Soil carbon stocks and their rates of ac
GLOBAL BIOGEOCHEMICAL CYCLES, VOL. 12, NO. 4, PAGES 687-701, DECEMBER 1998
Soil carbon stocks and their rates of accumulation
and loss
in a boreal forest landscape
Gloria
Rapalee
l'2,3,Susan
E.Trumbore
4,EricA.Davidson
l,
Jennifer
W.Harden
5,andHugo
Veldhuis6
Abstract. Borealforestsandwetlandsare thoughtto be significantcarbonsinks,andthey
couldbecomenet C sourcesas theEarth warms. Most of the C of borealforestecosystems
is
storedin themosslayerandin thesoil. The objectiveof thisstudywasto estimatesoilC
stocks
(including
moss
layers)
and
rates
ofaccumulation
andloss
fora733km2area
ofthe
BORealEcosystem-Atmosphere
Studysitein northernManitoba,usingdatafrom smaller-scale
intensivefield studies.A simpleprocess-based
modeldevelopedfrommeasurements
of soilC
inventoriesandradiocarbonwasusedto relatesoilC storageanddynamicsto soil drainageand
foreststandage. Soil C stockscovarywith soildrainageclass,with thelargestC stocks
occurringin poorlydrainedsites.Estimatedratesof soilC accumulation
or lossare sensitive
to theestimateddecomposition
constants
for the largepoolof deepsoilC, andimproved
understanding
of deepsoilC decomposition
is needed.While the uppermosslayersregrow
andaccumulate
C after fires,thedeepC dynamicsvary acrossthe landscape,
from a smallnet
sink
toasignificant
source.
Estimated
ne_2t
soi{
Caccumulation,
averaged
fortheentire
733
km2area,
was20gCrn-2yr-• (28gCrn yr- accumulation
insurface
mosses
offset
by8gC
rn yr lostfromdeepC pools)in a yearwith no fire. Most of theC accumulated
in poorly
andverypoorlydrainedsoils(peatlandsandwetlands).Burningof themosslayer in only 1%
of uplandswouldoffsetthe C storedin the remaining99% of the area. Significantinterannual
variabilityin C storageis expectedbecauseof the irregularoccurrence
of fire in spaceand
time. The effectsof climatechangeandmanagement
on fire frequencyandon decomposition
of immensedeepsoilC stocksare key to understanding
futureC budgetsin borealforests.
Inverse models that calculate the latitudinal
1. Introduction
distribution
of
CO2sources
and
sinks
from
observed
CO2 and13CO2
distribu-
Boreal forestsplay an importantrole in the global carbon
budget.
Ofthe1500
petagrams
(1Pg= 10•Sg)ofC
stored
in
soils globally and 600 Pg of C in aboveground
biomass[Reeburgh,1997],aboutonethirdof thisglobalterrestrialC is found
in thebiomass,detritus,soil, and peatpoolsof the borealforest
biome, with about three fourths of the C in boreal forest biome
storedbelowground[Appset al., 1993; Peng et al., 1998].
Predictinghow thesevaststoresof soil C in the borealforest
biome will be affectedby future climate changerequiresan
understanding
of the factorscontrollingthe production,decomposition,and storageof organicC in borealecosystems.
tions suggestthat borealregionsmay be significantC sinks
[Tanset al., 1990; Ciais et al., 1995]. An alternativeapproach
to determining
C exchangeat the landscapeor regionalscaleis
to developprocessmodelsof C dynamicsand to use those
models,togetherwith a knowledgeof the spatialdistributionof
importantforcingfactors,to extrapolatefield measurements
of
C losses and gains to regional scales. Because it is
process-based,
the latterapproachmay alsobe usedto predict
theresponse
of the landscapeto changesin forcingparameters.
Responses
of borealforestsoilsto warming,changesin drainage, or changesin fire frequencyhaveall beenproposedto be
importantfor terrestrialC storage[Bonan, 1993; Moore and
•TheWoods
HoleResearch
Center,
Woods
Hole,Massachusetts.
Knowles, 1990; Gorham, 1991; Kasischke et al., 1995; Kurz
:Department
of Geography,
SalemStateCollege,
Salem, andApps,1995].
Ma•ssachusetts.
ONow with Departmentof Earth SystemScience,Universityof
California,
Ëvine.
Departmentof Earth SystemScience,Universityof California,
Irvine.
56U.S.
Geological
Survey,
Menlo
Park,
California.
Agricultureand Agri-Food Canada,BrandonResearchCentre,
We usedfieldstudiesof C input,storage,andturnoverin the
northem
regionof theborealforest,madeaspartof theBOReal
Ecosystem-Atmosphere
Study(BOREAS), to developsimple
modelslinking soilC storageandratesof accumulation
to two
major factors: soil drainageclassand the time sincelast fire
Manitoba Land ResourceUnit, Universityof Manitoba, Winnipeg,
[Trumbore and Harden, 1997; Harden et al., 1997]. These
Canada.
modelshavepreviously
beencombined
with a soildrainagemap
for a 120 year old black sprucestand to comparethe soil
component
of net C storagewith tower-based,
eddycovariance
measurementsof net ecosystemproduction[Harden et al.,
1997]. In this paper, we expand this scaling approachto
Copyright1998 by the AmericanGeophysical
Union.
Papernumber98GB02336.
0886-6236/98/98GB-02336512.00
687
688
RAPALEE ET AL. ßSOIL C DYNAMICS
IN A BOREAL FOREST LANDSCAPE
estimate
thesoilCstorage
andaccumulation
rates
fora733km2
2. Methods and Data Sources
area within the BOREAS Northern Study Area (NSA), in
2.1. Study Area
northern
Manitoba,Canada.More specifically,we (1) estimate
totalC stocksby horizonfor commonsoil serieson the basisof
The BOREAS Northern Study Area, located between
soil surveydata and analysesof data from individual soil Thompsonand NelsonHouse,Manitoba, is nearthe northern
profiles; (2) estimateratesof soil C accumulationon the basis limitof theclosed-crown
borealfore_st
(Figure1a) [Sellerset al.,
of C stocksanda simplemodelof C turnoverderivedfrom field
1994].
Wefocus
hereona 733km2study
areawithin
theNSA
radiocarbon
studies;(3) relatepatternsof C stocksandaccumu- (Figure 1b), for whicha soil surveymapwasgeneratedduring
lationratestopatterns
of drainage,mosscover,andfire history; the 1994BOREAS field season[Veldhuisand Knapp, 1998b].
and (4) generatearea-weightedmap•sof soil C stocksand Cornercoordinates
of the studyareaare: 56.00øN, 98.69øW;
accumulation
rates
across
the733km2study
areain 1994,the
56.00øN, 98.17øW; 55.80øN, 98.18øW; 55.80øN, 98.69øW.
yearin whichmostBOREAS field studieswere conducted.We
alsoidentifyareasof greatest
sensitivityanduncertainty
in these
Much of the regionis underlainby poorlydrained,varved
clay depositedby Glacial Lake Agassizand interspersed
with
frequentbedrockoutcrops. A few minor exposuresof well-
estimates.
...-Canadian
BorealForestMap
ß
•.."
(A)
L •X •-;;-.
• •
•.
• x" -.......
:
-.•:i:
.•
:
..":':)"
.. ':•..'
::.:::.::.:::.
:..::.:...•:::..•..
..... .......,,.-.............
...
(B)
....
•:. ...:
.....•,..•
..•'" •
'::"
:....
''......
,,
Nx
••
..' •icp•rt
.
•::
• o• . '
Figure
1. (a)MapofCanadian
boreal
forests
sh2owing
theBOREAS
study
region
andthenorthern
andsouthern
sites
and(b)map
of BOREASNorthernStudyArea. The733 km studyareais drawnin black. The eddyflux towersitesare: Old JackPine (OJP),
Old Black Spruce(OBS), Fen (FEN), YoungJackPine (YJP), andBeaverPond(BP). Thesemapsweremodifiedfrom those
publiclyavailablecourtesyof BOREAS (http://boreas.gsfc.nasa.gov).
RAPALEE ET AL.: SOIL C DYNAMICS
drained,
sandyglacialtill occur,largelyin theeastempartof the
studyarea,wheretwomajorhillscomposed
of sandandgravel
(kamedeposits)
runnorth-south
with reliefup to 60 m [Sellers
IN A BOREAL FOREST LANDSCAPE
689
[Bonan, 1993; Moore and Knowles, 1990; Gorham, 1991;
Kasischke
etal.,1995•Kurz
andApps,
1995].
A 1981
fire
burnedabout128 km of the predominantly
spruce-forested
et al., 1994]. (SeeFigure2a andPlate 1b.) Extensiveareasof
southernpart of the studyarea. (SeePlate 1a.) Smallerbums
the clayeydeposits
areoverlainby deepandshalloworganic
deposits
(peat). Permafrost
occurssporadically
andat variable
depths
in clayeyuplandsoilsandin veneerbogs[Sellerset al.,
1994; Rubec, 1988; Veldhuisand Knapp, 1998b], and is
continuous
in collapsescarandpeatplateaubogsandpalsas
[Rubec,1988; Veldhuisand Knapp, 1998b]. The area is
undulatingwith longgentleslopes.Foreststandagesrange
from 13 to 140 yearsor moreon uplands,markingtheperiod
totaling
about
33km2 occurred
in 1956and1964along
the
elapsedsincethe lastfire.
Stratifyingthe studyareaby drainageclassis an effective
meansof identifyinglandcovertype. We identifiedsix broad
land cover types on the basis of similaritiesin landform,
drainage,
depthof organics
(peat),anddominantforestandmoss
cover.In thefollowingdescriptions,
theCanadiansoilclassification[AgricultureCanadaExpertCommitteeon SoilSurvey,
1987] is followed (in parentheses)
by the U.S. classification
[Soil SurveyStaff, 1975, 1996 (also availableon the World
WideWebat http://www.statlab.iastate.edu/soils/keytax/)]:
(1)
upland, well-drained, sandy Eluviated Dystric Brunisols
(Cryochrepts)with jack pine (Pinus banksiana)cover; (2)
upland,moderatelywell-drained,clayeyOrthicGrayLuvisols
(Cryoboralfs)with mixed black spruce(Picea mariana) and
hardwood
standswith feathermoss(Pleurozium,Hylocomium
spp.)asthedominant
groundcover;(3) transitional,
imperfectly
to poorly drained, clayey Gleyed Gray Luvisols and Luvic
Gleysols(Cryoboralfs)with black spruceand a mixtureof
featherandsphagnum(Sphagnum
spp.)mosses;(4) poorlyto
verypoorlydrained,levelto gentlysloping,clayey,peatyLuvic
Gleysolsand Terdc Mesic Fibrisols(Cryoboralfsand Cryofibrists) with black spruceand sphagnummossmixed with
featherandothermosses;(5) peatlands
consisting
of varying
peatmaterials
thatarewell- to poorlydrainedat the surfaceand
havefrozenpeatand/ormineralmaterialat depth,with Fibric
and Mesic OrganicCryosols(Cryofibristsand Cryohemists),
andblackspruceandsphagnum/feather
mossvegetation
cover.
(Palsas
aremounds
ofuplandpeatlands
underlainby perennially
frozenpeatandmineralsoil,with roughandunevensurfaces,
andarecharacterized
by havingdomedsurfaces.Peatplateau
bogs,alsounderlainby continuous
permafrost,
rise abruptly
fromsurrounding
unfrozen
fensandarecharacterized
by having
relatively flat and even surfaces[Rubec,1988; Veldhuisand
Knapp,1998b]); and(6) verypoorlydrainedfensandpermafrostcollapse
bogswith deepTypicFibrisols(Cryofibrists),
and
sedgeandbrownmossvegetation.(Fensarewetlandsfed by
groundwaterthat is relativelynutrient rich (comparedto
precipitation),
whereas
bogsareoligotrophic,
havingtheirupper
peat layer (rootingzone) abovethe groundwaterflow, and
therefore
havea rootingzonelargelydependent
onprecipitation
for its water and nutrientsupply. Vegetationof the fensis
characteristically
sedge(Carex,Eriophorumspp.)andbrown
moss(Depranocladus,
Tomenthypnum
spp.). Sphagnum
and/or
feathermosses
occurin bogs.Collapsescarbogsarenutrient
poor circularor ovaldepressions
causedby subsidence
of the
peatlandsurfacedue to thawingof permafrost[Rubec,1988;
Veldhuisand Knapp,1998b]).
In additionto soil drainage,the incidenceof fire is an
important
factorcontrollingannualaccumulation
ratesof soilC
north-south
ridgeneartheeastern
boundary
whereprimarilyjack
pinehasregenerated.
In 1989,largeareasburnedin thewestern
and northernsections
of the NSA, outsideof the studyarea.
Noneof the studyregionburnedin 1994.
2.2. Soil C Storage
Modelsof soilC dynamics
weredeveloped
usingdataonC
inventories
andradiocarbon
aspartof theBOREAS field effort
[Trumbore
andHarden,1997; Hardenet al., 1997]. Upland
soil profilesweredividedintotwo layers(Figures2a and2b)
that aredistinctlydifferentin theirC dynamics:
(1) a surface
detritalor mosslayer,madeup of relativelyundecomposed
organic
material,
and(2) a deeplayerconsisting
of morehighly
decomposed
organicmatter(humiclayer),in which plant
macrofossils
are rare,and whereminoramounts
of organic
matter are incorporatedinto the mineralsoil A horizon. The
mineralB horizonwasalsoincluded
in thisdeeplayer.
2.2.1. Uplandsurfacesoillayers. Uplandsurface
layers
accumulate
C between
fireevents.OrganicC is inputdirectly
tothesurface
layerasdetritus
fromtreesandmosses.
Decompositionratesaregenerally
slow(meanresidence
timesof theorder
of decades),
sothatthickorganicmatsaccumulate
in shallow
layers before lossesthroughdecomposition
offset annual
additions.Feathermosscoveron moderately
well- to imperfectlydrainedsoilstendsto dryoutin summerbecause
it does
notoverlie
saturated
soilsandhencewill bumnearlycompletely
to mineralsoilin stand-killing
fires. In contrast,
sphagnum
mosscoveris associated
with imperfectly
andpoorlydrained
soils that tendto staymoistat depthandthereforebum less
completely.
We modeledsoil C accumulation
of the surfacelayers
betweenfiresas a simplebalanceof C inputsandfirst-order
decomposition
usingthefollowingequations
fromJennyet al.
[ 1949] and Harden et al. [ 1992, 1997]:
dC/dt= I- (k x C)
(1)
Solvingfor C in Equation(1) yields:
C(t) = I/k x (1 - e-kt)
(2)
whereC iscarbon
massin unitsof massperarea,t is numberof '
years
sincethelaststand-killing
fire,I is inputratein massper
unitarea
•per
year,
and
kisadecomposition
coefficient
inunits
of years-.
Input I ratesanddecomposition
k constants
(Table1) were
determined
usingtwo approaches.
First,verticalaccumulation
ratesof mosswereobtainedusingradiocarbon
analyses
to
determinethe ageof accumulated
C [Trumboreand Harden,
1997]. Second,surveys
of theC inventory
in organicmatter
abovethe mostrecentcharcoallayerin the soil profilewere
madeacrossa seriesof sites(chronosequence)
thatdifferedin
timesincelastfire [Hardenet al., 1997]. The estimatesof I and
k basedon chronosequences
wereusedfor uplandsiteswith
feather and sphagnummosseson clay parentmaterialto
690
RAPALEE ET AL.: SOIL C DYNAMICS IN A BOREAL FOREST LANDSCAPE
(A)
(C)
(e)
LFH
-.:.:a6s-s::
0
Ae
of
FIBRIC
Of
($phaõnum)
3O
i:C•;•" i Oh
240 &
Om
6O
and/
9O
WELL-DRAINED
.SAND
ß
[ c
:?'
•
POORLY D•INED
O•6A•IC/C•Y
Oh
VP - FEN
DEEP ORGANIC
Figure2. Typicalborealforestsoilprofilesfromthreesites:(a) well-drainedsand,(b) poorlydrainedorganicmaterialsandclay
soil,and(c)verypoorlydrained
fenwithdeeporganic
material.Profiledepthsarein centimeters.Soilprofilenomenclature
is from
theAgricultureCanadaExpertCommitteeon Soil Survey[1987] andSoil SurveyStaff[1975, 1996].
representboth vertical and lateral accumulationof moss
ments showthat mostof the C in the deeporganiclayershas
[Harden et al., 1997]. The radiocarbon-derived
I andk values
tumovertimesof centuriesor millennia [Trumbore and Harden,
wereusedforjack pinesiteson sandysoilsandfor wetlands,
whereno chronosequence
datawereavailable.
Althoughthe variouscomponents
of litter layers,suchas
leaves,
moss,androots,decompose
at differentrates,theuseof
a single
k valuedoesa reasonable
job of describing
thedecadal
process
ofaccumulation
of C in regrowing
mossanddetritusin
the BOREAS NorthernStudyArea [Trumboreand Harden,
1997;Hardenet al., 1997].Themajorityof themassof C in the
slow-growing
detritallayeris deadmoss,and the k values
derivedfrom field studieslargelydescribedecomposition
of
mossratherthantheotherlittercomponents,
whichconstitute
a
smaller fractionof the total C in the layer and which do not
continueto accumulate
substantially
overdecades.
1997]andthattheselayershaveaccumulatedslowly,influenced
by many fire eventsover the millennia sincethe drying of
GlacialLake Agassiz. Input I anddecomposition
k ratesused
for deep soil layers (Table 1) were derived using the
accumulation
ofC and]4Csince
deglaciation
[Trumbore
and
Harden, 1997].
We madethe assumption
thatdeepC poolsin uplandsoils
gain C immediatelyfollowingfire and experiencenet C loss
betweenfire eventsthroughdecomposition
[Harden et al.,
1997]. Oneexception
to thisassumption
wasmadefor thejack
pine sites, where Trumbore and Harden [1997] found
radiocarbon in the deep sandy soils from atmospheric
thermonuclear
weaponstestingduring the 1950s and 1960s.
This model assumes that the C stock in surface moss and
HenceC inputsmusthaveoccurred
to thesesoilhorizonsduring
detritusis zeroimmediatelyfollowingfire andthatinputrates thelast30 years,probablyfrom sourcesotherthanfire residues,
anddecomposition
constants
donotchangeovertimeasmoss suchas inputsof C frompinerootsandinputsfromdissolved
layers
regrowandaccumulate
C aftertire. Hardenet al. [1997] organiccarbon.
2.2.3. Wetland soils. Wetlandsoilscan alsobe split into
foundthatallowingthevalueof k to changewith timedid not
two layerswith distinctlydifferentC dynamics[Clymo,1984]:
significantly
improvethefit to chronosequence
dataoverthatof
a fixed k value.
(1) the surfacelayers,which includethe acrotelm(the aerobic
2.2.2. Upland deep soil organiclayers. The deepsoil portionof the soil profileabovethe watertable)and the upper
anaerobicportionbelow
layersare composed
of decomposed
organicmatterand the portionof thecatotelm(the submerged,
and
mineral A and B horizons. In contrastto the surfacelayers, thewatertable)whereorganicmaterialis still recognizable,
whicharemadeup of recognizable
plantfragments,
deepsoil (2) the deephorizonsof the catotelmwhereorganicmatteris
(Figure2c).
organicmatterconsistsof humifiedmaterial(dark-colored moredecomposed
The surfacelayerof the wetlandcan dry out or grow above
organicmatternotassociated
withmineralsin whichpiecesof
moss and litter are no longer identifiable),charcoal,and the water table. Decompositionratesof mossesin this upper
to thosein uplandsurfacelayers(Table 1).
mineral-associated
C. ThedeepsoilC is derivedfromorganic zonearecomparable
matterthatwasoriginallyaddedas surfacedetritalandmoss Forfensandcollapsescarbogs,we assumethat the surfacelayer
is pushed
layersbutwaslatertransferred
to deeporganiclayersthrough is at steadystateand that any C not decomposed
the accumulation of charred remnants of rites and the downward
below the water tableto be addedto the deeperlayers.
TrumboreandHarden[1997] foundthattheinventoryof the
percolation
of solubleorganicmaterial.Radiocarbon
measure-
RAPALEE ET AL.: SOIL C DYNAMICS
Table 1. InputI andDecomposition
k Constants
for
SurfaceandDeepSoil Layers
I,
k,
IN A BOREAL FOREST LANDSCAPE
691
presence or absenceof permafrost. Carbon inputs and
decompositioncoefficients(Table 1) were determinedfrom
radiocarbon
analysesand C inventorydatafor bothsurfaceand
deeplayers[Trumboreand Harden, 1997]. For the palsasand
peat plateaubogs, we set deep decomposition
rates to zero,
because
theselayersarefrozensolidyear-round.Wetlandsbum
onlyinfrequently,typicallydownto the watertable.
Soil
Horizon Class
Drainage
kgCm-2yr-1
Surface
Well
0.06
0.07
Moderatelywell
0.08
0.013
Imperfect
0.07
0.0105
The spatial base for our analysesis a soil polygonmap
[Veldhuisand Knapp, 199_8b]from an extensiveand detailed
resolution,
in rasterform,andusestheBOREAS grid coordinate
system(onthebasisof the AlbersEqual-AreaConicprojection)
in the 1983 North AmericanDatum (NAD83). Accompanying
thesoilpolygonrasterimageis an attributetablethatdescribes
in detail each componentof each respectivesoil polygon.
Spatialvariationis indicatedby estimatesof thepercentareaof
thetotalpolygonoccupiedby the dominantsoil seriesandeach
of the minor soil seriesinclusionswithin eachmap polygon.
2.3.1. Map of forest stand age. Using a classifiedLandsat
Thematic Mapper image from August 20, 1988 [Hall and
Knapp,1998],we identifiedthebum scarof the large 1981 fire
that encompassed
mostof the southwestportionof the study
area. The perimeterof the regenerationarea was screendigitizedandoverlaidontothesoilpolygonmap. Using 1994 as
the base year, we assigneda standage of 13 yearsto those
polygonsandportionsof polygonswithin the 1981 bum. We
assumedanyinclusionclassifiedas a fen or collapsescarbog
did notbum in thisor anyotherfire.
yr-1
Poor
soilsurvey
ofthe733km2 study
area.Themapis at30m
Sphagnum
moss
0.06
0.008
Palsa
0.08
0.013
Fen
0.0324 a
0.02 a
Collapse
scarbog
0.0324a
0.02a
Well
0.015
0.01
Moderatelywell
0
0.003
Imperfect
0
0.002
Sphagnum
moss
0
0.0007
Palsa
0
0
Very poor
Deep
2.3 Spatial Data
Poor
From the Canadian Forest Service series of hand-drawn fire
maps dating back to the 1930s, we determinedapproximate
Very poor
locationsof the majorbumsof 1956 and 1964. We thenused
digitalmapsfromtheNaturalResources
Manitoba(NRM) 1988
Fen
0.064
0.0004
forestinventory,which includesdata layersof speciescover,
cuttingclass,and siteclass(whichincludesagerangesfor the
Collapsescarbog
0.064
0.0004
threemajorforestcovertypes)[Knappand Tuinhoff,1998]. We
Information in this table is from Trumbore and Harden
thengenerated
a datalayerthatassignedan ageor agerangefor
[1997].
eachforeststand.Thesedatalayerswerethenusedto isolate30
to 40 yearold uplandjack pine standsthat occupythe northaThese
values
arefixed
foraCinventory
of13kgCm-2,so
south ridge in the easternportionof the studyarea and that
astogive0.064
kgCm-2yr-l,theinput
tothedeep
layers.
correspondto the sitesidentifiedin thehand-drawnCanadian
ForestServicefire mapsas the 1964 and 1956 bums.
For areas along Highway 391 (see Figure lb), which
surface
layers
offens
ranged
from
6 to13kgC m-2.Forthis essentiallyis an east-westtransectthroughthe studyarea,we
study,
wechosetheinventoryof thesurfacelayerto be 13 kg C
usedtreecoredatacollectedin theHalliwell and Apps [ 1996]
2
m' , becausethis
is
the
C
inventory
required
to
calculate
the
extensivebiometricsurveyof the BOREAS NorthernStudy.
ß
-2
-1
ß
observed
deep•nputvalueof 0.064kg C m yr , usingsurface TheHalliwellandApps[1995] tie-in pointswereoverlaidwith
the soil polygonmap, and a mean forest stand age was
values
ofI andk given
inTable1 (and/deep
=/surface
- (ksurface
foreachrelevantpolygon.Wherecoresweretakenat
X Csurface)
). This inventorycorresponds
to a depthof calculated
approximately
70 cm (Figure2c), wheretheorganicmatteris breastheight,we added10 yearsfor spruceand5 yearsforjack
visiblymoredecomposed
andwherebulk densityincreases.The pineandaspen(G. Peterson,personalcommunication,
1997)to
top70 cmincrement
includes
C thathasbeenfixedoverroughly the numberof recordedrings.
the last50-100 years[TrumboreandHarden, 1997].
To estimatestandagesfor the remainingareaswherethere
Below the water table, oxygen limitation slows wereneitherfirescarsnortreecoredata,we reliedon the ranges
decomposition
rates[Clymo,1984]. Unlikeuplandsoils,where of stand ages reportedin the Natural ResourcesManitoba
inputstothedeepsoilaretiedto fire,C is addedcontinuously
to (NRM) 1988 forest inventory [Knapp and Tuinhoff, 1998].
deeplayersof wetlandsas the growingsurfacelayerof moss Notingthatthemoresitespecificagesderivedfromthe fire scar
pushesorganicmatterbelowthewaterandasthewatertable andtreecoredatawereat thehighendof the agerangesreported
riseswiththeaccumulation
of the peatmaterial. Decomposition in the 1988 NRM inventory,we usedthe highendof the NRM
in thedeepwetlandsoilsis affectedby surfacedrainageandthe age ranges for all other areas. For areas that were not
692
RAPALEE
ET AL. ' SOIL C DYNAMICS
IN A BOREAL FOREST LANDSCAPE
(A) FORESTSTANDAGE
TIME,(YRS),
SINCEFIRE
•
13
30 - 43
45 - 65
65- 90
> 90
FEN
LAKE
N
METERS
5OO0
(B) SOIL DRAINAGE
DRAINAGE
CLASS
WELL
•.r• MOD WELL
[---I IMPERFECT
POOR
VERY POOR
ROCK
!--1 LAKE
N
METERS
50O0
Plate
1. (a)Forest
stand
agemapof733km2study
area
compiled
from
satellite
images,
firemaps,
forest
inventory,
andtreecore
data. Age rangesrepresenttime sincelast fire. The referenceyearis 1994. Fensarethosesoil polygonsfrom the Veldhuisand
Knapp[ 1998b]soil surveyfor which50% or moreof the areais classifiedas fen and/orcollapsescarbogandis assumednotto
have
burned.
(b)Soildrainage
mapof733km2study
area.Themaprepresents
soildrainage
bydominant
soilseries
ofsoil
polygonsfrom the Veldhuisand Knapp [ 1998b]soil survey.
RAPALEE ET AL. ßSOIL C DYNAMICS IN A BOREAL FOREST LANDSCAPE
inventoiledin the 1988 NRM survey,becauseof their
unproductive
wetlandstatus,we assumed
standageswerethe
sameascontiguous
standswith similarcharacteristics
for which
data were available.
The resultingforeststandagemap(Platel a) wasusedfor
modelinput.Finiteageswereassigned
to thebumedareas.We
(A) SURFACE LAYERS
693
assumed
thatfiresoccurinfrequently
in fensandcollapse
bogs
anddidnotassign
stand
ages.Forall otherpolygons,
a rangeof
probable
ageswasdetermined,
andthemidpoint
of thatrange
wasusedformodel
inputforeachinclusion
of eachsoilpolygon.
2.3.2. Map of soildrainage.Thesoildrainage
map(Plate
I b) represents
the drainageclassof the dominantsoilseriesof
kg C m'2
eachsoilpolygon,
usingdatacollected
in thesoilsurvey[VeldhuisandKnapp,1998b].Notshownonthismaparetheminor
inclusions
of othersoilserieswithineachpolygon,
butthese
[Z•30
accompanying
dataon thespatialextentof eachsoilseriesthat
---':•
.... 4-7
•
7-1o
•'•o-'•3
was used in model calculations are described below.
2.3.3. Carbonstockmaps. Forgenerating
thesoilC stock
maps(Figure
3),wefollowed
themethods
outlined
byDavidson
[1995] and Davidson and Lefebrve [1993], with a few
adaptations
to account
for differencesin the BOREAS datasets.
By employing
area-weighted
estimates
for soilpolygons
and
numericallyaccounting
for minorinclusionsof unnamedsoil
serieswithineachnamedsoilpolygon,soilseriesareincluded
that are relativelyunimportant
spatiallybut that may make
METERS
5060
significantcontributionsto soil C stocks[Davidsonand
Lefebrve,1993].
Datafi'omsoil pitswerenotusedto calculatesoilC stockof
the surface
layers,because
thedepthof thesurface
layersis
influenced more by time since the last fire than it is a
characterization
of a givensoilseries.Instead,surface
layerC
stocks
(kgCm':)were
calculated
foreach
fi'actional
component
-:• 15 - 25
•
•s - so
15o•os
•TER$
.......
5800
(i.e., soilseriesinclusion)
of eachsoilpolygon,
usingI andk
valuesappropriate
to eachdrainageclass(Table1) andt values
fi'omtheforest
stand
agesof thefirehistorymap(Plate1a). [See
alsoRapalee
etal., 1998.]Thearea-weighted
averages
foreach
polygonwerethenmapped(Figure3a).
The averagedeepsoil C inventoryfor eachsoil serieswas
calculated
fi'omfieldobservations
of bulkdensity
andC content
madeby Trumboreet al. [1998]and Veldhuis
and Knapp
[1998a]for humicand mineralhorizons(belowthe layerof
charredmaterial). Theseaverageswere used to calculate
area-weighted
deepC stocksforeachpolygon(Figure3b). For
about12%of thestudyareawhereno datawereavailablefor the
soilseries,we usedtheaverageC inventory
for deepsoilsfi'om
similar soil series.
•>0
........... - 15
•-'•
.... 15 - 30
Becausemanysoil profiledescriptions
hadmissingdataon
bulkdensity
(BD), we developed
nonlinear
regressions
relating
bulkdensity
toC content
forthesoilprofiledescriptions
thathad
both.Severalpublished
regressions
forpredicting
bulkdensity
fi'omC concentrations
[Huntington
et al., 1989;Curtisand Post,
1964;Alexander, 1989] could not be usedbecauseour data set
includedorganicsoilswith high C contentand mineralsoils
withverylowC content(0 to 30
I
>30 to 61
modestandis withintherangeof errorandinterannualvariation
reportedby Gouldenet al. [1998].
While studies
on interannualvariabilityof C fluxesin boreal
forests
haveconcentrated
on extrapolatingstand-levelresponses
of photosynthesis
and respirationto differencesin climatic
conditions
[e.g.Frolking et al., 1996; Frolking, 1997; Goulden
et al., 1998], our results show that significantinterannual
variability in regionalC exchangeratesmay also resultfrom
interannualdifferencesin the occurrence
of fire. For example,
the 1981 fire burnedapproximately17% of the studyarea(128
of733km2).Theaverage
soilC accumulation
ratein1981is
estimated
byourmodel
tohave
been
27gm-2.Assuming
that
buming
ofthemoss
layer
inthis128km2firereleased
2000gC
net
Cyr
flux
1981
was
a
loss
ofX
317
m-2thel•
m_•
yr_ the
(27
gCsoil
m-2
-1xin
0.83
- 2000
gnet
Crd
2y/l
0.17).
gC
For comparison,Frolking [1997] estimatedthat interannual
differences
in C-2 storage
at theOBS tower
sitevariedbetween0
-1
ß
.
and125g C m yr becauseof physiological
response
to interannual anomaliesin weather. Hence, averagedover large
regions,interannualvariabilityin C exchangemayresultfrom
changesin the lateralextentof fire as muchor morethandirect
responses
of the vegetationto climaticconditions.
Puttingthe importance
of fire anotherway, the releaseof C
fromthemosslayerby a firecovetingonly 1% of the areain any
year wouldoffsetthe amountof C storedin soilsof the restof
thearea(20g C m-2yr-l x 0.99--2000g C m-2yfflx 0.01).
The fire return interval
ß
.:..................
.[•:-:::..:-:--..
....
=======================
......
::
..... ..:•=•;.::'•:
...
50•0
Fibre5. Rates
ofsoilcarbon
accumulation
orloss
inthe733
km studyarea. The mapsrepresent
area-weighted
averages
of
(a) surfacelayers,includingmoss,(b) deeplayers,and(c) total
profile. Seetextfor discussion
of surfaceanddeeplayers.
for Alaskan
boreal forests has been
estimated
to be 150 years[Kasischkeet al., 1995], with a range
in variousNorth Americanborealforesttypesof 70-500 years
[Payette,1993]. If thefire cyclewereabout100 years,thenour
carbonbudgetfor a yearwith no fire mightsuggestthatthe soil
C stockmayberoughlyat steadystateoverlargeareasand long
timescales. Over millennial timescales,soil C storagerates
sincethedry•.ng
of GlacialLakeAgassizareof theorderof only
-2
-1
a fewg C m yr [Trumboreand Harden, 1997;Harden et al.,
1997]. However,theincidence
of fire is variableoverspaceand
time andcouldbe changingin responseto climatechangeand
managementpractices.For an annualC budgetof a localsite,
the time since the last fire is critical.
For decadal or centennial
scalebudgetsoverthe region,fire frequencyandits variability
overthe landscapeareclearlyimportant.
A majoruncertainty
in determining
thepresentandfuturerole
4. Discussion
of soilsas C sourcesor sinksis the role of the deeporganic
ratesof
The modelestimates
that average2net soil C storagewas matterin theC budgetof uplandsoils. Decomposition
2
I
about20 g C m- yr- overthe 733 km studyareaof northern deeporganiclayersare an orderof magnitudeslowerthanthose
in surfacelayers(Table 1), whichmightleadto theassumption
boreal forest in 1994 (Table 3), which is a modest sink of
atmospheric
CO2. Usinganeddycovariance
method,Goulden thatthedeepsoil is not importantin C cycling. However,large
698
RAPALEE ET AL. ßSOIL C DYNAMICS IN A BOREAL FOREST LANDSCAPE
(A) SURFACE LAYERS
8O
ß-:::=:.-.:•:.."•:.:.
-•.'::i;•.
ø-•':"""••i:•
.........
:.i•
.......
•::""
.
'E 40
o 20
0
.........
!=•
=i.-•,=•:•
;;.....
or- Palsa
Poor- Sphagnum
.•-•
,•.•,:;:.:.,;..
.....•,•
.•,••i:
"'•"•":'••._.•
Imperfect moss
133:•":•34'•-•.•.•.
•'•'""":•'=::••Mod.
well
s5•o-'>•• Well
Time (Yrs.) Since Fire
(B) DEEP LAYERS
..
-40• .,..::•,::•:,
...•"•"•..•
• .• ?:/••---.•.....
..... ..:
....
•.....
:......
:?:.:.-•.,,•:-•
•.•.,....,
................
Well
0 •..••
••••'•
............
P ' -Sphagnummoss
•' •e•-•0 ••
>
Poor - Palsa
Time (Yrs.) Since Fire
Figure6. (a)Rates
ofC accumulation
(positive
values)
in surface
layers
(including
moss)
fordrainage
class
andbytimesince
fire
(stand
age).(b)Rates
ofCloss
(negative
values)
from
deep
organic
layers
(below
moss)
andmineral
soila_•
a-lfunction
of
- soil
drainage
andtimesince
fire.Fensandcollapse
scar
bogs
(notshown)
aregaining
anestimated
29 and12g C m yr , respectively,
in the deeplayers.
storesof C are presentin thedeeplayers,soevenat slowrates
of decomposition,
losseson an annualbasisaresignificant,
and
interannualdifferencesin decomposition
of deep soil C can
affect the C balance of a whole forest stand [Goutden et at.,
1998]. We haveassumedthat the deepsoil C poolsshownet
losses
of C in betweenfireevents.Comparisonof soilC balance
with eddycovariance
measurements
of NEP [Hardenet at.,
1997], evidenceof increasesin late summersoil respiration
[Goutdenet at., 1998], and radiocarbonmeasurements
in soil
CO2 [Winston
et at., 1997]all support
thishypothesis.
Considerableuncertaintiesremain about the origin and
dynamics
of thisimportant
deepsoilC pool. Theradiocarbon
datausedby TrumboreandHarden[ 1997]to calculateinputs
anddecomposition
constants
mayhavebeeninfluenced
moreby
the frequencyand severity of buming than by actual
decomposition
rates of the humic materials. Incubation
measurements
[Fries et el., 1997; Moore and Knowlee,1990]
showthatdeeporganicC decomposes
rapidlyif warmedand
dried.Thesefindings
suggest
thattheslowdecomposition
rates
we inferredfromradiocarbon
studiesare constrained
by physical
RAPALEE
ET AL. ßSOIL C DYNAMICS
IN A BOREAL FOREST LANDSCAPE
699
15.0
10.0
o 5.0
Poor
$phagnurn P•or
Mod Well Imperfect
moss
DRAINAGE
Paisa
VeryPoor •
Fen & Bog Total Area
CLASS
Figure
7. Total
annual
Csoilaccumulation
(>0)orloss
(
Soil carbon stocks and their rates of accumulation
and loss
in a boreal forest landscape
Gloria
Rapalee
l'2,3,Susan
E.Trumbore
4,EricA.Davidson
l,
Jennifer
W.Harden
5,andHugo
Veldhuis6
Abstract. Borealforestsandwetlandsare thoughtto be significantcarbonsinks,andthey
couldbecomenet C sourcesas theEarth warms. Most of the C of borealforestecosystems
is
storedin themosslayerandin thesoil. The objectiveof thisstudywasto estimatesoilC
stocks
(including
moss
layers)
and
rates
ofaccumulation
andloss
fora733km2area
ofthe
BORealEcosystem-Atmosphere
Studysitein northernManitoba,usingdatafrom smaller-scale
intensivefield studies.A simpleprocess-based
modeldevelopedfrommeasurements
of soilC
inventoriesandradiocarbonwasusedto relatesoilC storageanddynamicsto soil drainageand
foreststandage. Soil C stockscovarywith soildrainageclass,with thelargestC stocks
occurringin poorlydrainedsites.Estimatedratesof soilC accumulation
or lossare sensitive
to theestimateddecomposition
constants
for the largepoolof deepsoilC, andimproved
understanding
of deepsoilC decomposition
is needed.While the uppermosslayersregrow
andaccumulate
C after fires,thedeepC dynamicsvary acrossthe landscape,
from a smallnet
sink
toasignificant
source.
Estimated
ne_2t
soi{
Caccumulation,
averaged
fortheentire
733
km2area,
was20gCrn-2yr-• (28gCrn yr- accumulation
insurface
mosses
offset
by8gC
rn yr lostfromdeepC pools)in a yearwith no fire. Most of theC accumulated
in poorly
andverypoorlydrainedsoils(peatlandsandwetlands).Burningof themosslayer in only 1%
of uplandswouldoffsetthe C storedin the remaining99% of the area. Significantinterannual
variabilityin C storageis expectedbecauseof the irregularoccurrence
of fire in spaceand
time. The effectsof climatechangeandmanagement
on fire frequencyandon decomposition
of immensedeepsoilC stocksare key to understanding
futureC budgetsin borealforests.
Inverse models that calculate the latitudinal
1. Introduction
distribution
of
CO2sources
and
sinks
from
observed
CO2 and13CO2
distribu-
Boreal forestsplay an importantrole in the global carbon
budget.
Ofthe1500
petagrams
(1Pg= 10•Sg)ofC
stored
in
soils globally and 600 Pg of C in aboveground
biomass[Reeburgh,1997],aboutonethirdof thisglobalterrestrialC is found
in thebiomass,detritus,soil, and peatpoolsof the borealforest
biome, with about three fourths of the C in boreal forest biome
storedbelowground[Appset al., 1993; Peng et al., 1998].
Predictinghow thesevaststoresof soil C in the borealforest
biome will be affectedby future climate changerequiresan
understanding
of the factorscontrollingthe production,decomposition,and storageof organicC in borealecosystems.
tions suggestthat borealregionsmay be significantC sinks
[Tanset al., 1990; Ciais et al., 1995]. An alternativeapproach
to determining
C exchangeat the landscapeor regionalscaleis
to developprocessmodelsof C dynamicsand to use those
models,togetherwith a knowledgeof the spatialdistributionof
importantforcingfactors,to extrapolatefield measurements
of
C losses and gains to regional scales. Because it is
process-based,
the latterapproachmay alsobe usedto predict
theresponse
of the landscapeto changesin forcingparameters.
Responses
of borealforestsoilsto warming,changesin drainage, or changesin fire frequencyhaveall beenproposedto be
importantfor terrestrialC storage[Bonan, 1993; Moore and
•TheWoods
HoleResearch
Center,
Woods
Hole,Massachusetts.
Knowles, 1990; Gorham, 1991; Kasischke et al., 1995; Kurz
:Department
of Geography,
SalemStateCollege,
Salem, andApps,1995].
Ma•ssachusetts.
ONow with Departmentof Earth SystemScience,Universityof
California,
Ëvine.
Departmentof Earth SystemScience,Universityof California,
Irvine.
56U.S.
Geological
Survey,
Menlo
Park,
California.
Agricultureand Agri-Food Canada,BrandonResearchCentre,
We usedfieldstudiesof C input,storage,andturnoverin the
northem
regionof theborealforest,madeaspartof theBOReal
Ecosystem-Atmosphere
Study(BOREAS), to developsimple
modelslinking soilC storageandratesof accumulation
to two
major factors: soil drainageclassand the time sincelast fire
Manitoba Land ResourceUnit, Universityof Manitoba, Winnipeg,
[Trumbore and Harden, 1997; Harden et al., 1997]. These
Canada.
modelshavepreviously
beencombined
with a soildrainagemap
for a 120 year old black sprucestand to comparethe soil
component
of net C storagewith tower-based,
eddycovariance
measurementsof net ecosystemproduction[Harden et al.,
1997]. In this paper, we expand this scaling approachto
Copyright1998 by the AmericanGeophysical
Union.
Papernumber98GB02336.
0886-6236/98/98GB-02336512.00
687
688
RAPALEE ET AL. ßSOIL C DYNAMICS
IN A BOREAL FOREST LANDSCAPE
estimate
thesoilCstorage
andaccumulation
rates
fora733km2
2. Methods and Data Sources
area within the BOREAS Northern Study Area (NSA), in
2.1. Study Area
northern
Manitoba,Canada.More specifically,we (1) estimate
totalC stocksby horizonfor commonsoil serieson the basisof
The BOREAS Northern Study Area, located between
soil surveydata and analysesof data from individual soil Thompsonand NelsonHouse,Manitoba, is nearthe northern
profiles; (2) estimateratesof soil C accumulationon the basis limitof theclosed-crown
borealfore_st
(Figure1a) [Sellerset al.,
of C stocksanda simplemodelof C turnoverderivedfrom field
1994].
Wefocus
hereona 733km2study
areawithin
theNSA
radiocarbon
studies;(3) relatepatternsof C stocksandaccumu- (Figure 1b), for whicha soil surveymapwasgeneratedduring
lationratestopatterns
of drainage,mosscover,andfire history; the 1994BOREAS field season[Veldhuisand Knapp, 1998b].
and (4) generatearea-weightedmap•sof soil C stocksand Cornercoordinates
of the studyareaare: 56.00øN, 98.69øW;
accumulation
rates
across
the733km2study
areain 1994,the
56.00øN, 98.17øW; 55.80øN, 98.18øW; 55.80øN, 98.69øW.
yearin whichmostBOREAS field studieswere conducted.We
alsoidentifyareasof greatest
sensitivityanduncertainty
in these
Much of the regionis underlainby poorlydrained,varved
clay depositedby Glacial Lake Agassizand interspersed
with
frequentbedrockoutcrops. A few minor exposuresof well-
estimates.
...-Canadian
BorealForestMap
ß
•.."
(A)
L •X •-;;-.
• •
•.
• x" -.......
:
-.•:i:
.•
:
..":':)"
.. ':•..'
::.:::.::.:::.
:..::.:...•:::..•..
..... .......,,.-.............
...
(B)
....
•:. ...:
.....•,..•
..•'" •
'::"
:....
''......
,,
Nx
••
..' •icp•rt
.
•::
• o• . '
Figure
1. (a)MapofCanadian
boreal
forests
sh2owing
theBOREAS
study
region
andthenorthern
andsouthern
sites
and(b)map
of BOREASNorthernStudyArea. The733 km studyareais drawnin black. The eddyflux towersitesare: Old JackPine (OJP),
Old Black Spruce(OBS), Fen (FEN), YoungJackPine (YJP), andBeaverPond(BP). Thesemapsweremodifiedfrom those
publiclyavailablecourtesyof BOREAS (http://boreas.gsfc.nasa.gov).
RAPALEE ET AL.: SOIL C DYNAMICS
drained,
sandyglacialtill occur,largelyin theeastempartof the
studyarea,wheretwomajorhillscomposed
of sandandgravel
(kamedeposits)
runnorth-south
with reliefup to 60 m [Sellers
IN A BOREAL FOREST LANDSCAPE
689
[Bonan, 1993; Moore and Knowles, 1990; Gorham, 1991;
Kasischke
etal.,1995•Kurz
andApps,
1995].
A 1981
fire
burnedabout128 km of the predominantly
spruce-forested
et al., 1994]. (SeeFigure2a andPlate 1b.) Extensiveareasof
southernpart of the studyarea. (SeePlate 1a.) Smallerbums
the clayeydeposits
areoverlainby deepandshalloworganic
deposits
(peat). Permafrost
occurssporadically
andat variable
depths
in clayeyuplandsoilsandin veneerbogs[Sellerset al.,
1994; Rubec, 1988; Veldhuisand Knapp, 1998b], and is
continuous
in collapsescarandpeatplateaubogsandpalsas
[Rubec,1988; Veldhuisand Knapp, 1998b]. The area is
undulatingwith longgentleslopes.Foreststandagesrange
from 13 to 140 yearsor moreon uplands,markingtheperiod
totaling
about
33km2 occurred
in 1956and1964along
the
elapsedsincethe lastfire.
Stratifyingthe studyareaby drainageclassis an effective
meansof identifyinglandcovertype. We identifiedsix broad
land cover types on the basis of similaritiesin landform,
drainage,
depthof organics
(peat),anddominantforestandmoss
cover.In thefollowingdescriptions,
theCanadiansoilclassification[AgricultureCanadaExpertCommitteeon SoilSurvey,
1987] is followed (in parentheses)
by the U.S. classification
[Soil SurveyStaff, 1975, 1996 (also availableon the World
WideWebat http://www.statlab.iastate.edu/soils/keytax/)]:
(1)
upland, well-drained, sandy Eluviated Dystric Brunisols
(Cryochrepts)with jack pine (Pinus banksiana)cover; (2)
upland,moderatelywell-drained,clayeyOrthicGrayLuvisols
(Cryoboralfs)with mixed black spruce(Picea mariana) and
hardwood
standswith feathermoss(Pleurozium,Hylocomium
spp.)asthedominant
groundcover;(3) transitional,
imperfectly
to poorly drained, clayey Gleyed Gray Luvisols and Luvic
Gleysols(Cryoboralfs)with black spruceand a mixtureof
featherandsphagnum(Sphagnum
spp.)mosses;(4) poorlyto
verypoorlydrained,levelto gentlysloping,clayey,peatyLuvic
Gleysolsand Terdc Mesic Fibrisols(Cryoboralfsand Cryofibrists) with black spruceand sphagnummossmixed with
featherandothermosses;(5) peatlands
consisting
of varying
peatmaterials
thatarewell- to poorlydrainedat the surfaceand
havefrozenpeatand/ormineralmaterialat depth,with Fibric
and Mesic OrganicCryosols(Cryofibristsand Cryohemists),
andblackspruceandsphagnum/feather
mossvegetation
cover.
(Palsas
aremounds
ofuplandpeatlands
underlainby perennially
frozenpeatandmineralsoil,with roughandunevensurfaces,
andarecharacterized
by havingdomedsurfaces.Peatplateau
bogs,alsounderlainby continuous
permafrost,
rise abruptly
fromsurrounding
unfrozen
fensandarecharacterized
by having
relatively flat and even surfaces[Rubec,1988; Veldhuisand
Knapp,1998b]); and(6) verypoorlydrainedfensandpermafrostcollapse
bogswith deepTypicFibrisols(Cryofibrists),
and
sedgeandbrownmossvegetation.(Fensarewetlandsfed by
groundwaterthat is relativelynutrient rich (comparedto
precipitation),
whereas
bogsareoligotrophic,
havingtheirupper
peat layer (rootingzone) abovethe groundwaterflow, and
therefore
havea rootingzonelargelydependent
onprecipitation
for its water and nutrientsupply. Vegetationof the fensis
characteristically
sedge(Carex,Eriophorumspp.)andbrown
moss(Depranocladus,
Tomenthypnum
spp.). Sphagnum
and/or
feathermosses
occurin bogs.Collapsescarbogsarenutrient
poor circularor ovaldepressions
causedby subsidence
of the
peatlandsurfacedue to thawingof permafrost[Rubec,1988;
Veldhuisand Knapp,1998b]).
In additionto soil drainage,the incidenceof fire is an
important
factorcontrollingannualaccumulation
ratesof soilC
north-south
ridgeneartheeastern
boundary
whereprimarilyjack
pinehasregenerated.
In 1989,largeareasburnedin thewestern
and northernsections
of the NSA, outsideof the studyarea.
Noneof the studyregionburnedin 1994.
2.2. Soil C Storage
Modelsof soilC dynamics
weredeveloped
usingdataonC
inventories
andradiocarbon
aspartof theBOREAS field effort
[Trumbore
andHarden,1997; Hardenet al., 1997]. Upland
soil profilesweredividedintotwo layers(Figures2a and2b)
that aredistinctlydifferentin theirC dynamics:
(1) a surface
detritalor mosslayer,madeup of relativelyundecomposed
organic
material,
and(2) a deeplayerconsisting
of morehighly
decomposed
organicmatter(humiclayer),in which plant
macrofossils
are rare,and whereminoramounts
of organic
matter are incorporatedinto the mineralsoil A horizon. The
mineralB horizonwasalsoincluded
in thisdeeplayer.
2.2.1. Uplandsurfacesoillayers. Uplandsurface
layers
accumulate
C between
fireevents.OrganicC is inputdirectly
tothesurface
layerasdetritus
fromtreesandmosses.
Decompositionratesaregenerally
slow(meanresidence
timesof theorder
of decades),
sothatthickorganicmatsaccumulate
in shallow
layers before lossesthroughdecomposition
offset annual
additions.Feathermosscoveron moderately
well- to imperfectlydrainedsoilstendsto dryoutin summerbecause
it does
notoverlie
saturated
soilsandhencewill bumnearlycompletely
to mineralsoilin stand-killing
fires. In contrast,
sphagnum
mosscoveris associated
with imperfectly
andpoorlydrained
soils that tendto staymoistat depthandthereforebum less
completely.
We modeledsoil C accumulation
of the surfacelayers
betweenfiresas a simplebalanceof C inputsandfirst-order
decomposition
usingthefollowingequations
fromJennyet al.
[ 1949] and Harden et al. [ 1992, 1997]:
dC/dt= I- (k x C)
(1)
Solvingfor C in Equation(1) yields:
C(t) = I/k x (1 - e-kt)
(2)
whereC iscarbon
massin unitsof massperarea,t is numberof '
years
sincethelaststand-killing
fire,I is inputratein massper
unitarea
•per
year,
and
kisadecomposition
coefficient
inunits
of years-.
Input I ratesanddecomposition
k constants
(Table1) were
determined
usingtwo approaches.
First,verticalaccumulation
ratesof mosswereobtainedusingradiocarbon
analyses
to
determinethe ageof accumulated
C [Trumboreand Harden,
1997]. Second,surveys
of theC inventory
in organicmatter
abovethe mostrecentcharcoallayerin the soil profilewere
madeacrossa seriesof sites(chronosequence)
thatdifferedin
timesincelastfire [Hardenet al., 1997]. The estimatesof I and
k basedon chronosequences
wereusedfor uplandsiteswith
feather and sphagnummosseson clay parentmaterialto
690
RAPALEE ET AL.: SOIL C DYNAMICS IN A BOREAL FOREST LANDSCAPE
(A)
(C)
(e)
LFH
-.:.:a6s-s::
0
Ae
of
FIBRIC
Of
($phaõnum)
3O
i:C•;•" i Oh
240 &
Om
6O
and/
9O
WELL-DRAINED
.SAND
ß
[ c
:?'
•
POORLY D•INED
O•6A•IC/C•Y
Oh
VP - FEN
DEEP ORGANIC
Figure2. Typicalborealforestsoilprofilesfromthreesites:(a) well-drainedsand,(b) poorlydrainedorganicmaterialsandclay
soil,and(c)verypoorlydrained
fenwithdeeporganic
material.Profiledepthsarein centimeters.Soilprofilenomenclature
is from
theAgricultureCanadaExpertCommitteeon Soil Survey[1987] andSoil SurveyStaff[1975, 1996].
representboth vertical and lateral accumulationof moss
ments showthat mostof the C in the deeporganiclayershas
[Harden et al., 1997]. The radiocarbon-derived
I andk values
tumovertimesof centuriesor millennia [Trumbore and Harden,
wereusedforjack pinesiteson sandysoilsandfor wetlands,
whereno chronosequence
datawereavailable.
Althoughthe variouscomponents
of litter layers,suchas
leaves,
moss,androots,decompose
at differentrates,theuseof
a single
k valuedoesa reasonable
job of describing
thedecadal
process
ofaccumulation
of C in regrowing
mossanddetritusin
the BOREAS NorthernStudyArea [Trumboreand Harden,
1997;Hardenet al., 1997].Themajorityof themassof C in the
slow-growing
detritallayeris deadmoss,and the k values
derivedfrom field studieslargelydescribedecomposition
of
mossratherthantheotherlittercomponents,
whichconstitute
a
smaller fractionof the total C in the layer and which do not
continueto accumulate
substantially
overdecades.
1997]andthattheselayershaveaccumulatedslowly,influenced
by many fire eventsover the millennia sincethe drying of
GlacialLake Agassiz. Input I anddecomposition
k ratesused
for deep soil layers (Table 1) were derived using the
accumulation
ofC and]4Csince
deglaciation
[Trumbore
and
Harden, 1997].
We madethe assumption
thatdeepC poolsin uplandsoils
gain C immediatelyfollowingfire and experiencenet C loss
betweenfire eventsthroughdecomposition
[Harden et al.,
1997]. Oneexception
to thisassumption
wasmadefor thejack
pine sites, where Trumbore and Harden [1997] found
radiocarbon in the deep sandy soils from atmospheric
thermonuclear
weaponstestingduring the 1950s and 1960s.
This model assumes that the C stock in surface moss and
HenceC inputsmusthaveoccurred
to thesesoilhorizonsduring
detritusis zeroimmediatelyfollowingfire andthatinputrates thelast30 years,probablyfrom sourcesotherthanfire residues,
anddecomposition
constants
donotchangeovertimeasmoss suchas inputsof C frompinerootsandinputsfromdissolved
layers
regrowandaccumulate
C aftertire. Hardenet al. [1997] organiccarbon.
2.2.3. Wetland soils. Wetlandsoilscan alsobe split into
foundthatallowingthevalueof k to changewith timedid not
two layerswith distinctlydifferentC dynamics[Clymo,1984]:
significantly
improvethefit to chronosequence
dataoverthatof
a fixed k value.
(1) the surfacelayers,which includethe acrotelm(the aerobic
2.2.2. Upland deep soil organiclayers. The deepsoil portionof the soil profileabovethe watertable)and the upper
anaerobicportionbelow
layersare composed
of decomposed
organicmatterand the portionof thecatotelm(the submerged,
and
mineral A and B horizons. In contrastto the surfacelayers, thewatertable)whereorganicmaterialis still recognizable,
whicharemadeup of recognizable
plantfragments,
deepsoil (2) the deephorizonsof the catotelmwhereorganicmatteris
(Figure2c).
organicmatterconsistsof humifiedmaterial(dark-colored moredecomposed
The surfacelayerof the wetlandcan dry out or grow above
organicmatternotassociated
withmineralsin whichpiecesof
moss and litter are no longer identifiable),charcoal,and the water table. Decompositionratesof mossesin this upper
to thosein uplandsurfacelayers(Table 1).
mineral-associated
C. ThedeepsoilC is derivedfromorganic zonearecomparable
matterthatwasoriginallyaddedas surfacedetritalandmoss Forfensandcollapsescarbogs,we assumethat the surfacelayer
is pushed
layersbutwaslatertransferred
to deeporganiclayersthrough is at steadystateand that any C not decomposed
the accumulation of charred remnants of rites and the downward
below the water tableto be addedto the deeperlayers.
TrumboreandHarden[1997] foundthattheinventoryof the
percolation
of solubleorganicmaterial.Radiocarbon
measure-
RAPALEE ET AL.: SOIL C DYNAMICS
Table 1. InputI andDecomposition
k Constants
for
SurfaceandDeepSoil Layers
I,
k,
IN A BOREAL FOREST LANDSCAPE
691
presence or absenceof permafrost. Carbon inputs and
decompositioncoefficients(Table 1) were determinedfrom
radiocarbon
analysesand C inventorydatafor bothsurfaceand
deeplayers[Trumboreand Harden, 1997]. For the palsasand
peat plateaubogs, we set deep decomposition
rates to zero,
because
theselayersarefrozensolidyear-round.Wetlandsbum
onlyinfrequently,typicallydownto the watertable.
Soil
Horizon Class
Drainage
kgCm-2yr-1
Surface
Well
0.06
0.07
Moderatelywell
0.08
0.013
Imperfect
0.07
0.0105
The spatial base for our analysesis a soil polygonmap
[Veldhuisand Knapp, 199_8b]from an extensiveand detailed
resolution,
in rasterform,andusestheBOREAS grid coordinate
system(onthebasisof the AlbersEqual-AreaConicprojection)
in the 1983 North AmericanDatum (NAD83). Accompanying
thesoilpolygonrasterimageis an attributetablethatdescribes
in detail each componentof each respectivesoil polygon.
Spatialvariationis indicatedby estimatesof thepercentareaof
thetotalpolygonoccupiedby the dominantsoil seriesandeach
of the minor soil seriesinclusionswithin eachmap polygon.
2.3.1. Map of forest stand age. Using a classifiedLandsat
Thematic Mapper image from August 20, 1988 [Hall and
Knapp,1998],we identifiedthebum scarof the large 1981 fire
that encompassed
mostof the southwestportionof the study
area. The perimeterof the regenerationarea was screendigitizedandoverlaidontothesoilpolygonmap. Using 1994 as
the base year, we assigneda standage of 13 yearsto those
polygonsandportionsof polygonswithin the 1981 bum. We
assumedanyinclusionclassifiedas a fen or collapsescarbog
did notbum in thisor anyotherfire.
yr-1
Poor
soilsurvey
ofthe733km2 study
area.Themapis at30m
Sphagnum
moss
0.06
0.008
Palsa
0.08
0.013
Fen
0.0324 a
0.02 a
Collapse
scarbog
0.0324a
0.02a
Well
0.015
0.01
Moderatelywell
0
0.003
Imperfect
0
0.002
Sphagnum
moss
0
0.0007
Palsa
0
0
Very poor
Deep
2.3 Spatial Data
Poor
From the Canadian Forest Service series of hand-drawn fire
maps dating back to the 1930s, we determinedapproximate
Very poor
locationsof the majorbumsof 1956 and 1964. We thenused
digitalmapsfromtheNaturalResources
Manitoba(NRM) 1988
Fen
0.064
0.0004
forestinventory,which includesdata layersof speciescover,
cuttingclass,and siteclass(whichincludesagerangesfor the
Collapsescarbog
0.064
0.0004
threemajorforestcovertypes)[Knappand Tuinhoff,1998]. We
Information in this table is from Trumbore and Harden
thengenerated
a datalayerthatassignedan ageor agerangefor
[1997].
eachforeststand.Thesedatalayerswerethenusedto isolate30
to 40 yearold uplandjack pine standsthat occupythe northaThese
values
arefixed
foraCinventory
of13kgCm-2,so
south ridge in the easternportionof the studyarea and that
astogive0.064
kgCm-2yr-l,theinput
tothedeep
layers.
correspondto the sitesidentifiedin thehand-drawnCanadian
ForestServicefire mapsas the 1964 and 1956 bums.
For areas along Highway 391 (see Figure lb), which
surface
layers
offens
ranged
from
6 to13kgC m-2.Forthis essentiallyis an east-westtransectthroughthe studyarea,we
study,
wechosetheinventoryof thesurfacelayerto be 13 kg C
usedtreecoredatacollectedin theHalliwell and Apps [ 1996]
2
m' , becausethis
is
the
C
inventory
required
to
calculate
the
extensivebiometricsurveyof the BOREAS NorthernStudy.
ß
-2
-1
ß
observed
deep•nputvalueof 0.064kg C m yr , usingsurface TheHalliwellandApps[1995] tie-in pointswereoverlaidwith
the soil polygonmap, and a mean forest stand age was
values
ofI andk given
inTable1 (and/deep
=/surface
- (ksurface
foreachrelevantpolygon.Wherecoresweretakenat
X Csurface)
). This inventorycorresponds
to a depthof calculated
approximately
70 cm (Figure2c), wheretheorganicmatteris breastheight,we added10 yearsfor spruceand5 yearsforjack
visiblymoredecomposed
andwherebulk densityincreases.The pineandaspen(G. Peterson,personalcommunication,
1997)to
top70 cmincrement
includes
C thathasbeenfixedoverroughly the numberof recordedrings.
the last50-100 years[TrumboreandHarden, 1997].
To estimatestandagesfor the remainingareaswherethere
Below the water table, oxygen limitation slows wereneitherfirescarsnortreecoredata,we reliedon the ranges
decomposition
rates[Clymo,1984]. Unlikeuplandsoils,where of stand ages reportedin the Natural ResourcesManitoba
inputstothedeepsoilaretiedto fire,C is addedcontinuously
to (NRM) 1988 forest inventory [Knapp and Tuinhoff, 1998].
deeplayersof wetlandsas the growingsurfacelayerof moss Notingthatthemoresitespecificagesderivedfromthe fire scar
pushesorganicmatterbelowthewaterandasthewatertable andtreecoredatawereat thehighendof the agerangesreported
riseswiththeaccumulation
of the peatmaterial. Decomposition in the 1988 NRM inventory,we usedthe highendof the NRM
in thedeepwetlandsoilsis affectedby surfacedrainageandthe age ranges for all other areas. For areas that were not
692
RAPALEE
ET AL. ' SOIL C DYNAMICS
IN A BOREAL FOREST LANDSCAPE
(A) FORESTSTANDAGE
TIME,(YRS),
SINCEFIRE
•
13
30 - 43
45 - 65
65- 90
> 90
FEN
LAKE
N
METERS
5OO0
(B) SOIL DRAINAGE
DRAINAGE
CLASS
WELL
•.r• MOD WELL
[---I IMPERFECT
POOR
VERY POOR
ROCK
!--1 LAKE
N
METERS
50O0
Plate
1. (a)Forest
stand
agemapof733km2study
area
compiled
from
satellite
images,
firemaps,
forest
inventory,
andtreecore
data. Age rangesrepresenttime sincelast fire. The referenceyearis 1994. Fensarethosesoil polygonsfrom the Veldhuisand
Knapp[ 1998b]soil surveyfor which50% or moreof the areais classifiedas fen and/orcollapsescarbogandis assumednotto
have
burned.
(b)Soildrainage
mapof733km2study
area.Themaprepresents
soildrainage
bydominant
soilseries
ofsoil
polygonsfrom the Veldhuisand Knapp [ 1998b]soil survey.
RAPALEE ET AL. ßSOIL C DYNAMICS IN A BOREAL FOREST LANDSCAPE
inventoiledin the 1988 NRM survey,becauseof their
unproductive
wetlandstatus,we assumed
standageswerethe
sameascontiguous
standswith similarcharacteristics
for which
data were available.
The resultingforeststandagemap(Platel a) wasusedfor
modelinput.Finiteageswereassigned
to thebumedareas.We
(A) SURFACE LAYERS
693
assumed
thatfiresoccurinfrequently
in fensandcollapse
bogs
anddidnotassign
stand
ages.Forall otherpolygons,
a rangeof
probable
ageswasdetermined,
andthemidpoint
of thatrange
wasusedformodel
inputforeachinclusion
of eachsoilpolygon.
2.3.2. Map of soildrainage.Thesoildrainage
map(Plate
I b) represents
the drainageclassof the dominantsoilseriesof
kg C m'2
eachsoilpolygon,
usingdatacollected
in thesoilsurvey[VeldhuisandKnapp,1998b].Notshownonthismaparetheminor
inclusions
of othersoilserieswithineachpolygon,
butthese
[Z•30
accompanying
dataon thespatialextentof eachsoilseriesthat
---':•
.... 4-7
•
7-1o
•'•o-'•3
was used in model calculations are described below.
2.3.3. Carbonstockmaps. Forgenerating
thesoilC stock
maps(Figure
3),wefollowed
themethods
outlined
byDavidson
[1995] and Davidson and Lefebrve [1993], with a few
adaptations
to account
for differencesin the BOREAS datasets.
By employing
area-weighted
estimates
for soilpolygons
and
numericallyaccounting
for minorinclusionsof unnamedsoil
serieswithineachnamedsoilpolygon,soilseriesareincluded
that are relativelyunimportant
spatiallybut that may make
METERS
5060
significantcontributionsto soil C stocks[Davidsonand
Lefebrve,1993].
Datafi'omsoil pitswerenotusedto calculatesoilC stockof
the surface
layers,because
thedepthof thesurface
layersis
influenced more by time since the last fire than it is a
characterization
of a givensoilseries.Instead,surface
layerC
stocks
(kgCm':)were
calculated
foreach
fi'actional
component
-:• 15 - 25
•
•s - so
15o•os
•TER$
.......
5800
(i.e., soilseriesinclusion)
of eachsoilpolygon,
usingI andk
valuesappropriate
to eachdrainageclass(Table1) andt values
fi'omtheforest
stand
agesof thefirehistorymap(Plate1a). [See
alsoRapalee
etal., 1998.]Thearea-weighted
averages
foreach
polygonwerethenmapped(Figure3a).
The averagedeepsoil C inventoryfor eachsoil serieswas
calculated
fi'omfieldobservations
of bulkdensity
andC content
madeby Trumboreet al. [1998]and Veldhuis
and Knapp
[1998a]for humicand mineralhorizons(belowthe layerof
charredmaterial). Theseaverageswere used to calculate
area-weighted
deepC stocksforeachpolygon(Figure3b). For
about12%of thestudyareawhereno datawereavailablefor the
soilseries,we usedtheaverageC inventory
for deepsoilsfi'om
similar soil series.
•>0
........... - 15
•-'•
.... 15 - 30
Becausemanysoil profiledescriptions
hadmissingdataon
bulkdensity
(BD), we developed
nonlinear
regressions
relating
bulkdensity
toC content
forthesoilprofiledescriptions
thathad
both.Severalpublished
regressions
forpredicting
bulkdensity
fi'omC concentrations
[Huntington
et al., 1989;Curtisand Post,
1964;Alexander, 1989] could not be usedbecauseour data set
includedorganicsoilswith high C contentand mineralsoils
withverylowC content(0 to 30
I
>30 to 61
modestandis withintherangeof errorandinterannualvariation
reportedby Gouldenet al. [1998].
While studies
on interannualvariabilityof C fluxesin boreal
forests
haveconcentrated
on extrapolatingstand-levelresponses
of photosynthesis
and respirationto differencesin climatic
conditions
[e.g.Frolking et al., 1996; Frolking, 1997; Goulden
et al., 1998], our results show that significantinterannual
variability in regionalC exchangeratesmay also resultfrom
interannualdifferencesin the occurrence
of fire. For example,
the 1981 fire burnedapproximately17% of the studyarea(128
of733km2).Theaverage
soilC accumulation
ratein1981is
estimated
byourmodel
tohave
been
27gm-2.Assuming
that
buming
ofthemoss
layer
inthis128km2firereleased
2000gC
net
Cyr
flux
1981
was
a
loss
ofX
317
m-2thel•
m_•
yr_ the
(27
gCsoil
m-2
-1xin
0.83
- 2000
gnet
Crd
2y/l
0.17).
gC
For comparison,Frolking [1997] estimatedthat interannual
differences
in C-2 storage
at theOBS tower
sitevariedbetween0
-1
ß
.
and125g C m yr becauseof physiological
response
to interannual anomaliesin weather. Hence, averagedover large
regions,interannualvariabilityin C exchangemayresultfrom
changesin the lateralextentof fire as muchor morethandirect
responses
of the vegetationto climaticconditions.
Puttingthe importance
of fire anotherway, the releaseof C
fromthemosslayerby a firecovetingonly 1% of the areain any
year wouldoffsetthe amountof C storedin soilsof the restof
thearea(20g C m-2yr-l x 0.99--2000g C m-2yfflx 0.01).
The fire return interval
ß
.:..................
.[•:-:::..:-:--..
....
=======================
......
::
..... ..:•=•;.::'•:
...
50•0
Fibre5. Rates
ofsoilcarbon
accumulation
orloss
inthe733
km studyarea. The mapsrepresent
area-weighted
averages
of
(a) surfacelayers,includingmoss,(b) deeplayers,and(c) total
profile. Seetextfor discussion
of surfaceanddeeplayers.
for Alaskan
boreal forests has been
estimated
to be 150 years[Kasischkeet al., 1995], with a range
in variousNorth Americanborealforesttypesof 70-500 years
[Payette,1993]. If thefire cyclewereabout100 years,thenour
carbonbudgetfor a yearwith no fire mightsuggestthatthe soil
C stockmayberoughlyat steadystateoverlargeareasand long
timescales. Over millennial timescales,soil C storagerates
sincethedry•.ng
of GlacialLakeAgassizareof theorderof only
-2
-1
a fewg C m yr [Trumboreand Harden, 1997;Harden et al.,
1997]. However,theincidence
of fire is variableoverspaceand
time andcouldbe changingin responseto climatechangeand
managementpractices.For an annualC budgetof a localsite,
the time since the last fire is critical.
For decadal or centennial
scalebudgetsoverthe region,fire frequencyandits variability
overthe landscapeareclearlyimportant.
A majoruncertainty
in determining
thepresentandfuturerole
4. Discussion
of soilsas C sourcesor sinksis the role of the deeporganic
ratesof
The modelestimates
that average2net soil C storagewas matterin theC budgetof uplandsoils. Decomposition
2
I
about20 g C m- yr- overthe 733 km studyareaof northern deeporganiclayersare an orderof magnitudeslowerthanthose
in surfacelayers(Table 1), whichmightleadto theassumption
boreal forest in 1994 (Table 3), which is a modest sink of
atmospheric
CO2. Usinganeddycovariance
method,Goulden thatthedeepsoil is not importantin C cycling. However,large
698
RAPALEE ET AL. ßSOIL C DYNAMICS IN A BOREAL FOREST LANDSCAPE
(A) SURFACE LAYERS
8O
ß-:::=:.-.:•:.."•:.:.
-•.'::i;•.
ø-•':"""••i:•
.........
:.i•
.......
•::""
.
'E 40
o 20
0
.........
!=•
=i.-•,=•:•
;;.....
or- Palsa
Poor- Sphagnum
.•-•
,•.•,:;:.:.,;..
.....•,•
.•,••i:
"'•"•":'••._.•
Imperfect moss
133:•":•34'•-•.•.•.
•'•'""":•'=::••Mod.
well
s5•o-'>•• Well
Time (Yrs.) Since Fire
(B) DEEP LAYERS
..
-40• .,..::•,::•:,
...•"•"•..•
• .• ?:/••---.•.....
..... ..:
....
•.....
:......
:?:.:.-•.,,•:-•
•.•.,....,
................
Well
0 •..••
••••'•
............
P ' -Sphagnummoss
•' •e•-•0 ••
>
Poor - Palsa
Time (Yrs.) Since Fire
Figure6. (a)Rates
ofC accumulation
(positive
values)
in surface
layers
(including
moss)
fordrainage
class
andbytimesince
fire
(stand
age).(b)Rates
ofCloss
(negative
values)
from
deep
organic
layers
(below
moss)
andmineral
soila_•
a-lfunction
of
- soil
drainage
andtimesince
fire.Fensandcollapse
scar
bogs
(notshown)
aregaining
anestimated
29 and12g C m yr , respectively,
in the deeplayers.
storesof C are presentin thedeeplayers,soevenat slowrates
of decomposition,
losseson an annualbasisaresignificant,
and
interannualdifferencesin decomposition
of deep soil C can
affect the C balance of a whole forest stand [Goutden et at.,
1998]. We haveassumedthat the deepsoil C poolsshownet
losses
of C in betweenfireevents.Comparisonof soilC balance
with eddycovariance
measurements
of NEP [Hardenet at.,
1997], evidenceof increasesin late summersoil respiration
[Goutdenet at., 1998], and radiocarbonmeasurements
in soil
CO2 [Winston
et at., 1997]all support
thishypothesis.
Considerableuncertaintiesremain about the origin and
dynamics
of thisimportant
deepsoilC pool. Theradiocarbon
datausedby TrumboreandHarden[ 1997]to calculateinputs
anddecomposition
constants
mayhavebeeninfluenced
moreby
the frequencyand severity of buming than by actual
decomposition
rates of the humic materials. Incubation
measurements
[Fries et el., 1997; Moore and Knowlee,1990]
showthatdeeporganicC decomposes
rapidlyif warmedand
dried.Thesefindings
suggest
thattheslowdecomposition
rates
we inferredfromradiocarbon
studiesare constrained
by physical
RAPALEE
ET AL. ßSOIL C DYNAMICS
IN A BOREAL FOREST LANDSCAPE
699
15.0
10.0
o 5.0
Poor
$phagnurn P•or
Mod Well Imperfect
moss
DRAINAGE
Paisa
VeryPoor •
Fen & Bog Total Area
CLASS
Figure
7. Total
annual
Csoilaccumulation
(>0)orloss
(