia, and thus peattransfer to the catotelm was smaller. Because Zanox
had a small magnitude (0.05 m), 15% and 30%changes were also small and bog peat accumulationsensitivity was small. In additional simulations withZanox at 0.0 m and 0.1 m, bog peat accumulationsshowed strong negative sensitivity.
Sensitivities to vegetation productivity. Using thebase case scenarios for bog and fen (Table 1), weadjusted the fraction of vascular plant litter inputthat was root tissue, keeping total vascular NPPconstant (0.6 kg m22 y21 for fens, 0.4 kg m22 y21
for bogs). As the root fraction increased, peat accu-mulation also increased because more litter wasinput closer to the anoxic zone and thus had lesstime to decompose aerobically (Figure 5a). For bothbog and fen, the sensitivity was stronger for lowroot fractional inputs and weaker if most vascularlitter was input as roots.
We also adjusted the base case bog and fen sce-narios by changing the moss fraction of total NPP,again holding total NPP and the vascularaboveground:belowground NPP ratio fixed. Be-cause moss litter had a lower decomposition rate
Figure 4. Fraction of peat layers that is moss (nonvascu-lar) peat for bog and fen base case scenarios. For most ofthe profile, this fraction increases with depth as mossdecomposability is less than that of vascular tissue. Nearthe surface this trend is reversed as significant root litterinputs from vascular plants decreases the moss fraction ofthe peat layer.
Peat Decomposition and Accumulation 487
than vascular litter as the fraction of total NPP dueto mosses increased, the total accumulated mass ofpeat also increased (Figure 5b). The bog scenariowas more sensitive to this ratio because the initial
decomposition rate of bog moss litter was lowerthan the rate for fen moss litter, although bothscenarios had the same initial decomposition ratefor vascular litter.
Figure 5. (a) Sensitivity of peat accumulation to fraction ofvascular litter input as roots for bog and fen base casescenarios. Total vascular NPP was unchanged (0.6 kg m22
y21 for the fen, 0.4 kg m22 y21 for the bog). (b) Sensitivityof peat accumulation to moss fraction of total NPP for bogand fen base case scenarios. Total NPP was unchanged (0.7kg m22 y21 for the fen, 0.6 kg m22 y21 for the bog). (c)Sensitivity of peat accumulation to total NPP (moss 1vascular). Moss:vascular and aboveground:belowground
productivity ratios were unchanged (see Table 1). (d) Sen-sitivity of peat accumulation to nutrient status. Improved(diminished) nutrient status was modeled by increasing (de-creasing) both NPP and decomposition rates by the samefraction. Moss:vascular and aboveground:belowground pro-ductivity ratios were unchanged (see Table 1). In eachpanel, the vertical lines represent baseline scenario values.Note the different scales on some panels.
488 Frolking and others
In a third test, we adjusted total productivity forboth the bog and fen scenarios (Figure 5c). Peataccumulation in the fen scenario had a linear re-sponse to total productivity, with a 10% increase inproductivity leading to a 12% increase in peat ac-cumulation. Peat accumulation in the bog scenariohad a stronger sensitivity to total productivity and aslightly nonlinear response.
Finally, to approximate changes in nutrient sta-tus, we adjusted both total productivity and decom-position rates up and down by the same propor-tions. Higher nitrogen and other nutrient inputleads to increased productivity (Aerts and others1992; Jauhiainen and others 1999) and to morereadily decomposable fresh litter (Johnson andDamman 1993; Verhoeven and Toth 1995); how-ever, there is some suggestion that chronic nitrogenloading may alter the decomposer community (Wil-liams and Silcock 1997; Gilbert and others 1998).Again, responses were roughly linear (Figure 5d)over a 6 30% range, with bog peat accumulationmore sensitive than fen peat accumulation.
Sensitivities to climate. To evaluate the model’ssensitivity to climate, we considered eight climatescenarios—(warmer, cooler, wetter, drier, warmer/wetter, warmer/drier, cooler/wetter, and cooler/drier)—and four scenarios of ecophysiological sen-sitivity to temperature—(low and high sensitivityfor both productivity and decomposition). Temper-ature changes were 6 3°C and moisture changeswere 6 25% in water table depth. Ecophysiologicalsensitivities were modeled by assuming an expo-nential response (low; Q10 5 2; high, Q10 5 3) forboth total productivity or litter inputs (ms,a 1 mr,a 1ms,b in Table 1) and total decomposition rate (k0,a
and k0,b in Table 1). Thus, for example, in thewarmer/wetter scenario with high sensitivity forproductivity and low sensitivity for decompositionrate, the simulation would multiply baseline valuesof ms,a, mr,a, and ms,b by 1.390, multiply baselinevalues of k0,a and k0,b by 1.231, and multiply ZWT by0.75; by contrast, in the cooler/drier scenario, themultipliers would be 0.719, 0.812, and 1.25. Notethat these scenarios do not represent sensitivity toclimate variability or change, since the modifiedconditions were in effect for the entire period ofpeat development.
The fen scenario showed very little sensitivity tochanges in water table depth (Figure 6a, and Table1), so the effect of a water table shift was very small.The bog scenario had much larger sensitivities toboth temperature and water table differences (Fig-ure 6b). In both scenarios, warmer and wetter con-ditions were most conducive to peat accumulation.Both bog and fen sensitivities to wetter or drier
conditions were relatively insensitive to the effect ofwarming or cooling on the ecophysiological sce-nario. When productivity and decomposition ratesensitivities to temperature were the same (bothlow or both high), productivity effects dominatedand warmer conditions led to enhanced accumula-tion whereas cooler conditions led to reduced accu-mulation. Only when decomposition rates had aQ10 of 3 and productivity had a Q10 of 2 did cooler
Figure 6. Sensitivity to climate of peat accumulation for(a) fen and (b) bog under four ecophysiological scenariosrepresenting high and low sensitivity to temperature forboth productivity (total NPP) and decomposition rates. Ineach panel, each group of six bars represents the sensi-tivity to climate of a particular ecophysiological scenario,modeled as a Q10 response of 2 (low sensitivity) or 3(high sensitivity). Climate scenarios were changes of 63°C and 6 25% in water table depth. The base scenariovalue (standard climate) is represented by a horizontalline labeled “base run”. Because the ecophysiological sce-narios represented sensitivity to temperature changeonly, the wetter and drier scenarios (with no temperaturechange) are also represented by horizontal lines.
Peat Decomposition and Accumulation 489
temperatures lead to enhanced accumulation andwarmer temperatures cause reduced accumulation.
Comparison with Peat Core Data
PDM simulations were compared to age–depth pro-files from cores collected in four peatlands in east-ern Canada, selected because of their various pastand present environmental conditions and becausewe have firsthand information about their develop-mental history. Mer Bleue and Mirabel bogs arelocated approximately 150 km apart in the St. Law-rence Lowlands, near Ottawa and Montreal, respec-tively. Frontenac peatland is found another 150 kmto the east of Montreal, in the Lower Appalachianphysiographic region, near Sherbrooke. Finally,Malbaie bog lies in the central part of the Lauren-tian Highlands about 100 km north of Quebec City.
The individual peatlands span a range of climateconditions (Table 2). Malbaie differs strikingly fromthe other peatlands in that it is a bog patterned bynumerous pools. It is also the peatland where precip-itation is highest and temperature lowest. Like Fron-
tenac peatland, Malbaie bog occupies a tilted basin inthe regional topography; in both areas, peat has accu-mulated over an irregular till plain with scattered localpools. Frontenac peatland is covered by a domed bog(80%) to which a poor fen (20%) is adjoined down-slope. Mirabel and Mer Bleue are both slightly domedbogs without pools, lying in flat topography. In allpeatlands, black spruce (Picea mariana (Mill.) BSP)and larch (Larix laricina (Du Roi) Koch.) are dispersedover a microtopographic pattern of hummocks andhollows. Shrub heath communities are dominant(Chamaedaphne calyculata, Kalmia angustifolia, Ledumgroenlandicum). Sedge lawns are widespread only fur-ther away from the dome at Frontenac (Ball 1996).The bryophyte layer is composed almost entirely ofSphagnum mosses.
The postglacial development of these four peatlandsis well known through previous or ongoing studies.Peat thickness and type were observed at 50–140spots at Frontenac, Malbaie, and Mirabel; and a300-m transect of peat depth from the margin of thebog to the center is available at Mer Bleue. The peat
Table 2. General Site Characteristics for the Peatland Sites with Peat Core Data used to Test the PDMModel
Site Frontenac Mer Bleue Mirabel Malbaie
Physiographic region Appalachians St. Lawrence Lowlands LaurentiansLatitude 45°589N 45°259N 45°689N 47°369NLongitude 71°089W 75°409W 74°079W 70°589WElevation (m) 360 65 75 800Mean annual temp. (°C)a 4.2 5.8 6.1 0.0Mean July temp. (°C)a 18.5 20.8 20.8 14.8Mean January temp. (°C)a 211.4 210.8 210.3 215.3Degree-days . 5°Ca 1500 2050 2070 850Total precipitation (mm)a 1130 910 740 1530Annual snowfall (%)a 24 30 29 39Regional vegetation type Maple/Birch/Fir Sugar Maple/Hickory Sugar Maple/Hickory Spruce/FirPeatland size (km2) 1 28 2 1Peatland types Domed bog, fen Bog Bog Bog with poolsRegional topography Incised plateau Terraces Terraces Hilly plateauLocal topography Tilted Horizontal Horizontal TiltedBasin margins Smooth-sided Steep-sided Smooth-sided Smooth-sidedUnderlying deposit Till Marine clay Marine clay TillInitial conditions Scattered pools Shallow lake Shallow lake Scattered pools
aEnvironment Canada 1994.bDates are either radiocarbon dates or from pollen correlation with well-dated events in the area.
490 Frolking and others
cores were analyzed with a set of paleoecological tech-niques (loss on ignition, radiocarbon dating, and anal-yses of pollen and spores, testate amoebae, mosses,and plant macrofossils) to reconstruct local and re-gional vegetation history, local hydrological condi-tions at the peat surface, net sediment accumulationrate and decomposition events, and the overall devel-opmental history of the peatlands since deglaciation(Lavoie 1998; Lavoie and Richard 2000a; S. D. Mullerand P. J. H. Richard unpublished). Additional paleo-hydrological information (lake level changes) wasavailable to assess the regional water balance in thearea (Lavoie and Richard 2000b; S. D. Muller andP. J. H. Richard unpublished).
For comparison with the PDM results, we se-lected peat cores for which an adequate chronologyof peat accumulation was available. All dates re-ported are calibrated radiocarbon dates (Stuiver andothers 1998) or dates obtained through pollen cor-relation with well-dated events in pollen diagramsfrom neighboring sites (Mott and Camfield 1969).The palynological and plant macrofossil record forthese cores indicated that the Mer Bleue and Mira-bel bogs, located in the St. Lawrence Lowlands,began as large ponds left after the drainage of thepostglacial Champlain Sea. Rich fens developedover the organic deposits left by marshes (telmaticpeat), then changed to bogs at some point in theirdevelopment (Table 2). Frontenac and Malbaie de-veloped directly as fens on till plains (glacial depos-its) with only small pools around and then evolvedinto bogs, except at the Frontenac poor-fen site.
We adopted two approaches for the developmentof model profiles for the two-stage peatlands. Ineach case, we generated a bog peat profile from thepresent (surface) back to the transition time (2400–6800 BP, depending on the core). To generate theunderlying fen peat, in one case we simply gener-ated a fen peat profile from the present (surface)back to the basal date of the core (8000–11,600 BP,depending on the core) and took the lower portionof this fen peat from the transition time to the base.We call this the “bog/fen scenario (b/f)”. Our secondmethod of generating a fen peat profile was to “bogthe fen.” Deeper fen peat was subjected to bog-anoxic conditions (that is, fanox 5 0.025 rather than0.1). A linear transition from fen-anoxic to bog-anoxic conditions occurred in the 0.45 m below thefen root zone; this layer, extending from 0.3 m to0.75 m, corresponded to about 200 years of accu-mulation. The portion of this bogged-fen peat fromthe transition time to the base was then placedunder the bog peat to generate a bog/bogged fenprofile (b/b2f).
We also generated a bog/bogged rich fen (b/b2rf)
profile, where initial fen parameters were set tovalues representative of current rich fen data. Wechose to test a rich-fen parameterization for theinitial fen stage because higher base cation (Ca, Mg,K) concentrations in bulk peat samples from thedeep portion of the Mer Bleue core (P. J. H. Richardunpublished) indicate that the site initially had con-ditions appropriate for a rich fen. Similar initialconditions are deduced from palynological and pa-leobotanical analyses of the other cores, indicatinghigher nutrient availability after ice retreat (Fron-tenac, Malbaie) or after drainng of the postglacialChamplain Sea (Mer Bleue, Mirabel). The parame-ter changes from Table 1 were ms,a 5 0.35 kg m22
y21; mr,a 5 1.0 kg m22 y21, k0,a 5 0.38 y21, ms,b 50.15 kg m22 y21; and k0,b 5 0.07 y21. These werebased on NPP estimates for rich fens in westernCanada (Thormann and Bayley 1997) and mass lossdecomposition rates for rich-fen species (T. R.Moore unpublished). Overall peat accumulation forthe rich-fen scenario was significantly greater thanfor the standard-fen scenario (Figure 2), but most ofthis additional peat was less than 2000 years old.
In each case we compared PDM simulations withour default parameterizations (Table 1 or rich-fenvalues above) to the core date profiles. We did nottry to adjust parameters to fit site data because thereare no measurements at these sites to constrainwater table depth or total and component NPPrates. All of these values are site specific and have astrong influence on peat accumulation.
Frontenac peatland. The PDM baseline fen sce-nario compares well to the Frontenac fen core for thelast 6000 years (FRON-1), (Figure 7a). However,PDM greatly underestimated the peat remaining fromthe peatland’s first 1500 years. Accumulation ratesinferred from the core were then much higher (per-haps related to higher NPP due to higher nutrientlevels). We tested a second scenario in which theoldest peat (6000–7500 BP) was from a rich-fen sim-ulation (see Figure 2). However, this made little dif-ference and the discrepancy between PDM simulationand core data remained (Figure 7a).
We generated the following three model scenar-ios for the FRON-2 bog core: (a) b/f: 5300 years ofbog peat over 6500 years of fen peat, (b) b/b2f:5300 years of bog peat over 6500 years of bogged-fen peat, and (c) b/b2rf: 5300 years of bog peat over6500 years of bogged–rich-fen peat. For theFRON-3 bog core, we used only the b/b2rf scenario.The bog portion of these scenarios generated a verygood portrayal of net peat accumulation behaviorobserved in the FRON-2 and FRON-3 cores for theirbog period to 5300 BP (Figure 7b). The b/f scenariounderestimated the peat remaining from peatland’s
Peat Decomposition and Accumulation 491
first 6500 years as a fen, but subjecting this fen peatto bog-anoxic conditions (b/b2f) improved themodel fit to the data. The b/b2rf scenario gave thebest approximation to the observed age–depth pro-file of the FRON-2 core and had reasonable agree-ment with the FRON-3 core (Figure 7b).
Mer Bleue and Mirabel bogs. We generated ab/b2rf scenario for both the Mer Bleue and Mirabelsites (Figure 7c). The default bog scenario was in
very good agreement with field observations at MerBleue (over the last 6500 years), but the observedrich-fen peat accumulation was much higher thanthat simulated by the model between 8400 to 6500years ago (Figure 7c). The thickness of the under-lying bogged–rich-fen peat as simulated by PDMwas more similar to that observed 7200 years ago inthe case of Mirabel core (Figure 7c) when submittedto bog conditions 6800 years ago, but the model
Figure 7. PDM profiles (lines) compared to dated peatcores (dated depths correspond to points on each panel)for (a) Frontenac fen core; (b) Frontenac bog cores; (c)Mer Bleue and Mirabel bog cores, and (d) Malbaie bogcores. See Table 2 for site descriptions. Notation for model
profiles (heavy lines) is as follows: f 5 fen, f/rf 5 fen overrich fen, b/f 5 bog over fen, b/b2f 5 bog over bogged-fen,and b/b2rf 5 bog over bogged-rich fen (see Table 2 forfen-to-bog transition dates).
492 Frolking and others
overestimated the bog peat accumulated in thatcore. The most important characteristic that distin-guishes the two cores is their position relative togroundwater (and nutrient) supply. The Mer Bleuebog core is located laterally with respect to theentire peatland, in the northernmost fingerlike ex-panse through which the peatland drains west-ward, while the Mirabel bog core is centrally lo-cated within a much smaller system (Table 2). TheMirabel core was the only one in which we foundevidence for local fires (conspicuous charcoal lay-ers, for example, at 25 cm and 80 cm) (S. Mullerunpublished); the core age–depth curve may thusdepict reduced peat accumulation, either because ofthe peat that burned, or due to postfire conditionsnot immediately conducive to peat growth, or both.
Malbaie bog. We compared model b/b2f andb/b2rf scenarios to MAL-2 and the b/b2rf scenario toMAL-3, again using our baseline parameterizations.The bog scenario accumulated peat more rapidlythroughout the profile than indicated by two cores(Figure 7d). At MAL-2, the apparent fen peat accu-mulation rate was similar for the model and the coredata, but the model had generated 2.5 m of bog peatin the past 5000 years, whereas the core containedonly about 1 m of bog peat (Figure 7d). For MAL-3,the model accumulated peat at about twice the ob-served rate through the bog and fen stages (Figure7d). The fact that the observed peat accumulationrates during both (rich) fen and the overlying bogstages are lower than at the other sites strongly pointsto a climate effect, with temperature dominating be-cause Malbaie is both colder and wetter than theother sites (Table 2). This conclusion is also supportedby the results of model sensitivity to climate, unlessdecomposition is more sensitive to temperature thanproductivity (Figure 6). Clymo and others (1998) alsoreported lower peat accumulation rates in cooler cli-mates. In addition, the widespread pools found allover Malbaie bog may well have consumed some ofthe peat accumulated in the past by causing decom-position with little production. The position of thepools has shifted constantly during the peatland’s his-tory (Lavoie and Richard 2000a). The absence ofslowly-decomposing Sphagnum tissues until 1000years ago at MAL-2 probably means that tissue de-composition rates were faster before that time andthus also deeper than 1000 years (around 0.8 m) intothe peat. Faster decomposition rates lead to loweraccumulation rates.
DISCUSSION AND CONCLUSIONS
Several features in PDM distinguish it from previ-ously published models of peat accumulation. PDM
used observed vegetation productivity and fresh lit-ter decomposition rates as key parameters. Themodel then included three mechanisms for slowingdecomposition rate with depth in the peat profile.First, as litter decomposed, it became more resistantto decomposition. By the bottom of the profile,cohorts had lost about 96.5% (bog) to 98.5% (fen)of their initial mass, reducing decomposition ratesby this same amount. Second, a temperature effect,based on observed monthly mean temperature pro-files and a Q10 5 2 temperature rate multiplierfunction, reduced decomposition rates by 20% be-low a depth of about 1 m. depth. Finally, there wasa reduction in decomposition rates by a factor of 10(fen) or 40 (bog) due to anoxic conditions. Theproduct of these three effects was to reduce decom-position rates from 0.05–0.2 y21 (surface) to0.00012 y21 (bottom of fen) and 0.000076 y21
(bottom of bog), or by about a factor of 1000, sim-ilar to the estimate of Belyea and Clymo (1999).This relation of decomposition to observable pro-cesses at any depth in the peat profile allows PDMto predict a peat profile based on surface conditionsand does not require fitting any parameters to peatcore age–depth data sets.
A second new feature in PDM is the explicitinclusion of roots and fresh root litter input into thepeat column. For both bog and fen scenarios, themean water table depth determined the depth ofthe fully oxic zone. In the bog scenario, the meanwater table depth was also the rooting depth. In thefen scenario, a partially oxic zone extended fromthe mean water table depth down to the rootingdepth. Root litter inputs affected peat accumulationprimarily by incorporating fresh litter at some depthin the peat profile, closer to the anoxic zone. Thisroot material had a shorter transit time through theoxic zone, and more of the material was transferredto the anoxic zone. In both the bog and fen scenar-ios, increasing the proportion of vascular NPP thatwas deposited as litter below the surface led toincreased total peat accumulation. A third new fea-ture in PDM was the explicit treatment of littertissue from two different vegetation types (vascularand moss), with different initial decompositionrates. Total peat accumulation depended on thefraction of total NPP attributed to mosses and vas-cular plants.
With these three features, PDM was able to de-velop scenarios representing both bog and fen con-ditions. Although fen productivity was 17% higher,peat accumulation was less for the fen scenario after2000 years, primarily because the assumption ofmore severe anaerobic conditions at depth in thebog reduced mass loss by the deeper peat. Two
Peat Decomposition and Accumulation 493
other parameters also contributed to greater peataccumulation in the bog: One was a higher initialmass loss rate for fen moss than bog moss; the otherwas a higher proportion of NPP by moss for the bog.These were offset to some degree by the greaterproportion of total NPP from roots for the fen sce-nario.
Clymo and others (1998) described three differ-ent measures of a peatland carbon balance: thelong-term apparent rate of carbon accumulation(LARCA 5 total peat mass divided by basal age), thecurrent rate of carbon accumulation (TRACA), andthe rate of carbon transfer from the oxic zone to theanoxic zone (P*) (see Eq. [1]). They also estimatedthe mean decay rate of the peat in the anoxic zone(a*, equivalent to a mass-weighted average k in ourmodel). They generated P* and a* parameters byfitting to peat core data a simple model with con-stant P* and decomposition as in Eq. (3) (a 5 1 or2). All of these indicators of peat accumulation canbe derived in PDM without curve fitting. LARCAvalues for PDM were similar, whereas TRACA val-ues were lower for both the bog and the fen sce-narios than reported by Clymo and others (1998) intheir analysis of 795 peat cores from Finland (Table3). Peat transfer rate to the anoxic zone was similarfor the bog scenario, but the fen scenario had muchgreater transfer to the anoxic zone, primarily be-cause of its significant root inputs into its partiallyanoxic zone between the water table and the bot-tom of the root zone. PDM mean decompositionrates for the anoxic zone were faster for the fen andslower for the bog than the curve-fit values Clymoand others (1998) derived for a linear decay model(as used in PDM) and a mean annual temperatureof 5°C. Overall, we conclude that with surface ob-servations of NPP and decomposition rates and asimple formulation for the change in tissue decom-
posability as peat moves down the profile, PDM cangenerate critical peat-producing parameters similarto those observed in northern peatland systems.
When PDM peat accumulation results were com-pared with observed peat core age–depth profiles,there were cases of both good and poor agreement.Although we used a very general parameterizationand did not fit any parameters, PDM generated veryrealistic profiles for the most recent 5000–6000years at one fen (FRON-1) and three bog sites(FRON-2, FRON-3, and MB-930). This same pa-rameterization also overestimated peat accumula-tion during the past 2500–5000 years at three othersites (MAL-2, MAL-3, and MIR-1). There are sev-eral factors that may have contributed to theseoverestimations, some of which are demonstratedby the PDM sensitivity results (Table 1 and Figures5 and 6).
First, PDM overall accumulation was quite sensi-tive to total vegetation NPP, particularly for bogs.Peatland NPP generally decreases with cooler tem-peratures (Moore 1989), so it is likely that Malbaiehas relatively low productivity and thus lower ac-cumulation. Moore and others (2001) presented arelationship between NPP and mean annual airtemperature that suggests that Malbaie NPP shouldbe about 55%–70% of that at the other sites. Sec-ond, variations in the relative NPP of mosses tovascular plants will also influence productivityrates. This ratio is not known for many peatlands,nor is its variability over millennial time scales.Third, bog accumulation rates are very sensitive towater table depth, and little is known about itshistorical variability. Mirabel shows evidence of lo-cal fires, which may have interrupted or tempo-rarily reversed peat accumulation. Finally, from2000 to 7000 cal. BP, the MAL-2 core shows veryhigh regional pollen concentrations (excluding
Table 3. Comparison of Characteristics of Peat Accumulation as Simulated by the PDM Model and asDeveloped by Clymo and Others (1998)
Parameter Units
Present study
Clymo and Others 1998Bog Fen
Age y 8000 8000 ,1000–;11,000P* kg m22 y21 0.059 0.17 0.072a
a* y21 0.00014 0.0011 0.0004a
LARCA kg m22 y21 0.036 0.031 0.024–0.072b
TRACA kg m22 y21 0.022 0.010 0.054c
The Clymo and others (1998) values are based on fitting curves to peat age–depth profile data for several hundred peat cores from Finland.aBased on Clymo and others (1998; Figure 17) and a mean annual temperature of 5°CbBased on Clymo and others (1998; Figure 15a) for peat ages of around 8000 yearscBased on P* value for the linear decay model and Figure 15b of Clymo and others 1998
494 Frolking and others
peatland plants) (Lavoie 1998; Lavoie and Richard2000a), indicating for MAL-2 either high decompo-sition rates (relative to production) or intermittentperiods of peat consumption. When pools arepresent in a peatland, there will be significant de-composition in the peat under the pool, but negli-gible productivity in the pool (Seppala and Kouta-niemi 1985). This could lead to both enhancedpollen concentrations in the underlying peat and anage–depth profile similar to that observed in theMAL-2 core.
Because PDM is essentially a static model, theonly dynamic feature we could simulate was a sin-gle transition from fen to bog, a common phenom-enon in northern peatlands (Davis 1984; Janssensand others 1992; Kuhry and others 1993) that oc-curred at all of our bog core sites. For the Frontenacbog cores, PDM successfully simulated the underly-ing fen peat accumulation when this peat was sub-jected to bog-anoxic conditions, and its productivityand root:shoot ratio were adjusted to rich-fen val-ues. The same adjustment failed to produce bettersimulations for the Mer Bleue and the Mirabel bogcores. At the Malbaie core 2 site, the underlyingbogged-fen peat accumulation was similar to thatobserved. For the Malbaie 3 core, PDM overesti-mated fen peat accumulation; whereas at Mirabeland Mer Bleue, PDM underestimated fen peat ac-cumulation. In all PDM sensitivity scenarios, fenpeat accumulation was less responsive than bogpeat accumulation. Nonetheless, changes in rootcontribution to total productivity, moss contribu-tion to total productivity, temperature, or total pro-ductivity itself could each change fen peat accumu-lation by 6 20% or more, so scenarios of greater orlesser accumulation could be derived.
Although the representation of fen-to-bog tran-sitions in PDM is very simplistic, it generally im-proves model performance, suggesting that futuredevelopment in this area would be fruitful. Amongthe additional complexities that may be importantare details of the nature and rate of shifts in pro-ductivity and decomposition during the vegetationtransition from fen to bog, and rate of developmentof bog-anoxic conditions down the underlying fenpeat. The extremely high rate of accumulation from6000 to 8000 cal. BP observed in the Frontenac fencore (FRON-1) is coincident with a transition frombrown moss to herbaceous peat. A rich-fen param-eterization could accumulate peat rapidly (Figure2), but conditions for decomposition of this earlypeat would have had to be much less favorable topreserve such a large amount of this old peat. Thetransition from initial marsh (telmatic peat) to truefen peat, when it applies, is also a matter of concern
because of the tremendous differences in produc-tivity and decomposition.
The PDM value for fanox for the bog scenarios(0.025) is lower than reported values, and the fenscenario fanox value (0.10) is at the low end of re-ported values. Scanlon and Moore (2000) reporteda mean aerobic/anaerobic decomposition rate ratioof 0.1 for 12-day incubations of intact cores. Inshort-term slurry incubations, Moore and Dalva(1997) and Bridgham and Richardson (1992) ob-served values around 0.4. Longer incubations alsoshow a range in values—for example, Magnusson(1993): 26 weeks, 0.09–0.2; Updegraff and others(1995):80 weeks, 0.25–0.5; Bridgham and others(1998): 59 weeks, 0.13–0.25. Based on the PDMmodel results, we hypothesize that these observa-tions underestimate the degree of anoxicity in deeppeat, particularly for bogs. All laboratory incuba-tions involve significant disturbance of the in situpeat. The isolation of deep peat from the surface forcenturies to millennia, due to extremely low hy-draulic conductivities (see, for example, Ivanov1981), cannot be re-created in the lab. In addition,the longer incubations may underestimate optimalaerobic decomposition due to developing substratelimitations. Another possible explanation of the ap-parent success of the low anoxic decompositionrates in PDM is that they compensate for an under-estimation in the decline of cohort decomposabilitywith mass loss (that is, a . 2 in Eq. [3]).
One important factor affecting peat accumula-tion at any site that was not included in PDM wasthe impact of varying environmental conditionsover the millennia of peatland development andthe effects of autogenic processes such as micro-topographic development. Retreat of the ice sheetleft bare soils whose nutrient content was initiallyhigh, then was progressively diminished throughleaching (Willis and other 1997; Iversen 1954).There have been significant changes in climaticconditions in eastern Canada over the past 10,000years (for example, see Wright and others 1993;Richard 1994; Lavoie and Richard 2000b). Thesechanges have probably influenced peatland plantproductivity (and thus litter input rates), peat soilclimate profiles (and thus decomposition rates),and the composition of peatland vegetation (andthus tissue decomposability). Variation in any orall of these ecosystem function rates would leadto significant variation in peat accumulation ratesover time (see Figures 5 and 6) (Clymo and others1998). In the PDM results presented here, how-ever, total NPP, the proportions of NPP from vas-cular and moss vegetation, and initial decompo-sition rates were all assumed to be constant
Peat Decomposition and Accumulation 495
through time; and thus, model age– depth profilesare always a smooth, monotonic curve. To morefully explore the influence of the climatic varia-tions on peat accumulation, the model wouldrequire dynamic water table and thermal re-gimes. These would influence plant demographicsand both productivity and decomposition ratesand thus most parameters in PDM.
Dynamics could be incorporated by merging PDMwith the Peat Accumulation Model (PAM) of Hil-bert and others (2000). PAM includes a simple wa-ter balance model that simulates the position of thewater table as a function of the inputs and outputsof water and the changing surface elevation of thepeatland. PAM also includes a simple description ofthe relationship between the position of water tableand NPP of a peatland. At present, PAM does notconsider the impact of variations in temperature,does not distinguish between moss and vascularplants or belowground and aboveground litter in-puts, and has an overly simplified decompositionsubmodel.
To examine the impact of variations in climate,two additional pieces of information are required.First, an approximation of the date of peatland ini-tiation is needed. In most major peatland regionsthere are estimates of the basal dates for the initia-tion of peat accumulation. The second requirementis a realistic chronology of the relative variations inmoisture input to the peatland from the time ofinitiation to the present day, obtained from a datasource independent of the peatlands. These inde-pendent chronologies are not readily available inmost peatland areas, but regional reconstructionsbased on palynological and macrofossil analyses oflake sediments in some peatland regions are forth-coming (for example, see Lavoie and Richard2000b; S. D. Muller and P. J. H. Richard unpub-lished).
ACKNOWLEDGMENTS
This work has been supported by grants from theNASA Terrestrial Ecology Program’s BOREASGuest Investigator Program to S. F. and an NSERCStrategic Grant to N.T.R. Paleoecological resultswere acquired through the NSERC-funded pro-gram Climate System History and Dynamics(Team 4: Konrad Gajewski, University of Ottawa,and P. J. H. R.). We thank Peter Lafleur for accessto unpublished data. We also thank Goran Agrenand an anonymous reviewer for suggestions andcomments that helped to focus and improve themanuscript.
REFERENCES
Aerts R, Wallen B, Malmer N. 1992. Growth-limiting nutrientsin Sphagnum-dominated bogs subject to low and high atmo-spheric nitrogen supply. J Ecoly 80:131–140.
Agren GI, Bosatta E. 1996a. Quality: a bridge between theoryand experiment in soil organic matter studies. Oikos 76:522–528.
Agren GI, Bosatta E. 1996b. Theoretical ecosystem ecology: un-derstanding element cycles. Cambridge (UK): Cambridge Uni-versity Press. 234 p.
Backeus I. 1990. Production and depth distribution of fine rootsin a boreal open bog. Ann Botan Fennici 27:261–5.
Ball T. 1996. Seasonal transitions in fluxes of carbon dioxide andmethane from an ombrotrophic peatland, Frontenc Bog,southern Quebec [thesis]. Montreal: Geography Department,McGill University. 142 p.
Bartsch I, Moore TR. 1985. A preliminary investigation of pri-mary production and decomposition in four peatlands nearSchefferville, Quebec. Can J Bot 63:1241–8.
Belyea LR. 1996. Separating the effects of litter quality andmacroenvironment on decomposition rates in a patternedpeatland. Oikos 77:529–39.
Belyea LR, Clymo RS. 1999. Do hollows control the rate of peatbog growth? In: Meade R, Standen V, Tallis JH, editors. Pat-terned mires: origin and development, flora and fauna. Lon-don: British Ecological Society, Mires Research Group. p 1–15.
Boelter DH. 1968. Important physical properties of peat material.Proceedings of the Third International Peat Congress., Na-tional Research Council of Canada, Quebec City. p. 150–156.
Bridgham SD, Richardson CJ. 1992. Mechanisms controlling soilrespiration (CO2 and CH4) in southern peatlands. Soil BiolBiochem 24:1089–99.
Bridgham SD, Updegraff K, Pastor J. 1998. Carbon, nitrogen, andphosphorus mineralization in northern wetlands. Ecology 79:1545–61.
Brinson MM, Lugo AE, Brown S. 1981. Primary productivity,decomposition, and consumer activity in freshwater wetlands.Annual Rev Ecol Syst 12:123–61.
Clymo RS. 1984. The limits to peat bog growth. Philos Trans RlSoc London B 303:605–54.
Clymo RS. 1992. Models of peat growth. Suo 43:127–36.
Davis AM. 1984. Ombrotrophic peatlands in Newfoundland,Canada: their origins, development and trans-atlantic affini-ties. Chem Geol 44:287–309.
Environment Canada. 1994. Canadian monthly climate data and1961–90 normals on CD-Rom (version 3.0). Ottawa; Atmo-spheric Environment Service.
Frolking S, Bubier JL, Moore TR, Ball T, Bellisario LM, BhardwajA, Carroll P, Crill PM, Lafleur PM, McCaughey JH, and others.1998. Relationship between ecosystem productivity and pho-tosynthetically-active radiation for northern peatlands. GlobalBiogeochem Cycles 12:115–26.
Gilbert D, Amblard C, Bourdier G, Francez A-J. 1998. Short-termeffect of nitrogen enrichment on the microbial communities ofa peatland. Hydrobiologia 373/374:111–9.
Gorham E. 1995. The biogeochemistry of northern peatlandsand its possible response to global warming. In: WoodwellGM, MacKenzie FT, editors. Biotic feedbacks in the global
496 Frolking and others
climatic system: will the warming speed the warming? NewYork: Oxford University Press. p 169–87.
Gorham E. 1991. Northern peatlands: role in the carbon cycleand probable responses to climatic warming. Ecol Appl. 1:182–95.
Gorham E, Janssens JA. 1993. The paleorecord of geochemistryand hydrology in northern peatlands and its relation to globalchange. Suo 43:117–26.
Heal OW, Anderson JM, Swift MJ. 1997. Plant litter quality anddecomposition: an historical overview. In: Cadisch G, GillerKE, editors. Driven by nature: plant litter quality and decom-position. New York: CABI. p 3–30.
Hilbert D, Roulet NT, Moore TR. 2000. Modelling and analysis ofpeatlands as dynamic systems. J Ecol 88:230–42.
Hogg EH. 1993. Decay potentials of hummock and hollow Sphag-num peats at different depths in a Swedish raised bog. Oikos66:269–78.
Ingram HAP. 1983. Hydrology. In: Gore, AJP editor. Mires:swamp, bog, fen, and moor; vol 4A. Amsterdam: ElsevierScientific Publishing p 67–158.
Irwin RW. 1968. Soil water characteristics of some (southern)Ontario peats. Proceedings of the Third International PeatCongress, National Research Council of Canada, Quebec City.
Ivanov KE. 1981. Water movement in mirelands. New York:Academic Press. p. 165–168.
Iversen J. 1954. The late-glacial flora of Denmark and its relationto climate and soil. Danmarks Geol Undersgelse 11(80):87–119.
Janssens JA, Hansen BCS, Glaser PH, Whitlock C. 1992. Devel-opment of a raised-bog complex. In: Wright HE, Jr, Coffin BA,Aaseng NE, editors. The patterned peatlands of Minnesota.Minneapolis: University of Minnesota Press. p 189–221.
Jauhiainen J, Silvola J, Vasander H. 1999. The effects of in-creased nitrogen deposition and CO2 on Sphagnum angustifo-lium and S. warnstorfii. Ann Bot Fennici 35:247–56.
Johnson LC, Damman AWH. 1993. Decay and its regulation inSphagnum peatlands. Adv Bryol 5:249–96.
Kuhry P, Nicholson BJ, Gignac LD, Vitt DH, Bayley SE. 1993.Development of Sphagnum dominated peatlands in boreal con-tinental Canada. Can J Bot 71:10–22.
Lafleur PM, Roulet NT, Admiral S. 2001. The annual cycle of CO2
exchange at a boreal bog peatland. J Geophys Res. 106:3071–3081.
Lahde E. 1967. Seasonal variation in the depth of the aerobiclimit and the ground water table in virgin and in drainedMyrtillus spruce swamp. Acta Forestalia Fennica 81:1–15.
Lavoie M. 1998. Dynamique de l’entourbement et fluctuationsdes niveaux lacustres postglaciaires au Quebec meridional[thesis,]. Montreal: Department de geographie, Universite deMontreal.
Lavoie M, Richard PJH. 2000a. Paleoecologie de la tourbiere dulac Malbaie, massif des Laurentides (Quebec): evaluation durole du climat sur l’accumulation de tourbe. Geogr Phys Quat.54:165–181.
Lavoie M, Richard PJH. 2000b. Postglacial water-level fluctu-ations of a small lake in Southern Quebec. Holocene 10:621–34.
Lavoie M, Richard PJH. 2000c. The role of climate on the devel-opmental history of Frontenac peatland, Southern Quebec.Can J Bot 78:668–84.
Magnusson T. 1993. Carbon dioxide and methane formation inforest mineral and peat soils during aerobic and anaerobicincubations. Soil Biol Biochem 25:877–83.
Moore TR. 1989. Growth and production of Sphagnum at five fensites, subarctic eastern Canada. Can J Bot 67:1203–7.
Moore TR, Bubier JL, Frolking S, Roulet NT. 2001. Plant biomassand production and CO2 exchange in an ombrotrophic bog. JEcol. Forthcoming.
Moore TR, Dalva M. 1997. CO2 and CH4 exchange potentials ofpeat soils aerobic and anaerobic laboratory incubations. SoilBiol Biochem 29:1157–64.
Mott RJ, Camfield M. 1969. Palynological studies in the Ot-tawa area. Paper 69 –38. Ottawa; Geological Survey of Can-ada. 16 p.
Paustian K, Agren GI, Bosatta E. 1997. Modelling litter qualityeffects on decomposition and soil organic matter dynamics. In:Cadisch G, Giller KE, editors. Driven by nature: plant litterquality and decomposition. New York: CABL. p 313–35.
Puranen R, Makila M, Saavuori H. 1999. Electric conductivityand temperature variations wi thin a raised bog in Finland:implications for bog development. Holocene 9:13–24.
Richard PJH. 1994. Postglacial palæophytogeography of the east-ern St. Lawrence River watershed and the climatic signal ofthe pollen record. Palæogeog Palæoclimatol Palæoecol 109:137–63.
Saarinen T. 1996. Biomass and production of two vascular plantsin a boreal mesotrophic fen. Can J Bot 74:934–8.
Scanlon D, Moore TR. 2000. CO2 production from peatland soilprofiles: the influence of temperature, oxic/anoxic conditionsand substrate. Soil Sci 165:153–60.
Seppala M, Koutaniemi L. 1985. Formation of a string and pooltopography as expressed by morphology, stratigraphy and cur-rent processes on a mire in Kuusamo, Finland. Boreas 14:287–309.
Siegel DI. 1993. Groundwater hydrology. In: Wright HE, Jr,Coffin BA, Aaseng NE, editors. The patterned peatlands ofminnesota. Minneapolis: University of Minnesota Press. p163–72.
Stuiver M, Reimer PJ, Bard E, Beck JW, Burr GS, Hughen KA,Kromer B, McCormac FG, van der Plicht, Spurk M. 1998.INTCAL98 radiocarbon age calibration, 24,000–0 cal. BP. Ra-diocarbon 40:1041–83.
Thormann MN, Bayley SE. 1997. Above-ground net primaryproduction along a bog-fen-marsh gradient in southern borealAlberta, Canada. Ecoscience 4:374–84.
Tolonen K, Vasander H, Damman AWH, Clymo RS. 1992. Pre-liminary estimate of long-term carbon accumulation and lossin 25 boreal peatlands. Suo 43:277–80.
Updegraff K, Pastor J, Bridgham SD, Johnston CA. 1995. Envi-ronmental and substrate quality controls over carbon andnitrogen mineralization in a beaver meadow and a bog. EcolAppl 5:151–63.
Verhoeven JT, Toth E. 1995. Decomposition and Carex and Sph-agnum litter in fens: effect of litter quality and inhibitions byliving tissue homogenates. Soil Biol Biochem 27:271–5.
Verry ES, Brooks KN, Barten PK. 1988. Streamflow responsefrom an ombrotrophic mire. Vol. 1. International Symposiumon the Hydrology of Wetlands in Temperate and Cold Regions,The Publications of the Academy of Finland, Helsinki. p. 52–59.
Peat Decomposition and Accumulation 497
Wallen B. 1986. Above and below ground dry mass of the threemain vascular plants on hummocks of a subarctic peat bog.Oikos 46:51–6.
Williams BL, Silcock DJ. 1997. Nutrient and microbial changes inthe peat profile beneath Sphagnum magellanicum in response toadditions of ammonium nitrate. J Ecol 34:961–70.
Willis KJ, Braun M, Sumegi P, Toth A. 1997. Does soil changecause vegetation change or vice versa? A temporal perspectivefrom Hungary. Ecology 78:740–50.
Wright HE, Kutzbach JE, Street-Perrott FA, Ruddiman WF,Webb T III. 1993. Global climates since the last glacial maxi-mum. Minneapolis: University of Minnesota Press.
