Wetland carbon cycle responses to hydrological change:
Impacts on regional and global carbon budgets
Benjamin N. Sulman University of Wisconsin-‐Madison
Department of Atmospheric and Oceanic Sciences
Special thanks: Ankur R. Desai Jonathan Thom Nicole M. Schroeder Nicanor Z. Saliendra Peter M. Lafleur Larry B. Flanagan Rob Scheller
Oliver Sonnentag D. ScoM Mackay Alan Barr Andrew Richardson NACP site synthesis parPcipants
Outline • What are wetlands? • How are they important to the global carbon cycle?
• How do they respond to hydrological variaPons? – Inter-‐annual Pme scales – Century Pme scales
• AddiPonal complicaPons • Conclusions
What are wetlands? U.S. Clean Water Act defini6on: Areas that are inundated or saturated by surface or ground water… sufficient to support … vegetaPon typically adapted for saturated soil condiPons (U.S. Army Corps of Engineers)
Peatlands:
Accumulate thick organic soil layers
Global distribuPon of wetlands
Forested bog
Nonforested bog
Forested Swamp
Nonforested swamp
Alluvial FormaPons
Other land
Water body
MaMhews and Fung, 1987, GBC
Northern peatland types Fen
• Groundwater and surface water fed
• Usually shrubs or sedges dominate
• Peat results from anaerobic soil
Tundra • Permafrost soils • Seasonal thawing leads to flooding of low areas
• Peat results from chronic freezing
Lost Creek (WI)
Barrow (AK) (Photo from specnet.info)
Bog • Rain-‐fed • Nutrient-‐poor • O`en dominated by mosses
• Peat results from anaerobic soil
Mer Bleue (ON)
Western Peatland (AB)
Outline • What are wetlands? • How are they important to the global carbon cycle?
• How do they respond to hydrological variaPons? – Inter-‐annual Pme scales – Century Pme scales
• AddiPonal complicaPons • Conclusions
The global peatland carbon pool is large
Mitra et al, 2005, Curr. Sci.
Boreal and subarcPc wetlands contain between 120 and 500 Pg soil carbon (Mitra et al, 2005) This is up to 1/3 of total global soil carbon pool (Gorham, 1991)
Wetlands in northern landscapes contain a large fracPon of total C
WI: Buffam et al., GCB (2011); MN: Weishampel et al., For. Ecol. Man. (2009) FracPons exclude lake area and carbon storage in lake sediments
0%
10%
20%
30%
40%
50%
60%
70%
80%
90%
100%
WI area WI C MN area MN C
Upland %
Wetland %
Peatland carbon is vulnerable to climate and hydrological change LETTERS
100
60
20
SOC
(kgC
m–2
)
2.0
1.0
0
Peat
dep
th (m
)4°C rise
0 1000 2000 3000 4000Years
0 1000 2000 3000 4000Years
6-year mean water table
Fibrous peat
Humic peat
Metabolic
4°C rise
Humic
Structural
Figure 3 A 4000-year simulation of peat SOC and peat depth at the BOREAS Fensite. a,b, Peat SOC (a) and peat depth (b). Meteorological data for 1994–2005 areused recursively for this long-term simulation. For years 0–2000, the simulated peatcolumn is in dynamic equilibrium under the current climate. A uniform rise oftemperature by 4 �C is applied at year 2000, indicated by downward arrows. Theblack line denotes total peat (fibrous plus humic), and the red line denotes theboundary between fibrous and humic peat.
decomposition. The change in temperature triggers this feedback,and the soil water–carbon system is eventually shifted to the newlow-SOC regime.
To study the transient behaviour of the system, we disturbedthe current equilibrium of the Fen simulation based on thetemperature and precipitation anomalies predicted by the generalcirculation model HadCM3 using scenario A2 of the SpecialReport on Emissions Scenarios for the period 2004–2099 (ref. 25).Transient responses of SOC to climate change strongly dependon the peat type2,18 (Fig. 4). The metabolic pool responds quicklyto climate change, and the decomposition rate of this pool iscontrolled by interannual variability in climate. Extended dryperiods are indicated during 2038–2045 and 2084–2087 due toclimate fluctuations generated by HadCM3 and the hydrologicalmemory of the peat system. The metabolic pool loses more than20% of SOC during each of these dry periods due to exposureof SOC to aerobic conditions10. Although the metabolic SOC is aminor portion of the total SOC, its fast temperature response isthe key process of interannual fluctuations in net ecosystem carbonexchange observed in northern peatlands4.
To single out the physical–biogeochemical interactions, weintentionally omitted ecophysical responses of peatland vegetationto environmental changes. In reality, however, plants willsensitively respond to changes in moisture and temperatureregimes, nutrient status, atmospheric CO2 and peat texture26,27,and changes in the wetland vegetation community and litterquantity and quality strongly influence peat decomposition
–0.2
0
0.2
0.4
0.6
Wat
er ta
ble
dept
h (m
)
1.0
0.8
0.6
0.4
Prop
ortio
nal c
hang
e
1980 2000 2020 2040 2060 2080 2100Years
1980 2000 2020 2040 2060 2080 2100Years
HumicStructuralMetabolic
Current climate
Climate change
Figure 4 Transient change in the water table at the BOREAS Fen site,2004–2099. a, Change in water table. b, Proportional changes in SOC. Before 2004,the model is in equilibrium under the 1994–2005 climate. Then, temperature andprecipitation anomalies projected by HadCM3 with Special Report on EmissionsScenarios A2 are used to force the model in and after year 2004. Shaded areasdenote extremely dry periods of 2038–2045 and 2084–2087.
and accumulation dynamics28. This study emphasizes that thehydrological–biogeochemical feedback inherent to peat has a strongpotential to increase climate sensitivities and avoids a studydesign that might be confounded by two new feedbacks, namelybiogeochemical and vegetation dynamics. The CO2 emissions fromthe peat collapse predicted by this study could be ameliorated orexacerbated by changes in ecosystem structure and function. Ournext research step is to include dynamic vegetation simulated bythe ED model framework15,16.
The transient resistance to peat decomposition observed inthe Fen site simulation is due mainly to microbial conversion oflabile SOC into more recalcitrant SOC29. The massive SOC lossinduced by the soil-condition–carbon feedback can be preventedif the temperature rise is reversed within a few hundred yearsor if a significant increase in precipitation maintains the currentlevels of the water table4. In summary, our modelling approachdemonstrates how the mechanistic linkages that exist between thephysical and biogeochemical dynamics of peatlands have strongimplications for the response of northern peatlands to climatechange30, including a large peat loss due to positive feedbacks inorganic soil.
METHODS
Air temperature, wind speed, net radiation and humidity observed for 12 years(1994–2005) at the BOREAS Northern Study Area OBS eddy-covariancetower site every 30 min (<http://www-as.harvard.edu/data >) are repeatedly
nature geoscience ADVANCE ONLINE PUBLICATION www.nature.com/naturegeoscience 3
• Peat carbon is preserved by cool temperatures and flooded condiPons
• Warming and drying can disrupt the process and lead to carbon loss
Ise et al 2008
Outline • What are wetlands? • How are they important to the global carbon cycle?
• How do they respond to hydrological variaPons? – Inter-‐annual Pme scales – Century Pme scales
• AddiPonal complicaPons • Conclusions
Effects of water table change
Saturated, anoxic Unsaturated, oxygenated
CH4 CO2 CH4 CO2
North American Carbon Program: A site and model intercomparison project
• Three peatland eddy covariance flux sites – Plus four addiPonal sites in a
site comparison
• Seven ecosystem models • Standardized meteorological
driver data • Time series of 3-‐8 years
Results presented in Sulman et al, GRL, 2010 and JGR-‐Biogeosciences, 2012
Western Peatland
Sandhill Fen
Lost CreekWilson FlowageSouth Fork
Mer Bleue
NACP Peatland Sites
Site Lost Creek shrub fen (WI)
Mer Bleue bog (ON)
Western Peatland treed fen (AB)
VegetaPon Primarily alder and willow
Sphagnum mosses with some shrubs
Stunted trees and shrubs, understory of mosses
Mean GEP 2.31 g/m2/day 1.68 2.36
Mean ER 2.10 g/m2/day
1.49 1.83
Mean NEE -‐0.21 g/m2/day -‐0.19 -‐0.53
Example Pmeseries
Lost Creek Shrub fen
Western Peatland tree/sedge fen
Mer Bleue (Eastern Peatland) Bog
NEE (gC/m2/day)
Water table (cm)
NEE (gC/m2/day)
Water table (cm)
NEE (gC/m2/day)
Water table (cm)
Hydrological effects in four fens
• Eddy-‐covariance summer carbon flux anomaly vs. water table anomaly for four northern fen sites
• Both ER and GEP increase with deeper water tables (long Pme scales)
• Drying over short Pme scale can lead to reducPon in GEP and net CO2 emission
• NEE has no significant correlaPon with water table
Sulman et al., GRL, 2010
ContrasPng effects in bogs:
• Bog C fluxes (white symbols) have lower magnitude and opposite sign correlaPon with water table
Sulman et al., GRL, 2010
How well did models simulate peatland processes?
Model name
Temporal resolu6on
Soil layers Soil C pools N cycle Max soil moisture
DLEM Daily 2 3 Yes SaturaPon
Ecosys Hourly 8 9 Yes SaturaPon (with water table)
LPJ Daily 2 2 No Field capacity
ORCHIDEE 30-‐min 2 8 No Field capacity
SiB 30-‐min 10 None No SaturaPon
SiBCASA 30-‐min 25 9 No SaturaPon
TECO 30-‐min 10 5 No SaturaPon
All models overesPmated GEP and ER
Annual
Summer
Sulman et al., JGR, 2012
Monthly residuals were correlated with observed water table
CorrelaPons with water table
CorrelaPon coefficient
Slope
Diurnal cycles not bad at fens
Lost Creek
Diurnal cycles not bad at fens
Western Peatland
Diurnal cycles significantly worse at bog
Conclusions: Interannual Pme scales
• Fens and bogs have opposite responses to water table variaPons
• Ecosystem models overesPmate peatland producPvity and respiraPon
• Water table variaPons contribute significantly to model error
• Models perform beMer at bogs than fens
Outline • What are wetlands? • How are they important to the global carbon cycle?
• How do they respond to hydrological variaPons? – Inter-‐annual Pme scales – Century Pme scales
• AddiPonal complicaPons • Conclusions
Effects of water table change
Saturated, anoxic Unsaturated, oxygenated
CH4 CO2 CH4 CO2
Long-‐term drying: model analysis
LANDIS-‐II model: • Species cohort based forest
succession model • Yearly Pme step • Tracks cohort biomass and
two soil C pools • ReproducPon: Seed dispersal
and establishment probability • NPP: Species maximum NPP,
maximum biomass, and compePPon
CO2
Litter Fast C Slow C
CO2
Atmosphere
NPP(max NPP, max biomass, competition)
Seed dispersal
SimulaPng wetlands
• Plants divided by flood tolerance
• Wet fracPons in grid cells determined with soil height distribuPon
• Growth parameters mulPplied by habitat surface fracPon in grid cell
!1.2
!1.0
!0.8
!0.6
!0.4
!0.2
0.0
0.2
0.4
Dep
th (m
)
Underwater: 61.4% of peat, 36.0% of surface
Upland: 0.0% of peat, 0.0% of surface
Sedges: 23.9% of peat, 41.0% of surface
Wet woody: 14.7% of peat, 57.0% of surface
Bimodal hummock/hollow topography (Eppinga et al. 2008)
Soil decomposiPon model
1000 3000 5000 7000Age (years)
−2.5
−2.0
−1.5
−1.0
−0.5
0.0
Dep
th in
soil
(m)
0.0 0.2 0.4 0.6 0.8 1.0Decomposition modifier
−2.5
−2.0
−1.5
−1.0
−0.5
0.0fTfWWater table
• DecomposiPon rate k depends on age, temperature, and water table factors
• Mean k calculated from 100 soil columns sampled from topography distribuPon
k(t) =k0
1 + k0t
k(z) = k(t)fT (z)fW (z) Model and profiles based on Frolking et al. 2001
Soil decomposiPon and plant community dependence on water table
0 50 100 150 200Decomposition rate (gC/m2/yr)
!2.0
!1.5
!1.0
!0.5
0.0
Wat
er ta
ble
(m)
UplandWetminDeep peatShallow peat
0.0 0.2 0.4 0.6 0.8 1.0Vegetation area fraction
!2.0
!1.5
!1.0
!0.5
0.0
UnderwaterUplandSedgesWet woodyGram peat initialShrub peat initialWet min initial
Peatland pools: Shallow peat scenario:
18.5 kgC/m2 45 cm depth
Deep peat scenario: 100kgC/m2 2.5 m depth
Low sensiPvity at deeper depths is due to older C
VegetaPon fracPon dependence on water table
Soil decomposiPon rate dependence on water table
Modeled landscape: Northern Wisconsin
MinnesotaWisconsin
Ecoregion Ac6ve area frac6on
Upland 38%
Mineral wetland
27%
Shrub peat 29%
Graminoid peat
5%
Price County, near Phillips, WI
Categorized based on remote sensing and soil inventories
Summary of simulaPons
40 cm 100 cm
40 y
ears
10 y
ears
Control Veg
Soil Both
Control Veg
Soil Both
Control Veg
Soil Both
Control Veg
Soil Both
Water table decline
Leng
th o
f dec
line
• Moderate and severe levels of water table decline
• Fast and slow water table decline
• SeparaPon of plant and soil effects
• These combinaPons were applied to both shallow and deep peat scenarios
Model results: control simulaPon fluxes
0100200300400500600
Up
lan
d
NPP
Shallow resp
Deep resp
0
50
100
150
200
250
Wetm
in0
20
40
60
80
100
Sh
rub
peat
0 50 100 150 200 250 300 350 400Model time (years)
010203040506070
Gra
mp
eat
Fluxes
(gC
/m2
-yr)
• Four ecoregions: – Upland forest – Mineral woody wetland – Peat shrub wetland – Peat graminoid wetland
• Upland was most producPve
• ProducPvity declines and respiraPon increases as forest ages
Water table effects on carbon balance
0 100 200 300 400
−10
0
10
20
30
Tota
l ca
rbon
Wetmin
0 100 200 300 400
−10
0
10
20
30
Bio
mass
0 100 200 300 400
−10
0
10
20
30
Soil
carb
on
ch
an
ge
0 100 200 300 400
−10
0
10
20
30
Shrubpeat
0 100 200 300 400
−10
0
10
20
30
0 100 200 300 400
−10
0
10
20
30
0 100 200 300 400
−10
0
10
20
30
Grampeat
0 100 200 300 400
−10
0
10
20
30
0 100 200 300 400
−10
0
10
20
30
0 100 200 300 400
−10
0
10
20
30
Landscape
0 100 200 300 400
−10
0
10
20
30
0 100 200 300 400
−10
0
10
20
30
Carb
on (
kg/m
2)
Model time (years)
Control shallow
Control deep
Veg test shallow
Veg test deep
Soil test shallow
Soil test deep
Net shallow
Net deep
Zero line
Water table decline caused: • Increased soil
decomposiPon • Increased biomass
accumulaPon • Net effect: Short
term increase in carbon, followed by long-‐term losses
Scenario: 100 cm WT decline over 40 years
Water table effects on carbon balance
0 100 200 300 400−10
−5
0
5
10
15
20
Tota
l ca
rbon
Wetmin
0 100 200 300 400−10
−5
0
5
10
15
20
Bio
mass
0 100 200 300 400−10
−5
0
5
10
15
20
Soil
carb
on
0 100 200 300 400−10
−5
0
5
10
15
20Shrubpeat
0 100 200 300 400−10
−5
0
5
10
15
20
0 100 200 300 400−10
−5
0
5
10
15
20
0 100 200 300 400−10
−5
0
5
10
15
20Grampeat
0 100 200 300 400−10
−5
0
5
10
15
20
0 100 200 300 400−10
−5
0
5
10
15
20
0 100 200 300 400−10
−5
0
5
10
15
20Landscape
0 100 200 300 400−10
−5
0
5
10
15
20
0 100 200 300 400−10
−5
0
5
10
15
20
Carb
on (
kg/m
2)
Model time (years)
40 cm 40 yrs
40 cm 10 yrs
100 cm 40 yrs
100 cm 10 yrs
Zero line
Peatlands: • 100 cm declines:
– Short term: C gain – Long term: C loss
• 40 cm declines – Short term: C neutral – Long term: C loss
Mineral wetlands: • C gain for both Whole landscape • Short-‐term: C increase • Long-‐term: C steady • Time scale of decline
made liMle difference
Net change from control run for shallow peat simulaPons: Different water table scenarios
Simple global upscaling
MaMhews and Fung, 1987, GBC
Forested bog
Nonforested bog
Forested Swamp
Nonforested swamp
Alluvial FormaPons
Other land
Water body
Boreal and subarcPc wetland area ≈ 2-‐4x1012 m2 (Mitra et al 2005)
Simple global upscaling
• Boreal/subarcPc wetland area ≈ 2-‐4x1012 m2 (Mitra et al 2005)
• Modeled changes: – Soil C loss of 5 kgC/m2
– Biomass C gain of 5-‐10 kgC/m2
• Anthro emissions ≈ 4-‐8 PgC/year (IPCC 2007) • Global equivalent
– Loss of 10-‐20 PgC (1-‐5% addiPonal emissions over 100 yrs) – Gain of 30-‐60PgC (4-‐15% lower emissions over 100 yrs)
Conclusions: Century Pme scales
• Plant community responses dominate response to drying
• Moderate drying leads to C loss in peatlands • Severe drying leads to short-‐term C gain followed by losses
• Drying leads to C gain in non-‐peat wetlands • Drying leads to significant C gain at landscape scale
• Magnitudes are significant at global scales
Outline • What are wetlands? • How are they important to the global carbon cycle?
• How do they respond to hydrological variaPons? – Inter-‐annual Pme scales – Century Pme scales
• AddiPonal complicaPons • Conclusions
AddiPonal complicaPons and future applicaPons
• Topography • Non-‐CO2 carbon fluxes • Changes in soil properPes over Pme • Climate-‐driven hydrology
Peatland topography
J. S. Aber, 2001. Accessed from hMp://www.emporia.edu/earthsci/estonia/estonia.htm, 1/13/2011. See Aber et al., Suo, 2002
Männikjärve bog, Estonia
Peatland topography
Sonnentag, PhD thesis (2008)
Microtopography in wet peatlands
What does water table depth mean, really?
Microtopography in wet peatlands
• Water table can vary by tens of cm at small scales
• Mean water table at a peatland does not capture the real range of variability
• Topographical variaPons lead to micro-‐ecosystems within the peatland
Measured effects
maintain soil ice later into the season than adjacent wetmicroforms and this may lead to the lower peat temper-atures following water table drawdown. Since Rtot ispositively related to temperature [Moore and Dalva,1993], this shift may help mediate increases in Rtot inducedby the larger oxic zone.[27] Productivity has not been altered significantly fol-
lowing the water table drawdown; however, the vegetationcommunity at experimental hollows changed and GEP atthese locations increased. GPmax increased linearly throughtime at experimental hollows and there is some indicationthat this is also the case at experimental lawns (Figure 5).Increases in the cover of vascular vegetation at lawns andhollows across the site support these trends. Manipulation
of water table position in peatland mesocosms suggests thatmaximum Sphagnum [Weltzin et al., 2001] and graminoidand forb productivity [Weltzin et al., 2000] occurs under wetconditions. These studies did not consider flooded condi-tions but only water table position at or below the surface.In contrast, descriptions from pools naturally drained by soilpipes suggest that the removal of standing water can inducean ecological succession that involves the colonization ofbare peat substrate by Sphagnum mosses and sedges [Fosteret al., 1988] and a similar process is occurring at theexperimental site.[28] A comparison of maximum GEP at the control site
and a nearby site drained for eight seasons [Strack et al.,2006] showed significantly higher productivity at drained
Figure 4. Log-transformed quotient experimental/control for seasonal total (a) GEP, (b) Rtot, and(c) CH4 efflux at individual sample plots through time. Sample plots are arranged from the driest at theleft to the wettest at the right. Vertical dashed lines indicate the transition from (left) hummocks to lawnsand (right) lawns to hollows. Horizontal lines at 1 indicate no difference between experimental andcontrol sites.
GB1007 STRACK AND WADDINGTON: PEATLAND C FLUX AFTER LOWER WATER TABLE
9 of 13
GB1007
WADDINGTON AND ROULET: ATMOSPHERE-WETLAND CARBON EXCHANGES 241
50.0
•. 37.5 2s.o •o m 12.5 0.0
i b 10
•o•' 10
z -200
Lawn
a) CH4
I 1992 :i:• 1993
b) NEE
1 1992
-30
1
0
1992 1993
d) -20 cm peat temperature (C)
I 1992 i'..'• 1993
Figure 4. Mesotopography scale ridge-lawn-pool sequence: (a) seasonal CH 4 flux, (b) seasonal NEE, (c) mean water table position, and (d) -20 cm peat temperature. Solid bars refer to 1992 data, while stippled bars refer to 1993 data. The overbar indicates 1 standard deviation above and below the mean.
The NEE results indicate that the MEF estimates (- 17 g CO 2 m '2 in 1992, +33.8 g CO 2 m '2 in 1993) and the MAF estimates (+38.8 g CO 2 m '2 in 1992, +171.1 g CO2 m '2 , in 1993) differ greatly from each other and the MMF extrapolation (-33.5 and -4.2 g CO2 m '2 in 1992 and 1993, respectively). Furthermore, the MAF approach estimates a net loss of CO 2 from the peatland in both 1992 and 1993. The MMF estimate of NEE is similar
to results of NEE from vegetated lawns in various peatlands of the Hudson Bay lowland (HBL) [Whiting, 1994]. Comparison of these approaches indicates the necessity to consider all scales of peatland topography when making an estimate of NEE. NEE estimates are extremely sensitive to scaling rules not only because NEE differs by 2 orders of magnitude between ridges and pools, but also because the sign of the flux changes among mesoscale units. The approach that includes the maximum ecosystem diversity and weights the fluxes according to that diversity should yield the best estimate of the true flux.
The MMF estimate of NEE indicates a large interalmual variability in the exchange of CO2 at Stor-Amyran (-33.5 and -4.2 g CO2 m '2 in 1992 and 1993, respectively). The mean water table position was similar in the two years, and both years experienced a pronounced dry period (mid June in 1992, mid
July in 1993). The timing ofthese dry periods, however, appears to be the reason why the exchange differed between years. Dry conditions in July coincide with peak peat temperature. Peak soil respiration therefore coincides with peak CO 2 fixation by plants. Consequently, lower water table conditions in July may reduce net uptake because the respiration offsets the fixation.
An earlier dry condition in June had less impact on net CO 2 exchange because soil respiration is limited due to colder soils
in June. Asymmetry between respiration and plant fixation of CO2 has been shown to determine the net source-sink
relationship in boreal forest stands (S. Frolking, University of New Hampshire, personal communication, 1995). Furthermore,
there was a greater loss of CO 2 from the pools in 1993 because they were warmer than in 1992. Since the pool flux is so large, the pools greatly influence the peatland ecosystem level NEE even though they represent only a small proportion of the total surface area. Similar results were found by Hamilton et al. [1994] for peatlands in the Hudson Bay lowlands.
Scales of Variability in Atmosphere - Peatland Carbon Exchange
This study examined the exchange of carbon over spatial scales from tens of centimeters (microtopography; Figure 3), to
Strack and Waddington, Glob. Biogeochem. Cy., 2007
Waddington and Roulet, Glob. Biogeochem. Cy., 1996
CH4 and CO2 fluxes Effects of lowered water table
Non-‐CO2 carbon fluxes An example: Mer Bleue bog
not representative of the long-term rates because thepeat is still actively decomposing in the oxygenated,unsaturated zone (Turunen et al., 2004). Peat age of 400years corresponds to depths of 0.32m in core MB930and 0.45m in core MB775, and zero depth correspondsto the base of living Sphagna. Deposition time withinthis window varies between 16 and 42 yr cm!1 for theMB930 core and between 19 and 31 yr cm!1 for theMB775 core.
Results
Annual and seasonal climate variations
The climate for 1998–2004 was quite variable (Fig. 2):over half the mean seasonal temperatures and totalprecipitation were greater than 1 SD from the long-term
mean and totals for the 66-year record at the OttawaMacdonald-Cartier International Airport (45.191N lati-tude, 75.401W longitude, 114m a.m.s.l.), approximately12km southwest of the Mer Bleue peatland (Fig. 1). Weranked the seasonal temperature and precipitation withinthe 66 years of record (Fig. 3). The winters of 2002 and2004, the spring of 1999, the summers of 2001 and 2002,and autumn 2000 received significantly (i.e. in the upperor lower quartile) less precipitation, while the springs of2000 through 2002, and the autumns of 1999, 2003, and2004 experienced precipitation closer to the norm. Onlythree seasons were colder than normal (summer andautumn 1999, and winter 2002), while in 11 of theremaining 21 seasons, or 45% of the study period, thetemperatures were in the top quartile. Six out of a possible24 seasons experienced both large temperature and pre-cipitation anomalies (drier/wetter and warmer/cooler).
!120
!100
!80
!60
!40
!20
0
20
40
1998–1999 1999–2000 2000–2001 2001–2002 2002–2003 2003–2004 Six-yearMean
C (g
m!2
yr!1
)
Peatland loss
Peatland gain
NEE
CH4 DOC
C balance
!80
!60
!40
!20
0
20
40
1-Nov-98 1-Nov-99 1-Nov-00 1-Nov-01 1-Nov-02 1-Nov-03 1-Nov-04
WT
(cm
); T
a an
d Ts
(°C
)
Water table (cm) Air temperature (2 m) Hummock temperature (!0.1 m)
Fig. 5 Components of the annual carbon balance (top panel) and the trends on air, hummock peat temperature and water table for the
6 years (1 November 1998 to 31 October 2004) for the Mer Bleue peatland (bottom panel).
C ONT EMPORARY AND LAT E HOLOC ENE CAR BON ACCUMULAT I ON 405
r 2007 The AuthorsJournal compilation r 2007 Blackwell Publishing Ltd, Global Change Biology, 13, 397–411
• NEE was larger than other factors, but ignoring DOC and CH4 would lead to overesPmate of net carbon uptake
• High inter-‐annual variability leads to high uncertainty
Roulet et al., Glob. Change Biol., 2007
Northern Wisconsin landscape
age of wetlands relative to upland forests (Table 2). LakeCO2 evasion contributes about 3% of the regional totalnet exchange. Precipitation flux into and hydrologicflux out of the NHLD are on the order of 1% and 3%,respectively, of the regional total net exchange.The mismatch between pool sizes and current rates
points to widely disparate mean C turnover timesamong wetlands, lakes, and forests of the region.Assuming current rates are representative of long-termrates, the estimated average time required to build upeach major regional C pool was 65 years for forest, 1775years for peat wetlands, and 7364 years for lake sedi-ments.
Spatial distribution of C pools and fluxes on the landscape
Pools and fluxes of C were spatially heterogeneous at arange of spatial scales in the NHLD landscape. Themost dense pools of C were peat in peatlands andsediments in lakes (Fig. 3b). Patches of relatively youngconiferous forest gave by far the greatest local rates ofinflux of C into the land surface, while lakes gave rise to
patches of C evasion, with smaller lakes having thehighest evasion rates (Fig. 3c).
Discussion
Implications of magnitude and spatial variability of Cpools and fluxes
By constructing a C budget we incorporated surfacefreshwaters, wetlands, and upland forests into a singleframework. This framework establishes a context for Ccycling research in this and other surface-water richregions. An integrated view like this is needed to targetresearch and management strategies in a parsimoniousway.We determined that the largest current-day annual
land–atmosphere fluxes in the NHLD region are foundin forests, which are aggrading. This rate is an order ofmagnitude higher than surface–atmosphere exchangein wetlands or surface waters (Fig. 2, Table 2). Thus thecurrent behavior of the NHLD in terms of annual C
Forests: 64,000
Pool sizes in Gg-C
Wetlands: 158,000 Surface Waters: 162,000
Atmosphere
Flux rates in Gg-C yr-1
Forest NEE994
GPP 3233, R 2238
CH4 emission13
Wetland NEE124
GPP 878, R 754
CO2 evasion28
Runoff34
CH4 evasion2
Fossil fuels154
Precip12
Sed 22
Wetland runoff21
Forest runoff24
Wetland litter1
Forest litter2
Fig. 2 Schematic showing the three major ecosystem types of the Northern Highlands Lake District (NHLD), along with best estimates of
C flux rates and pool sizes. These estimates are associated with varying degrees of uncertainty (Tables 1–5). Forests make up 54% of the
NHLD area, wetlands 28% (including 20% peatlands and 8% other wetlands), and lakes 13%. NEE, net ecosystem exchange; GPP, gross
primary production; R, respiration.
12 I . B U F FAM et al.
r 2010 Blackwell Publishing Ltd, Global Change Biology, doi: 10.1111/j.1365-2486.2010.02313.x
Results for northern Wisconsin Wetland liMer + wetland runoff = 17.7% of wetland NEE LiMer + runoff + methane = 28% of wetland NEE Forest liMer + runoff = 2.6% of forest NEE Buffam et al., Glob. Change Biol., 2011
Future model improvements • Dynamic soils
– Peat layers treated as age cohorts – Soil subsidence and changes in bulk properPes
• InteracPve hydrology – Couple to climate-‐driven hydrological model – Landscape topography driven by digital elevaPon map
• Improved biology – Nitrogen cycle – ProducPvity coupled to climate – Explicit species biological responses to flooding
• Climate feedbacks – Albedo – Latent and sensible heat fluxes – Carbon cycle coupled to climate
Outline • What are wetlands? • How are they important to the global carbon cycle?
• How do they respond to hydrological variaPons? – Inter-‐annual Pme scales – Century Pme scales
• AddiPonal complicaPons • Conclusions
Summary of results
• Peatland community types and succession control carbon cycle responses to hydrological change
• Model simulaPons overes6mate produc6vity and respira6on and miss hydrology-‐driven variability in peatlands
• Responses to hydrological change vary greatly depending on 6me scale
How might wetlands surprise us?
• Slow and fast hydrological changes can have opposite effects on carbon fluxes
• Different types of northern wetlands can have opposite responses to similar forcings
• Tundra, northern wetlands, coastal wetlands, and tropical wetlands could have different behaviors
• MulPple micro-‐ecosystems within a peatland due to topography could lead to higher resilience than expected
Acknowledgements
• Ankur Desai • Jonathan Thom • Lab members and other grad students • Thanks for funding from the BART IGERT fellowship program • Natural Sciences and Engineering Research Council of Canada (NSERC), the
Canadian FoundaPon for Climate and Atmospheric Sciences (CFCAS), and BIOCAP Canada
• North American Carbon Program (NACP) and NASA Terrestrial Ecology Program • U.S. Department of Energy (DOE) Office of Biological and Environmental Research
(BER) NaPonal InsPtute for ClimaPc Change Research (NICCR) Midwestern Region Subagreement 050516Z19
• Thanks to my coauthors and all the contributors to the NACP site synthesis
References • Aber et al. PaMerns in Estonian bogs as depicted in color kite aerial photographs. Suo (2002) vol. 53 pp. 1-‐15 • Baird et al. Upscaling of peatland-‐atmosphere fluxes of methane: small-‐scale heterogeneity in process rates and the pitalls of “bucket-‐and-‐slab” models. In Carbon Cycling in Northern Peatlands, A. J. Baird, L. R. Belyea, X. Comas, A. S. Reeve, and L. D. Slater, eds. American Geophysical Union, Washington, D.C., 2009.
• Belyea, L. R. Nonlinear dynamics of peatlands and potenPal feedbacks on the climate system. In Carbon Cycling in Northern Peatlands, A. J. Baird, L. R. Belyea, X. Comas, A. S. Reeve, and L. D. Slater, eds. American Geophysical Union, Washington, D.C., 2009.
• Buffam et al. IntegraPng aquaPc and terrestrial components to construct a complete carbon budget for a north temperate lake district. Global Change Biology (2011) vol. 17 (2) pp. 1193-‐1211
• Eppinga, M. B., Rietkerk, M., Borren, W., Lapshina, E. D., Bleuten, W., & Wassen, M. J. (2008). Regular Surface PaMerning of Peatlands: ConfronPng Theory with Field Data. Ecoscience, 11(4), 520–536.
• Frolking, S., Roulet, N., Moore, T., Richard, P. J. H., Lavoie, M., & Muller, S. (2001). Modeling northern peatland decomposiPon and peat accumulaPon. Ecosystems, 4(5), 479–498.
• Mitra et al. An appraisal of global wetland area and its organic carbon stock. Current Science (2005) vol. 88 (1) pp. 25-‐35 • Roulet et al. Contemporary carbon balance and late Holocene carbon accumulaPon in a northern peatland. Global Change Biology (2007) vol. 13 pp. 397-‐411
• Strack and Waddington. Response of peatland carbon dioxide and methane fluxes to a water table drawdown experiment. Global Biogeochem. Cycles (2007) vol. 21 (1) pp. GB1007
• Sulman et al. CO2 fluxes at northern fens and bogs have opposite responses to inter-‐annual fluctuaPons in water table. Geophys Res LeM (2010) vol. 37 (19) L19702
• Sulman, B. N., Desai, A. R., Schroeder, N. M., Ricciuto, D., Barr, A., Richardson, A. D., Flanagan, L. B., et al. (2012). Impact of hydrological variaPons on modeling of peatland CO2 fluxes: results from the North American Carbon Program site synthesis. Journal of Geophysical Research, 117, G01031. doi:10.1029/2011JG001862
• Waddington and Roulet. Atmosphere-‐wetland carbon exchanges: Scale dependency of CO2 and CH4 exchange on the developmental topography of a peatland. Global Biogeochem. Cycles (1996) vol. 10 (2) pp. 233-‐245
• Waddington et al. Water table control of CH4 emission enhancement by vascular plants in boreal peatlands. J. Geophys. Res (1996) vol. 101 (D17) pp. 22775
• Weishampel et al. Carbon pools and producPvity in a 1-‐km2 heterogeneous forest and peatland mosaic in Minnesota, USA. Forest Ecology and Management (2009) vol. 257 (2) pp. 747-‐754
