Content uploaded by Sebastian Wolf
Author content
All content in this area was uploaded by Sebastian Wolf on Oct 28, 2017
Content may be subject to copyright.
Carbon sequestration potential of tropical pasture
compared with afforestation in Panama
SEBASTIAN WOLF
*
, WERNER EUGSTER
*
,CATHERINEPOTVINwz,
BENJAMIN L. TURNERzand NINA BUCHMANN
*
*
Institute of Agricultural Sciences, ETH Zurich, Universitaetsstrasse 2, 8092 Zurich, Switzerland, wMcGill University, Department
of Biology, 1205 Dr Penfield Avenue, Montre
´al H3A1B1, Que
´bec, Canada, zSmithsonian Tropical Research Institute, Apartado
0843-03092, Balboa, Ancon, Panama
Abstract
Tropical forest ecosystems play an important role in regulating the global climate, yet deforestation and land-use
change mean that the tropical carbon sink is increasingly influenced by agroecosystems and pastures. Despite this, it is
not yet fully understood how carbon cycling in the tropics responds to land-use change, particularly for pasture and
afforestation. Thus, the objectives of our study were: (1) to elucidate the environmental controls and the impact of
management on gross primary production (GPP), total ecosystem respiration (TER) and net ecosystem CO
2
exchange
(NEE); (2) to estimate the carbon sequestration potential of tropical pasture compared with afforestation; and (3) to
compare eddy covariance-derived carbon budgets with biomass and soil inventory data. We performed comparative
measurements of NEE in a tropical C
4
pasture and an adjacent afforestation with native tree species in Sardinilla
(Panama) from 2007 to 2009. Pronounced seasonal variation in GPP, TER and NEE were closely related to radiation,
soil moisture, and C
3
vs. C
4
plant physiology. The shallow rooting depth of grasses compared with trees resulted in a
higher sensitivity of the pasture ecosystem to water limitation and seasonal drought. During 2008, substantial
amounts of carbon were sequestered by the afforestation (–442 g C m
–2
, negative values denote ecosystem carbon
uptake), which was in agreement with biometric observations (–450 g C m
–2
). In contrast, the pasture ecosystem was a
strong carbon source in 2008 and 2009 (261 g C m
–2
), associated with seasonal drought and overgrazing. In addition,
soil carbon isotope data indicated rapid carbon turnover after conversion from C
4
pasture to C
3
afforestation. Our
results clearly show the potential for considerable carbon sequestration of tropical afforestation and highlight the risk
of carbon losses from pasture ecosystems in a seasonal tropical climate.
Keywords: carbon dioxide fluxes, eddy covariance, FLUXNET, grazing, land-use change, managed ecosystems, soil carbon
Received 5 January 2011; revised version received 2 May 2011 and accepted 6 May 2011
Introduction
Tropical ecosystems account for more than half of the
global terrestrial gross primary production (GPP) (Beer
et al., 2010), contain 40% of the carbon stored in the
terrestrial biosphere, and are considered to sequester
large amounts of carbon dioxide from the atmosphere
(Grace et al., 2001). However, the current role of tropical
ecosystems in terrestrial carbon sequestration remains
uncertain as ongoing deforestation and associated land-
use changes strongly reduce the area of tropical forests,
with cropland and pasture becoming more prevalent
(Fearnside 2005; Alves et al., 2009). Land-use change
from tropical forest to pasture has been reported to affect
ecosystem carbon budgets in the short-term through
increased inter and intra-annual variations in ecosystem
CO
2
fluxes and the sensitivity to seasonal drought
(Priante-Filho et al., 2004; von Randow et al., 2004;
Saleska et al., 2009). Moreover, one of the major long-
term effects of such land-use changes is the reduced
carbon sink strength of pasture ecosystems (IPCC 2007).
Despite the general importance of tropical ecosystems
for global climate and carbon cycling, eddy covariance
(EC)fluxmeasurementsinthetropicsremainscarceand
thus are globally under-represented. Tropical sites repre-
sent only 10% of the global FLUXNET measurement net-
work, with most sites located in neotropical forests and
only a few in tropical pastures or other land-use types
(http://www.fluxnet.ornl.gov). A recent FLUXNET synth-
esis highlighted the importance of C
4
vegetation for ter-
restrial GPP, accounting for 20% of global terrestrial GPP,
and emphasized the need for an expansion of observations
in these scarcely covered ecosystems (Beer et al., 2010).
EC flux measurements indicate that many tropical
forests act as carbon sinks, which is consistent with
Correspondence: Sebastian Wolf, tel. 141 44 632 3886, fax 141 44
632 1153, e-mail: sewolf@ethz.ch
Global Change Biology (2011) 17, 2763–2780, doi: 10.1111/j.1365-2486.2011.02460.x
r2011 Blackwell Publishing Ltd 2763
carbon uptake inferred from long-term biometric data
at some of these sites (Malhi et al., 1999; Loescher et al.,
2003; Luyssaert et al., 2007; Bonal et al., 2008). A few
tropical forests were reported to act as carbon sources,
although this might have been related to severe drought
or disturbance recovery (Saleska et al., 2003; Hutyra
et al., 2007). In Brazil, a transitional forest was found
to have an annual carbon budget close to equilibrium
(Vourlitis et al., 2001) while a tropical savanna appeared
to be a sink of carbon (Miranda et al., 1997). Published
results of carbon fluxes for other land-use types in the
neotropics are limited: again in Brazil, Grace et al. (1998)
and von Randow et al. (2004) found indications that a
tropical pasture sequestered carbon, as did Priante-
Filho et al. (2004) for a pasture under conversion to
afforestation. In contrast, chamber measurements by
Wilsey et al. (2002) showed carbon losses from tropical
pastures in Panama. It therefore remains unclear
whether tropical pastures are carbon sinks or sources.
Latin America has one of the highest deforestation
rates in the tropics, with land predominantly converted
to pasture for extensive grazing (Wassenaar et al., 2007).
Few of these pastures are later used for afforestation,
although this is considered an effective measure to
sequester carbon and mitigate increasing CO
2
concen-
trations in the atmosphere (FAO 2009). Malhi et al. (2002)
estimated the mitigation potential of tropical afforesta-
tion at 15% of global CO
2
emissions. Afforestation of
pasture may become more relevant for tropical countries
in the future within the international carbon accounting
of the Kyoto protocol, but this requires accurate infor-
mation on the carbon sequestration potential involved.
A comparative measurement design is needed to quan-
tify carbon dynamics involved in the land-use change
from pasture to afforestation, to account for confounding
factors of spatial divergence and variations in meteo-
rology (Don et al., 2009). In this study, we determined
the carbon budgets of tropical pasture and native tree
species afforestation at a site in Central Panama from 2007
to 2009 based on continuous measurements using two EC
flux towers. The objectives of our study were: (1) to
elucidate the environmental controls and the impact of
management on GPP, total ecosystem respiration (TER)
and net ecosystem CO
2
exchange (NEE); (2) to estimate the
carbon sequestration potential of tropical pasture com-
pared with afforestation; and (3) to compare EC -derived
carbon budgets with biomass and soil inventory data.
Material and methods
Site description
The Sardinilla site (Panama) is located at 91190N, 791380W and
70 m a.s.l., about 30 km north-east of Barro Colorado Island
(BCI). Sardinilla has a semi-humid tropical climate with a
mean annual temperature of 25.2 1C, 2289 mm annual precipi-
tation (2007–2009; long-term mean of nearby Salamanca 1972–
2009 is 2267 mm) and a pronounced dry season from January
to April. Dry season length in Central Panama varies among
years (134 19 days for 1954–2009; ACP, 2010) and is – along
with precipitation patterns – influenced by ENSO, the El Nin
˜o
Southern Oscillation (Graham et al., 2006; Lachniet, 2009).
Geologically, the site belongs to the Gatuncillo formation and
the bedrock is classified as tertiary limestone containing clayey
schist and quartz sandstone (ANAM 2010). Soils in the pasture
are Alfisols, based on their high base status and clay transloca-
tion in the profile. Soils in the afforestation are similar and
include as well large areas with cracking clays that exhibit
vertic properties. The Sardinilla site was logged in 1952/1953
and shortly used as arable land, before it was converted to
pasture (Wilsey et al., 2002). An improved afforestation (i.e.,
plantation using native tree species only) was established at
parts of the site (7.5 ha) in 2001 with an average of 1141 stems
per ha and without any particular soil preparation (plough-
ing). The six native tree species used for the afforestation site
were: Luehea seemanii, Cordia alliodora, Anacardium excel-
sum, Hura crepitans, Cedrela odorata, Tabebuia rosea. A
moderately dense understory vegetation (shrubs, grasses and
sedges) was present, which was cut once a year (typically in
December) by manual thinning and the residues left on-site.
Traditional grazing continued on an adjacent pasture (6.5 ha),
where vegetation is dominated by C
4
grasses, and consists of
(most abundant first): Paspalum dilatatum (C
4
), Rhynchospora
nervosa (sedge, C
3
), Panicum dichotomiflorum (C
4
) and Spor-
obolus indicus (C
4
). Mean canopy height was about 10 m in
the afforestation and 0.09 m in the pasture (in 2008). While
the afforestation site has an undulating topography (elevation
range o10 m), the adjacent pasture is homogeneously flat
with an overall slope of o21. Detailed footprint analyses
indicated that fluxes measured at both sites indeed originate
predominantly from the respective land-use type (Wolf et al.,
2011). For further details on the Sardinilla sites see Wolf et al.
(2011).
Instrumentation and data acquisition
Two flux towers were established in Sardinilla over a grazed
pasture (March 2007–January 2010), and an adjacent afforesta-
tion (February 2007–June 2009). Our micrometeorological mea-
surement systems consisted of an open path infrared gas
analyzer (IRGA, Li-7500, LI-COR, Lincoln, NE, USA) and a
three-dimensional sonic anemometer (CSAT3, Campbell Scien-
tific, Logan, IL, USA). Instruments were installed at a height of
3 m at the pasture and 15 m at the afforestation site. Data
acquisition was performed using an industry grade embedded
box computer (Advantech ARK-3381, Taipei, Taiwan), running
a Debian based Linux operating system (Knoppix 4.0.2, Knop-
per.Net, Schmalenberg, Germany). Ancillary meteorological
measurements included air temperature and relative humidity
(MP100A, Rotronic, Bassersdorf, Switzerland), incoming
shortwave radiation (R
G
, CM3, Kipp & Zonen, Delft, The
Netherlands), photosynthetic photon flux density (PPFD,
2764 S. WOLF et al.
r2011 Blackwell Publishing Ltd, Global Change Biology,17, 2763–2780
PAR Lite, Kipp & Zonen), precipitation (10116 rain gauge,
TOSS, Potsdam, Germany), soil temperature at 5 cm depth
(TB107, Markasub, Olten, Switzerland) and volumetric soil
water content (SWC) at 5 and 30 cm depth (EC-5, Decagon,
Pullman, WA, USA). Flux measurements were conducted at
20 Hz, meteorological measurements at 10 s and stored as half-
hourly averages (sums for precipitation) using data loggers:
CR23X at the afforestation and CR10X at the pasture site (both
Campbell Scientific). Precipitation and incoming shortwave
radiation were measured at one tower location only. Regular
cleaning of sensors and IRGA calibration checks were carried
out to assure data quality. Further details on the measurement
setup at the Sardinilla site are reported in Wolf et al. (2011).
Flux data processing
Data acquisition and quality filtering. Flux measurements
were recorded using the in-house software sonicreadHS and
raw data were processed to half-hourly averages using the
in-house EC software eth-flux (Mauder et al., 2008; source code
for Unix/Linux systems can be obtained from the authors).
During post-processing, fluxes were corrected for damping
losses (Eugster & Senn, 1995) and density fluctuations (Webb
et al., 1980). Data screening was done using the following
rejection criteria: (1) Optical sensor contamination (spider
eggs, rain) resulting in high window dirtiness of the IRGAs.
We used a 10% threshold above the mean background value of
the respective IRGA. (2) Filtering for stationarity following
Foken & Wichura (1996). We excluded fluxes whenever the
30 min average deviated by more than 100% from the
corresponding mean of 5 min averages. (3) CO
2
fluxes
outside the range of –50 to 50 mmol m
–2
s
–1
were excluded. (4)
Statistical outliers outside the 3 SD range of a 14 day
running mean window were removed. (5) Periods with low
turbulence conditions were excluded based on friction velocity
(u
*
). Seasonal and site-dependent u
*
-thresholds were
determined according to the method by Gu et al. (2005) and
Moureaux et al. (2006). These threshholds yielded u
*
o0.04 m s
–
1
during the dry season, u
*
o0.03 m s
–1
during the dry–wet
transition, while no thresholds were found during the wet
season and wet–dry transition periods for the pasture site. At
the afforestation site, the thresholds determined were
u
*
o0.02 m s
–1
during the dry season, u
*
o0.01 m s
–1
during
the wet season, u
*
o0.05 m s
–1
during the dry–wet transition,
while no threshold was found during the wet–dry transition
period. (6) Negative nighttime fluxes and a respective amount
of positive nighttime data were removed using a trimmed
mean approach.
Gap filling
Filling of data gaps was required to obtain a continuous time
series of flux data for budget assessments. At the pasture site,
data were available for 97.7% of the time between June 2007
and January 2010. After quality filtering, 54.6% of good to
excellent quality data remained (64.7% daytime, 43.6% night-
time data). At the afforestation site, data availability was 94.5%
between June 2007 and June 2009, with 47.6% of good to
excellent quality data remaining after quality filtering (65.4%
daytime, 28.3% nighttime data).
Gap filling of NEE
daytime
was based on non-linear light
response curves (LRC), i.e., the functional relationship be-
tween daytime CO
2
fluxes and photosynthetic photon flux
density (PPFD). We used a logistic sigmoid function as sug-
gested by Moffat (unpublished) that has been used by Eugster
et al. (2010) to determine light-response parameters for each
single day:
NEEdaytime ¼2Amax 0:51
1þe2aPPFD
Amax
þTERdaytime ð1Þ
A
max
denotes the maximum photosynthetic capacity of the
ecosystem (mmol CO
2
m
–2
s
–1
), athe apparent quantum yield
(mmol CO
2
mmol
1
photons), PPFD the photosynthetic photon
flux density (mmol photons m
–2
s
–1
, 90% quantile used to
exclude outliers) and TER
daytime
the daytime TER
(mmol CO
2
m
–2
s
–1
). The initial value of awas set to 0.03 and
the initial value of TER
daytime
was determined using a linear
least-squares regression. The applied sigmoid fit overcomes
some of the limitations of the widely used rectangular
(Michaelis–Menten equation) and nonrectangular hyperbolic
fits (Gilmanov et al., 2003). In particular, it was found to yield
the best light response approximation of all semiempirical
functions by properly describing the different phases of the
light response of NEE
daytime
(Eugster et al., 2010; Moffat
unpublished). For days, when the logistic sigmoid function
did not converge or the curvature in the relationship between
NEE
daytime
and PPFD was not significant, a linear least-squares
regression was used. Remaining daytime gaps (e.g. due to few
or no measurements) were filled using a gap model with
parameters estimated from the LRC of the days prior and
subsequent to the gaps, or using linear interpolation. During
nighttime, we found only a weak temperature sensitivity of
ecosystem respiration to soil and air temperatures (R
2
o0.02),
irrespective of the choice of nonlinear (Lloyd and Taylor
model, Q
10
) or linear functions. Therefore, we gap filled night-
time data using a 10-day running mean approach. Few night-
time gaps that still remained (o1%) at the afforestation site
were filled using linear interpolation.
Partitioning
To partition the comparably small flux of daytime NEE into its
much larger gross components, we used:
GPP ¼NEEdaytime þTER ð2Þ
with GPP (positive value) inferred from the difference of day-
time net ecosystem exchange (NEE
daytime
) and TER. TER was
inferred from mean nighttime data (as no temperature sensi-
tivity was observed, cf. Gap Filling section), when photosynth-
esis is zero (and thus GPP is zero). In cases when NEE
daytime
exceeded TER (resulting in negative GPP values), e.g. with
onset of turbulent mixing in the morning or after rainfall, we
replaced TER derived from nighttime data with NEE
daytime
and set GPP to zero. Since our daytime TER is inferred from
mean nighttime data without a temperature dependency
observed in Sardinilla, no diurnal variations in TER are
CARBON SEQUESTRATION OF PASTURE AND AFFORESTATION 2765
r2011 Blackwell Publishing Ltd, Global Change Biology,17, 2763–2780
assumed. In general, daytime TER inferred from nighttime
data should be considered as best estimate, since it neglects
photorespiration occurring during the day. While this is a
valid assumption for our pasture site which is dominated by
C
4
vegetation, this assumption is more critical for our affor-
estation site which is dominated by C
3
vegetation, although
TER during the day is typically dominated by soil respiration.
LAI, biomass, grazing and soil measurements
Auxiliary variables included leaf area index (LAI) and stand-
ing biomass measurements, stocking densities and soil sam-
pling. Leaf area index (LAI) was measured in campaigns with
a LAI-2000 (LI-COR, Lincoln, USA) in July 2008 and weekly till
bi-weekly from March to July 2009. At the afforestation site,
LAI was measured separately for the tree canopy (measured at
1 m above ground) and the total canopy including the unders-
tory (measured at ground level). In the pasture, aboveground
standing biomass was sampled bi-weekly (n510) from June
2008 to January 2010 using a 50 50 cm aluminium frame with
subsequent drying for at least 3 days at 60 1C. Since February
2009, photosynthetic active (green) biomass was separated
from senescent biomass. Based on the measurements from
2009 and 2010, we estimated the percentage of living biomass
before March 2009. Total aboveground biomass carbon at the
pasture was calculated by assuming that 50% of the dry weight
green biomass is carbon.
Grazing (i.e., stocking density) at the pasture was monitored
between June 2008 and January 2010 by counting the number
of grazing livestock (dominantly cattle with a few horses) on a
daily basis. We used coefficients reported by Chilonda & Otte
(2006) to calculate standardized livestock units (LU) per hec-
tare, with cattle accounting for 0.7 LU and horses for 0.5 LU in
Central America. Overgrazing was defined as 44LUha
–1
d
–1
,
which is rather conservative with respect to generally accepted
values of the carrying capacity of 1–2 LU ha
–1
in Europe.
At the afforestation site, standing biomass was assessed
using annually measured biometric data for trees (calculated
based on allometric equations; on 22 plots of 45 45 m size),
herbaceous plants, litter and coarse woody debris (CWD).
Details on the assessment of biometric data can be found in
Potvin et al. (2011). As herbaceous biomass data were not
available for 2009, we assumed no change from 2008. Data
on CWD were not available for the years 2007 and 2009, and
thus, we estimated CWD by averaging the data from available
years.
Since the year 2008 was the only calendar year with full data
coverage by EC measurements at the afforestation, our direct
comparison with inventory data was initially constrained to
that specific year. However, we used our EC measurements
from 2008 to estimate fluxes from January to May of the year
2007, and July to December 2009. This extrapolation made it
possible to compare 3 years of EC fluxes with the biomass
inventory at the afforestation.
Topsoil (0–10 cm) sampling at the afforestation was done in
March 2009 (n522) using a cylindrical corer 10 cm long with a
diameter of 6.8 cm. At the pasture site, three soil profiles from 0
to 100 cm depth were sampled horizontally in January 2010, in
10 cm increments, and additional samples in 5 cm depth.
Topsoil values at the pasture were derived by averaging the
samples from 5 and 10 cm depth. All samples were dried at
60 1C for at least 72 h before they were ground and analyzed
for C, N and d
13
C with an elemental analyzer (Thermo Flash
HT Soil Analyzer, Thermo Fisher Scientific, Waltham, USA)
coupled through a Conflo III interface to an isotope ratio mass
spectrometer (Delta V Advantage, Thermo Fisher Scientific,
Waltham, USA). To assess changes in soil parameters since
afforestation establishment, we compared our measurements
with samplings from June 2001 and 2002 by Abraham (2004).
We used bulk density (d
B
) values reported by Abraham (2004)
to calculate topsoil (0–10 cm) carbon and nitrogen pools, by
assuming no changes in d
B
since 2001/2002. This is supported
by a study of Seitlinger (2008) that found no changes of topsoil
d
B
in the afforestation between 2001 and June 2007. Data on
carbon pools below 10 cm at the afforestation site was
extracted from Abraham (2004), assuming changes in carbon
pools since 2001 occurred predominantly in the topsoil, as
deeper soil carbon pools are considered relatively stable
(Malhi & Davidson, 2009). At the pasture site, we used the
mean of the three soil profiles sampled in 2010 and the topsoil
d
B
reported by Abraham (2004) to estimate soil carbon pools
from 0–100 cm. A mixing model was used to assess the carbon
isotopic source contribution of organic matter in the soil, with
litter values reported by Abraham (2004) as –14.4%for pasture
and –29.5%for tree litter.
Statistical analyses and general conventions
All statistical analyses were carried out using the Rstatistics
software package, version 2.10.0 (R Development Core Team
2009, www.r-project.org). Daytime was defined as the period
when PPFD exceeded 5 mmol m
–2
s
–1
. Fluxes from the atmo-
sphere to the biosphere are marked with a negative sign
denoting carbon uptake by the ecosystem; positive fluxes
indicate carbon loss. In general, only seasons with full data
coverage were used for seasonal averaging. Separation of dry
from wet seasons and transition periods was done based on
daily precipitation sums using the methodology described in
Wolf et al. (2011): wet season was defined as the time span with
no periods of more than four consecutive days without rain,
and the dry season vice versa. Transition periods mark the time
span between both main seasons. When writing ‘seasonal
drought’, we refer to the plant physiological effects of soil
moisture deficiency during the dry season.
Results
Intra- and interannual variations in precipitation
We found a pronounced seasonality in the climate in
Sardinilla where most of the precipitation (498%) oc-
curred during the wet season from April to December
(Table 1, Fig. 1d). On average, November was the wettest
(4300 mm) and September was the driest month (about
200 mm) during the wet season. The dry season lasted
2766 S. WOLF et al.
r2011 Blackwell Publishing Ltd, Global Change Biology,17, 2763–2780
Table 1 Monthly values of, precipitation sum (P), mean soil water content (SWC, Sardinilla afforestation, at 5 cm depth), total photosynthetic photon flux density (PPFD), mean
air temperature (T
Air
), total net ecosystem exchange (NEE), total gross primary production (|GPP|) and total ecosystem respiration (TER) at Sardinilla, Panama from 2007 to 2009.
Measurements at the afforestation were discontinued after June 2009
Year Month P (mm) SWC (%) PPFD (mol m
–2
)T
Air
(1C)
Pasture Afforestation
NEE (g C m
–2
) |GPP| (g C m
–2
) TER (g C m
–2
) NEE (g C m
–2
) |GPP| (g C m
–2
) TER(g C m
–2
)
2007 Jun 278 40.5 802 25.5 9.3 216 225 –41.4 211 170
Jul 197 43.7 815 25.3 3.8 225 229 –26.7 200 174
Aug 223 45.9 814 25.0 6.9 215 222 –23.4 188 164
Sep 206 48.2 901 25.1 –13.5 239 226 –36.2 196 160
Oct 380 47.2 867 25.1 –13.8 240 226 –39.9 201 161
Nov 351 51.6 751 24.9 –19.5 219 199 –31.1 184 153
Dec 288 49.8 942 24.7 –23.9 253 229 –1.5 150 148
2008 Jan 11 28.3 1290 24.6 16.6 270 287 –27.6 168 140
Feb 11 24.0 1152 24.6 54.5 144 199 –42.2 155 112
Mar 4 22.3 1350 24.9 69.0 72 141 –27.0 135 108
Apr 72 22.1 1249 25.5 98.4 68 166 8.7 119 127
May 289 30.7 985 25.5 62.8 173 236 –3.4 183 180
Jun 230 42.8 781 25.1 8.6 210 219 –43.7 199 155
Jul 356 49.9 825 24.6 –19.5 239 220 –58.1 206 148
Aug 263 50.3 874 24.9 –9.8 225 215 –59.4 207 148
Sep 203 49.5 934 25.5 –7.9 225 218 –44.7 189 145
Oct 241 49.1 945 25.2 –16.0 260 245 –61.2 199 138
Nov 338 51.9 680 24.3 –9.2 202 193 –57.2 169 112
Dec 53 45.9 1149 25.1 13.4 257 270 –26.0 154 128
2009 Jan 13 30.2 1125 24.7 31.7 211 242 –30.7 153 122
Feb 20 29.1 1121 25.3 71.6 94 166 –40.7 133 93
Mar 12 25.8 1373 25.2 75.9 51 127 –11.2 114 103
Apr 94 23.9 1235 25.8 87.2 25 112 22.1 86 108
May 239 32.0 958 25.7 73.4 145 219 13.3 148 162
Jun 238 39.1 953 25.5 33.2 177 210 –60.8 212 151
Jul 309 – 982 26.1 –3.3 226 223 – – –
Aug 286 – 923 25.7 –22.1 233 210 – – –
Sep 211 – 913 26.0 –36.5 231 194 – – –
Oct 296 – 894 25.3 –40.7 251 211 – – –
Nov 486 – 595 24.7 –10.5 202 192 – – –
Dec 30 – 1112 25.5 –0.2 246 245 – – –
CARBON SEQUESTRATION OF PASTURE AND AFFORESTATION 2767
r2011 Blackwell Publishing Ltd, Global Change Biology,17, 2763–2780
from about January to April. The two transition periods
were characterized by highly variable but limited
amounts of precipitation (o4mmd
–1
). Compared with
the long-term annual mean from nearby Salamanca
(2267 mm, 1972–2009; derived from STRI 2010), Sardinilla
received above average rainfall in 2007 (2553 mm,113%),
below average rainfall in 2008 (2074 mm, –9%) and about
average rainfall in 2009 (2233 mm, 1%).
H
Wet season Dry Wet season Dry Wet season Dry
|GPP|, TER, NEE
(g C m−2 day−1)
−2
0
2
4
6
8
10
GPP
TER
NEE
(a)
Pasture
June 2007 − January 2010
T T
|GPP|, TER, NEE
(g C m−2 day−1)
−2
0
2
4
6
8
10 GPP
TER
NEE
(b)
Afforestation
PPFD
(mol m−2 day−1)
0
10
20
30
40
50 PPFD (c)
Precipitation
(mm wk−1)
0
50
100
150
0
10
20
30
40
50
SWC (%)
SWC (d)
Jun Aug Oct Dec Feb Apr Jun Aug Oct Dec Feb Apr Jun Aug Oct Dec Feb
2007 2008 2009 2010
Fig. 1 Daily total gross primary production (|GPP|, absolute value), total ecosystem respiration (TER) and net ecosystem exchange
(NEE; full shading indicates periods of carbon sinks, striped periods of carbon sources) of the Sardinilla pasture (a) and afforestation (b)
sites. Daily total of photosynthetic photon flux density (PPFD) is shown from June 2007 to January 2010 (c). |GPP|, TER, NEE and PPFD
are displayed as 14-days running means. Black arrows denote first day of periods with overgrazing (44LUha
–1
day
–1
) and ‘H’ combined
with grey arrow marks the day of herbicide application by the farmer at the pasture. ‘T’ combined with the black arrow indicates periods
of understory thinning at the afforestation. Weekly precipitation (grey bars) and weekly mean volumetric soil water content (SWC;
Sardinilla afforestation, at 5cm depth) are given (d). Measurements at the afforestation were discontinued after June 2009. The inserts at
the top indicate the different seasons (wet, dry) including transition periods (shaded areas).
2768 S. WOLF et al.
r2011 Blackwell Publishing Ltd, Global Change Biology,17, 2763–2780
Soil water content (SWC) at 5 cm depth exceeded 40%
during most of the wet season (mean 46%, maximum of
52% in November) and rapidly declined to below 30%
after the onset of the dry season (mean 26%, minimum
of 22% in April; Table 1, Fig. 1d). Following the first
rainfalls during the dry-wet transition period in April,
SWC started to increase swiftly and exceeded 40% by
June. During the dry season in 2009, SWC declined less
compared with 2008, which was related to moderate
precipitation events in February. SWC in deeper soil
layers (30 cm, not shown) was higher compared with
SWC in 5 cm depth during the dry season and about
similar during the wet season. Daily total photosyn-
thetic photon flux density (PPFD) ranged from a mini-
mum of 5.2 mol m
–2
d
–1
in November (wet season) to a
maximum of 58.5 mol m
–2
d
–1
in March (dry season; Fig.
1c). PPFD exceeded 40 mol m
–2
d
–1
during most of the
dry season (mean 41.8 mol m
–2
d
–1
) and was reduced
during the wet season (mean 28.6 mol m
–2
d
–1
). Seasonal
temperature variations at Sardinilla were small and
within 11C of the annual mean of 25.2 1C (2007–
2009), with the lowest values generally occurring dur-
ing November to March (Table 1). The diurnal tempera-
ture range in Sardinilla was larger than seasonal
variations and ranged from 22.2 1C (nighttime) to
27.4 1C (daytime) during the dry season and 23.5 1Cto
26.8 1C during the wet season, respectively.
Seasonal patterns in GPP, TER and NEE
We observed strong seasonal variations of GPP, TER and
net ecosystem exchange (NEE) in both pasture and
afforestation ecosystems. Seasonal variations at both sites
were larger during the year 2009 than in 2008 (Fig. 1).
At the pasture site, daily NEE ranged from a mini-
mum of –4.6 g C m
–2
d
–1
during the wet season to a
maximum of 8.3 g C m
–2
d
–1
during the dry-wet transi-
tion period (mean of 0.5 g C m
–2
d
–1
). During the wet
season, average GPP and TER were on the same order
of magnitude with seasonal means of 7.0 g C m
–2
d
–1
.
GPP dropped several times by up to 30% during the wet
seasons 2007 and 2008 (Fig. 1). We observed maximum
values for GPP and TER during the wet-dry transition
period, with 8.5 and 8.7 g C m
–2
d
–1
respectively (seaso-
nal means). During the dry season, GPP was limited by
water availability and declined to a minimum of
2.0 g C m
–2
d
–1
(seasonal mean) during the dry-wet tran-
sition period with predominantly senescent pasture
vegetation (LAI 50.6). TER was reduced during the
dry season as well (mean 5.6 g C m
–2
d
–1
) but exceeded
GPP (mean 3.7 g C m
–2
d
–1
), resulting in positive NEE
and thus, carbon release from the ecosystem. With the
first rainfalls during the dry-wet transition period, TER
increased rapidly and reached the level of the mean
seasonal TER within about 1 month (Fig. 1). However,
with most of the pasture grasses senescent, GPP did not
increase for another 1–2 weeks. Maximum carbon losses
occurred during the dry-wet transition period. Overall,
mean daily TER (6.89 1.38 g C m
–2
d
–1
) was similar to
GPP (6.39 2.59 g C m
–2
d
–1
).
At the afforestation site, we observed smaller seasonal
variations of NEE, GPP and TER than at the pasture site
with lower absolute values in general. Daily NEE ranged
from –5.4 g C m
–2
d
–1
during the wet season to
3.7 g C m
–2
d
–1
during the dry-wet transition period
(mean of 1.0 g C m
–2
d
–1
). During the wet season, GPP
consistently exceeded TER with 6.3 vs. 4.7 g C m
–2
d
–1
,
except during December 2007 (Fig. 1). Accordingly, NEE
was negative and the afforestation acted as a carbon sink
throughout the wet season. GPP and TER peaked during
the early wet season in June and July, after the leaves of
all tree species had fully developed (LAI of 6.0 in 2009).
Besides a second wet season maximum of GPP during
October, GPP and TER declined gradually during the
wet season. During the wet-dry transition, TER in-
creased while GPP remained stable. This increase was
particularly strong in December 2008 and reduced NEE
to –0.6 g C m
–2
d
–1
(seasonal mean). During the dry
season, GPP initially increased (in 2007) or remained
relatively constant (in 2009) during the early dry season,
while TER was declining. However, as GPP exceeded
TER, the afforestation maintained carbon uptake during
most of the dry season, with a mean NEE of –0.9 g C m
–
2
d
–1
. During the dry-wet transition, TER increased
strongly following the first rainfall while GPP did not
increase for another 1–2 weeks. During the dry-wet
transition period the afforestation was a carbon source
(mean NEE of 0.6 g C m
–2
d
–1
).
Overall, we observed a strong coupling of daily TER
with GPP during the dry season. This coupling was
stronger at the pasture (R
2
50.85, Po0.001) compared
with the afforestation (R
2
50.26, Po0.001). During the
wet season, however, daily variations in TER were only
weakly correlated with GPP at both sites
Carbon budgets
The pasture ecosystem was a substantial carbon source
and lost 470 g C m
–2
from June 2007 to December 2009.
Inter-annual variation in the carbon budget was small,
with 261 g C m
–2
yr
–1
in 2008 and 260 g C m
–2
yr
–1
in 2009
(Fig. 2). Using the mean NEE from January to May 2008
and 2009, we estimated an annual carbon budget of
251 g C m
–2
yr
–1
for 2007. The pasture ecosystem lost
carbon during most of the year except during the late
wet season. Seasonal carbon budgets indicated that more
carbon was lost during the dry season 2009 compared
with 2008 but losses were in reverse order during the wet
CARBON SEQUESTRATION OF PASTURE AND AFFORESTATION 2769
r2011 Blackwell Publishing Ltd, Global Change Biology,17, 2763–2780
season 2008 compared with 2009 (Table 2). Total monthly
NEE ranged from –2.3 g C m
–2
mo
–1
during the wet
season to 92.8 g C m
–2
mo
–1
during the dry-wet transition
period (overall m ean 15.2 g C m
–2
mo
–1
, Table 1).
The afforestation ecosystem was a strong carbon
sink from June 2007 to June 2009 (–750 g C m
–2
). Total
annual NEE was –442 g C m
–2
in 2008 and we estimated
the annual budgets (see pasture) for 2007 and 2009 to
–292 and –419 g C m
–2
, respectively (Fig. 2). The affor-
estation was a continuous carbon sink during most of
the year, except the end of the dry season, the dry-wet
transition period and in December 2007. Seasonal
budgets indicated higher carbon losses during the
dry-wet transition and onset of the wet season in
2009 compared with 2008 (Table 2). Carbon uptake
increased from 2007 to 2009 due to reductions in
TER, primarily during the wet season. Monthly NEE
ranged from –36.0 g C m
–2
mo
–1
during the wet season
to 15.4 g C m
–2
mo
–1
during the dry-wet transition
period (overall mean –30.0 g C m
–2
mo
–1
, Table 1).
Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec Jan
Dry Wet season
Cumulative NEE (g C m−2)
−400
−200
0
200
400 2007
2008
2009
Pasture
Afforestation
2007: 251 g C m−2
2008: 261 g C m−2
2009: 260 g C m−2
2007: −292 g C m−2
2008: −442 g C m−2
2009: −419 g C m−2
Source
Sink
Fig. 2 Cumulative annual net ecosystem exchange (NEE) of the Sardinilla pasture and afforestation sites from 2007 until 2009. Numbers
displayed denote annual total budgets with grey indicating years only partly measured. Annual budgets for those years have been
estimated using the respective periods from prior and following and years. Annual budget uncertainties are estimated to be
below 100 g C m
–2
yr
–1
.
Table 2 Season length (d), precipitation sum (P), mean of total daily photosynthetic photon flux density (PPFD) and seasonal total
net ecosystem exchange at the pasture (NEE.Pa) and afforestation site (NEE.Aff)
Dates Length
(d)
P
(mm)
PPFD
(mol m
–2
d
–1
)
NEE.Pa
(g C m
–2
)
NEE.Aff
(g C m
–2
)
2007
Wet season 23.04.–28.12. 250 2471 27.3
*
–43.3
*
–198.8
*
Wet-dry transition 29.12.–17.01. 20 17 40.8 3.5 –8.2
2008
Dry season 18.01.–03.04. 77 17 42.1 135.3 –89.6
Dry-wet transition 04.04.–28.04. 25 51 41.8 81.3 7.3
Wet season 29.04.–05.12. 221 1964 28.2 22.0 –337.2
Wet-dry transition 06.12.–05.01. 31 34 37.2 17.6 –18.4
2009
Dry season 06.01.–19.04. 104 42 41.6 208.9 –75.4
Dry-wet transition 20.04.–29.04. 10 37 34.8 46.5 15.2
Wet season 30.04.–30.11. 215 2122 29.0 –1.7 –44.8
*
Wet-dry transition 01.12.–03.01. 34 32 36.6 –1.4 –
*
Incomplete, only partial temporal coverage
2770 S. WOLF et al.
r2011 Blackwell Publishing Ltd, Global Change Biology,17, 2763–2780
During the entire dry season and the beginning of the
wet season, the pasture was a persistent and strong
source of CO
2
. In contrast, the afforestation was a net
carbon sink that was only occasionally interrupted
during this period (Fig. 1).
Environmental controls of GPP, TER and NEE
SWC and VPD. A strong correlation between monthly
mean SWC at 5 cm depth and monthly total NEE was
found at the pasture site (R
2
50.84, Po0.001; not
shown). Below a threshold of about 47% in monthly
mean SWC at 5 cm depth, the pasture ecosystem
became a source of carbon. The ecosystem response of
NEE to SWC at 30 cm depth was weaker (R
2
50.64,
Po0.001) compared with 5 cm depth at the pasture site.
SWC at 5 cm depth explained less variation in monthly
total GPP (R
2
50.59, Po0.001) and TER (R
2
50.20,
Po0.05) compared with NEE. In addition, we found
vapour pressure deficit (VPD) weakly related to NEE
(R
2
50.24, Po0.01) and GPP (R
2
50.23, Po0.01).
Similar but weaker relationships were found on
weekly but not on shorter timescales (daily and half-
hourly).
At the afforestation site, monthly mean SWC at 30 cm
depth showed a stronger relationship to monthly total
NEE (R
2
50.39, Po0.01) compared with SWC at 5 cm
depth (R
2
50.26, Po0.01). GPP was associated even
stronger with monthly variations in SWC at 30 cm
depth (R
2
50.57, Po0.001). TER was weakly related to
SWC at 5 cm depth only (R
2
50.21, Po0.05). Unlike at
the pasture site, no significant relationship of VPD with
NEE was observed, but with GPP (R
2
50.48, Po0.001)
and TER (R
2
50.49, Po0.001).
PPFD, Ecosystem light response of NEE. We observed a
strong seasonality in ecosystem light response to
photosynthetic photon flux density (PPFD) with
differences between the two sites (Fig. 3). The pasture
ecosystem showed a relatively weak light response
during the dry season with a high light compensation
point (LCP) of 400 mmol m
–2
s
–1
and light saturation
(due to limited carboxylation rate) at about
1500 mmol m
–2
s
–1
. This was even more pronounced
during the dry-wet transition, light compensation was
actually never achieved and TER constantly exceeded
photosynthesis. During the wet season, the
photosynthetic efficiency was most pronounced with a
Pasture
PPFD (µmol m−2 s−1)
NEEdaytime (µmol m−2 s−1)
0 500 1000 1500 2000 2500
−40
−20
0
20
40
Afforestation
PPFD (µmol m−2 s−1)
0 500 1000 1500 2000 2500
Fig. 3 Seasonally averaged light response curves (LRC) for Sardinilla pasture and afforestation sites from 2007 to 2009. Daytime net
ecosystem exchange (NEE
daytime
) is displayed as a function of photosynthetic photon flux density (PPFD). Symbols denote half-hourly
measurements. LRC were estimated using a nonlinear logistic sigmoid function with seasonal fitting parameters: maximum
photosynthetic capacity (A
max
), saturated photosynthetic capacity (A
sat
), daytime total ecosystem respiration (TER) and light compensa-
tion point (LCP, all in mmol m
–2
s
–1
). In addition, seasonal LCPs are represented by the dotted grey lines.
CARBON SEQUESTRATION OF PASTURE AND AFFORESTATION 2771
r2011 Blackwell Publishing Ltd, Global Change Biology,17, 2763–2780
low LCP of 242 mmol m
–2
s
–1
and high values of A
sat
(–19.4 mmol m
–2
s
–1
) and TER
daytime
(7.9 mmol m
–2
s
–1
).
Basically light saturation was not reached during
the wet-dry transition period, which exhibited the
highest rates of A
max
(–31.4 mmol m
–2
s
–1
) and A
sat
(–22.5 mmol m
–2
s
–1
) along with TER
daytime
(8.9 mmol
m
–2
s
–1
).
The afforestation ecosystem exhibited less seasonal
variation in light response and we observed overall
lower LCPs compared with the pasture (Fig. 3). During
the dry season, A
sat
(–8.4 mmol m
–2
s
–1
) exceeded
TER
daytime
(3.5 mmol m
–2
s
–1
) and light saturation was
reached at approximately 1500 mmol m
–2
s
–1
, similar to
the pasture. The highest LCP at the afforestation was
observed during the dry-wet transition (330mmol m
–2
s
–1
)
and light saturation was reached at about 1200 mmol
m
–2
s
–1
. During the wet season, the light response of
NEE at the afforestation ecosystem was most pronounced
with a low LCP (189 mmol m
–2
s
–1
) and the highest seaso-
nal values of A
max
(–22.0 mmol m
–2
s
–1
), A
sat
(–15.9 mmol
m
–2
s
–1
)andTER
daytime
(6.1 mmol m
–2
s
–1
). However, the
light response during the wet-dry transition period was
very limited and similar to the light response during the
dry season.
Management controls of GPP and TER
Grazing. Grazing was the main management factor that
influenced GPP and TER at the pasture site. Grazing
varied substantially from June 2008 to January 2010,
from zero to 75 livestock per day (median: 18.1) on the
6.5 ha pasture. This corresponds to a median and
maximum of 1.6 and 8.0 LU ha
–1
d
–1
, respectively
(Fig. 4b). Periods of average grazing were constrained
by isolated periods of overgrazing (44LUha
–1
d
–1
).
Lower stocking densities were observed in 2008
(median: 1.2 LU ha
–1
d
–1
) compared with 2009
(2.0 LU ha
–1
d
–1
). On the other hand, periods of
overgrazing persisted longer in 2008 (up to 9 days)
compared with 2009 (up to 3 days).
Overgrazing strongly reduced GPP of the pasture
ecosystem, particularly during the wet season 2008 (Fig.
1a). We observed a strong correlation between GPP and
grazing intensity during the major part of the wet
season 2008, with GPP being significantly reduced
when the pasture was overgrazed. The significant
reduction in GPP started immediately, reached a maxi-
mum after 4 days (R
2
50.42) of the start of overgrazing,
and lasted for an average for 6 days. TER was reduced
as well but was delayed with respect to the beginning of
overgrazing. A significant reduction in TER started after
2 days, had its maximum after 6 days (R
2
50.23) and
lasted on average for 9 days. Along with GPP, NEE
showed less net uptake, which was most pronounced
after 1 day (R
2
50.37) and lasted for 5 days. No signi-
ficant time lag of GPP vs. grazing was observed during
the wet season 2009. Furthermore, we observed an
apparent reduction in ecosystem light response during
and shortly after overgrazing: When excluding over-
grazing periods with confounding limitations by envir-
onmental controls (namely PPFD), the pasture exhibited
a reduction in daily photosynthetic capacity (A
max
),
Wet season Dry Wet season Dry
0
100
200
300 (a)
Biomass (g C m−2)
Pasture, June 2008 − January 2010
0
2
4
6
8(b)
Grazing (LU ha−1 day−1)
Jun Aug Oct Dec Feb Apr Jun Aug Oct Dec Feb
2008 2009 2010
Fig. 4 Aboveground green biomass (a) and grazing (in livestock units, LU; b) at the Sardinilla Pasture from June 2008 to January 2010.
The dashed black lines denote the overall median for biomass (a) and the annual median for grazing (b). The dotted grey line shows the
overgrazing threshold of 4 LU.
2772 S. WOLF et al.
r2011 Blackwell Publishing Ltd, Global Change Biology,17, 2763–2780
increased TER
daytime
and increased the LCP. For
instance, during overgrazing in September 2008 (DOY
267–270, 7.7 LU ha
–1
), A
max
was reduced from 31.5 to
26.9 mmol m
–2
s
–1
, TER increased from 10.7 to
14.5 mmol m
–2
s
–1
and the LCP increased from 311 to a
maximum of 597 mmol m
–2
s
–1
. We used this period of
overgrazing combined with the reduction in above-
ground biomass (DOY 259–274) and estimated an aver-
age forage consumption of 0.81 g C m
–2
d
–1
(for 1 LU) or
475 g C m
–2
yr
–1
, when using the median stocking den-
sity of 1.6 LU ha
–1
for Sardinilla.
Herbicide application. Another management factor at the
pasture site was herbicide application by the farmer.
The recovery of the vegetation after the dry season
was dominated by the fast growing pioneering weed
Croton hirtus, which inhibits the recovery of other
grasses. In 2009, herbicide was applied by the farmer
on June 1 and within about 2–3 weeks, the weed died
off. Following the herbicide application, we observed a
reduction in GPP by about 15%, an increase in TER by
about 10% and a reduction in LAI from 2.75 0.19
(June 8th) to 2.26 0.15 (June 18th). No exact dates
for herbicide application in 2007 and 2008 could be
obtained.
Understory thinning. Thinning of the understory (weed
removal) at the end of the year was the only
management intervention at the afforestation site. No
carbon was exported, and all clipped biomass was left
on site. A major thinning (full afforestation area) was
carried out in December 2007. This reduced GPP
substantially by about 50% and turned the
afforestation site into a carbon source (Fig. 1) for a
period of 17 days. Understory thinning in December
2008 was minor and included only parts of the
afforestation site. We observed an accompanying
reduction in GPP as well but less strong than in 2007.
This difference was largely due to lower TER in
December 2008. However, TER increased strongly
following the thinning in 2008, indicating enhanced
microbial activity due to decomposing litter.
Inventory data
Biomass. Significant seasonal variations in aboveground
green biomass were observed at the pasture site,
ranging from zero during the dry season and dry-wet
transition period (senescent vegetation) to a maximum
of 254 g C m
–2
during the wet-dry transition period,
with an overall median of 178 77 g C m
–2
(Fig. 4a).
With re-growing vegetation after onset of the wet
season, average aboveground biomass was reached
during July and lasted until January, except for
periods with pronounced overgrazing during the wet
season 2008. Substantial reductions in aboveground
biomass by up to 20% were observed following
periods of overgrazing, e.g. in September 2008 from
204 8.9 g C m
–2
(DOY 259) to 163 8.7 g C m
–2
(DOY
274), when overgrazing occurred from DOY 267–270.
Aboveground biomass and LAI were strongly
correlated at the pasture site (R
2
50.84, Po0.001).
Belowground biomass was determined once in March
2009 (n51) and was with 176 g C m
–2
similar to average
aboveground biomass.
At the afforestation, total aboveground biomass
more than tripled from 772 136 g C m
–2
in 2005 to
2449 891 g C m
–2
in 2009 (Table 3). Belowground bio-
mass (coarse roots) increased from 112 63 g C m
–2
in
2005 to 693 306 g C m
–2
in 2009. In 2008, tree biomass
(above and belowground) was the largest component of
ecosystem biomass with 1533 705 g C m
–2
(72.3%),
followed by understory vegetation with 546
89 g C m
–2
(25.7%), coarse woody debris (CWD) with
14.8 12.6 g C m
–2
(0.7%) and litter with 27.9
20.0 g C m
–2
(1.3%; Fig. 5). The annual increase in tree
biomass was highest in 2006 ( 169.4%) and 2009
(166.6%), while it was smaller in 2007 ( 155.9%) and
Table 3 Above- and belowground standing biomass at the Sardinilla afforestation from 2005 to 2009
2005 (g C m
–2
) 2006 (g C m
–2
) 2007 (g C m
–2
) 2008 (g C m
–2
) 2009 (g C m
–2
)
Trees 299 168 506 269 792 419 1116 522 1861 879
Herbaceous biomass 443 109 401 65 539 82 546 89 546 89
*
Coarse woody debris 2.1 21.8 18.3 14.8 14.8
*
Litter 27.9 20.0
Total aboveground biomass 772 136 957 285 1377 448 1705 537 2449 891
Roots 112 63 190 98 294 144 417 188 693 306
Total biomass 884 63 1147 377 1671 524 2122 719 3143 1189
Total biomass increment (g C m
–2
yr
–1
) – 263 225 524 275 450 324 1021 556
Values indicate mean standard deviation. No data on herbaceous biomass and coarse woody debris were available for 2009 and
thus, data from 2008 were used to estimate total biomass (
*
)
CARBON SEQUESTRATION OF PASTURE AND AFFORESTATION 2773
r2011 Blackwell Publishing Ltd, Global Change Biology,17, 2763–2780
lowest in 2008 (41.2%). The relative contribution
of understory vegetation to the total aboveground bio-
mass was reduced considerably from 50.1% in 2005 to
17.4% in 2009. However, total herbaceous biomass
increased strongly from 2006 to 2007 ( 134.5%), due
to the invasion of Saccharum spontaneum (Canal grass)
in parts of the afforestation site. Overall, annual
ecosystem biomass increments varied strongly from
263 225 g C m
–2
in 2006, to a maximum of
1021 556 g C m
–2
in 2009. The large increment in
2007 was primarily associated with the strong increase
in herbaceous biomass.
Soil. A strong reduction in topsoil (0–10 cm) carbon
pools was found over a period of 8 years since 2001/
2002 at both sites. This reduction was stronger at the
pasture site (reduction of 1250 g C m
–2
or –46%) as
compared with the afforestation (930 g C m
–2
or –28%,
Table 4). Correspondingly, nitrogen pools decreased
more strongly at the pasture compared with the
afforestation by 120 g N m
–2
(–44%) and 80 gN m
–2
(–30%), respectively. At the afforestation, the stable
carbon isotope ratio (d
13
C) became significantly
more depleted from 2001 (–17.0 0.8%) to 2009
(–20.9 1.2%). This change in d
13
C indicated that in
2001 about 83% of the organic matter in the topsoil was
derived from C
4
pasture vegetation, whereas this
contribution had decreased to 57% by 2009. As
expected, no significant change ind
13
C was observed
at the pasture site. The total soil carbon pool from
0–100 cm depth was clearly lower at the pasture
(5.36 0.18 kg m
–2
) than at the afforestation site
(7.64 1.63 kg m
–2
). At both sites, roughly one-third of
the carbon pool was concentrated in the top 10 cm of
soil (27.6% and 33.0% for pasture and afforestation,
respectively).
RPlant
AfforestationPasture
GPP
–2082
TER
1640
NEE
–442
GPP
–2345
TER
2606
NEE
261
BM 178
BM 176
BM 475
BM 1533 (+447)
RSoil
50%
BM 417 (+123)
BM 1116 (+324)
BM 43 (–3)
BM 546 (+7)
SOM 1480 (–156)
SOM 3880
SOM 2520 (–116)
SOM 5120
RSoil
51%
RPlant
Inventory increment +335
Fig. 5 Carbon stocks (g C m
–2
) and fluxes (g C m
–2
yr
–1
) in the Sardinilla pasture and afforestation in 2008. Grey numbers denote carbon
fluxes with net ecosystem exchange (NEE) and its components gross primary production (GPP), and total ecosystem respiration (TER).
Wide arrows indicate the size and direction of the fluxes, slim arrows the origin of the TER components soil (R
Soil
) and plant respiration
(R
Plant
); the percentage denotes the measured source contribution of soil respiration. Numbers in boxes denote carbon stocks with annual
increments (g C m
–2
yr
–1
) given in brackets. Soil organic matter (SOM) is reported for topsoil (0–10 cm) and 10–100 cm depth. Above-
ground biomass (BM
AG
) in the afforestation is separated in tree biomass (BM
T
), herbaceous biomass (BM
H
), coarse woody debris
(BM
CWD
) and litter (BM
L
). Belowground biomass (BM
BG
) is reported for trees only in the afforestation (BM
AG.T
). In the pasture, foraged
biomass (BM
F
) by grazing livestock was estimated using the biomass reduction during periods of overgrazing. The ‘Inventory increment’
incorporates biometric uptake minus soil carbon losses.
2774 S. WOLF et al.
r2011 Blackwell Publishing Ltd, Global Change Biology,17, 2763–2780
Discussion
The pasture under investigation was heavily grazed at
an intensity which cannot be considered sustainable
under current conditions. In addition, water limitations
led to strong and persistent carbon losses during the
dry season that continued into the first weeks of the wet
season. Due to overgrazing, carbon uptake of the
pasture during the wet season was not sufficient to
compensate carbon losses during the dry season. In
contrast, the afforestation site persistently sequestered
large amounts of carbon as measured with the eddy-
covariance method and supported by biometric obser-
vations.
Environmental controls
The main environmental controls of ecosystem CO
2
fluxes in Sardinilla were PPFD and SWC. Considerable
differences in the ecosystem response of pasture and
afforestation to seasonal limitations in soil moisture
were found, with the pasture ecosystem being more
susceptible to seasonal drought than the afforestation.
During the dry season, GPP at the pasture was strongly
reduced whereas the afforestation maintained a GPP
that exceeded TER well into the dry season (for about 75
days in 2008 and 81 days in 2009). This seems strongly
related to the rooting depth of grasses vs. trees as we
observed a mean rooting depth of only about 10–20 cm
at the pasture. In comparison, mean rooting depth at the
afforestation was 1.4 m in 2009 (Jefferson Hall, personal
communication, unpublished data).
Light reduction due to cloudiness during periods of
intense precipitation (e.g. in November) strongly
reduced GPP at both ecosystems. Similar reductions in
GPP were observed at tropical forest sites in Amazonia
(Malhi et al., 1998) and Costa Rica (Loescher et al., 2003)
during the wet season. However, the ecosystem
response to varying light and soil moisture conditions
was very different between pasture and afforestation
(Fig. 3). This seems predominantly related to the differ-
ent photosynthetic pathways of C
4
grasses and C
3
trees.
C
3
species are generally most active during the early
growing season while C
4
species increase photosyn-
thetic activity at warmer and drier conditions (Lambers
et al., 2008). An additional explanation for the limited
light response of the afforestation during the wet-dry
transition period could be the age of the foliage, as the
chlorophyll concentration per unit leaf area decreases
with age (Lambers et al., 2008). It is likely that this is also
the case with grasses but the grazed pasture vegetation
has higher turnover rates and thus persistent re-growth
of plant tissues.
GPP and TER patterns. Total annual GPP at the Sardinilla
pasture was higher compared with the afforestation
(Table 5) but lower than a tropical pasture in Brazil
reported by Gilmanov et al. (2010) and Grace et al.
(1998). Annual GPP at the afforestation was consistent
with results reported by Vourlitis et al. (2001) for a
transitional tropical forest in Brazil but lower than the
range reported from tropical forests (Table 5). This is
likely due to the young age of the afforestation, which is
still in its establishment phase.
Overall higher TER in the pasture compared with
afforestation (Fig. 5) seems to be caused by enhanced
biomass turnover including decomposition rates due to
grazing. The observed strong increases of TER during
the dry-wet transition period are likely a combination of
physical and physiological effects: Firstly, highly con-
centrated CO
2
is pushed out of the soil pore space
(macro-pores and desiccation cracks; Birch effect).
Secondly, large amounts of organic material that accu-
mulated during the dry season start to decompose
rapidly with the suddenly increasing soil moisture.
The stronger increase in TER at the pasture compared
with the afforestation seems related to additional accu-
mulated organic material (manure) by grazing livestock.
Table 4 Topsoil (0–10 cm) characteristics at the Sardinilla pasture (2002 to 2010) and afforestation sites (2001 to 2009)
d
B
(g cm
–3
) C (%) C (kg m
–2
) N (%) N (kg m
–2
)C:N d
13
C(%)
Pasture
2002 0.86 0.07 3.21 0.45 2.73 0.35 0.31 0.05 0.27 0.03 10.44 0.70 –17.80 1.86
2010 – 1.72 0.33 1.48 0.14 0.17 0.04 0.15 0.02 10.11 0.16 –18.93 0.56
Afforestation
2001 0.59 0.05 5.85 0.52 3.45 0.24 0.49 0.03 0.29 0.02 11.93 0.68 –17.01 0.78
2009 – 4.24 0.92 2.52 0.57 0.36 0.08 0.21 0.05 11.86 1.32 –20.86 1.23
Values indicate mean standard deviation. Bulk density (d
B
), carbon concentration (C), carbon pool (C), nitrogen concentration (N),
nitrogen pool (N), C : N ratio (C:N) and carbon isotope ratio (d
13
C). Data from 2001/2002 are derived from Abraham (2004). Soil
sampling at the pasture was done in January 2010 and at the afforestation in March 2009. No bulk density data were available for
2009/2010; hence, values from 2001/2002 were used to calculate soil pools.
CARBON SEQUESTRATION OF PASTURE AND AFFORESTATION 2775
r2011 Blackwell Publishing Ltd, Global Change Biology,17, 2763–2780
Table 5 Summary of ecosystem CO
2
flux studies in the tropics
Site
(Country)
Location
(Lat., Long.)
Rainfall
(mm yr
–1
)
Vegetation
type Time LAI
TER
night
(mmol m
–2
s
–1
)
NEE
day
(g C m
–2
d
–1
)
GPP
annual
(g C m
–2
yr
–1
)
NEE
annual
(g C m
–2
yr
–1
)
Biomass
(g C m
–2
yr
–1
) References
Xishuangbanna
(China)
211550N, 1011160E 1487 TMF 2003–2006 – – – 2594 119 359 Tan et al. (2010)
La Selva (Costa
Rica)
101260N, 841990W 4000 TRF 1998–2000 2.7–4.9 7.05 – 3097 242
*
– Loescher et al.
(2003)
Sardinilla
Afforestation
(Panama)
91180N, 791380W 2267 P 2007–2009 1.2–2.9 6.5 0.5 2345w261w– This study
Sardinilla
Pasture
(Panama)
91190N, 791380W 2267 YF 2007–2009 3.0–5.4 4.7 1.0 2082w442 450wThis study
Guyaflux (French
Guiana)
51170N, 521540W 3041 TRF 2004–2005 6.9 10.0–15.0 0.4 3911z150 102§ Bonal et al.
(2006, 2008)
Caxiuana
˜(Brazil) 11430S, 511280W 2500 TRF 1999 (108 d) 5–6 7.6 – 3630 560 – Carswell et al.
(2002)
Cuieriras, C14
(Brazil)
21350S, 601070W 2200 TRF 1995–1996 5–6 6.5 – 3040 590 230 Malhi et al.
(1998, 1999)
Tapajos km67,
Santare
´m
(Brazil)
21510S, 541580W 1920 TRF 2002–2005 6–7 9.2 0.8 3157 94 200}Hutyra et al.
(2007)
Ducke (Brazil) 21570S591570W 2431 TRF 1987 (12 d) 5–6 5.95 2.2 – 220k– Fan et al. (1990)
Tapajos km83,
Santare
´m
(Brazil)
3130S, 541560W 1920 TRF
**
2000–2001 6–7 6.0–7.0 – 3000 130 200 Saleska et al.
(2003)
Cotriaguacu
´
(Brazil)
91520S, 581140W 2000 P–A 2002 (10 mo) 1.0–2.7 5.9 4.6 – – – Priante-Filho
et al. (2004)
Jaru, Rondonia
(Brazil)
101050S, 611570W 2170 TRF 1992–1993
(55 d)
4 6.6 0.8 2440 102 – Grace et al.
(1995, 1996,
1998)
FNS-A, Rondonia
(Brazil)
101460S, 621210W 2170 P 1993 (11 d) 1.1–3.9 6.0 1.9 4000 – – Grace et al.
(1998)
– – 1999–2002 – 3.8 1.8 – – – von Randow
et al. (2004)
– – 1999 – – – 3471 400ww – Gilmanov et al.
(2010)
Sinop (Brazil) 111250S, 551200W 2037 TF 1999–2002 2.5–5.0 5.0 1.2 2062 5 – Vourlitis et al.
(2001, 2004,
2005)
2776 S. WOLF et al.
r2011 Blackwell Publishing Ltd, Global Change Biology,17, 2763–2780
Carbon budget synthesis. The pasture ecosystem was a
large carbon source from 2007 to 2009. As far as we are
aware, the average annual loss of 261 g C m
–2
yr
–1
is the
first quantitative estimate for tropical pasture that
covers more than 1 year (Table 5). Only three other
studies reported total NEE, but on a daily base. All
found carbon uptake for pastures: Grace et al. (1998)
observed –1.9 g C m
–2
d
–1
in Amazonia during 11 days
in May 1993, and von Randow et al. (2004) found similar
uptake of –1.8 g C m
–2
d
–1
at the same pasture during
1999–2002. Priante-Filho et al. (2004) found an even
larger carbon uptake of –4.6 g C m
–2
d
–1
in a pasture
in conversion to afforestation.
The large carbon losses at the Sardinilla pasture
could be either associated with soil carbon, dissolved
organic carbon (DOC) or hidden abiotic factors like
weathering of calcareous bedrock as suggested by
Serrano-Ortiz et al. (2010). The substantial reduction in
topsoil carbon seems to be the main source for the
strong carbon losses (Fig. 5). Potential causes for the
discrepancy to the EC measured carbon losses are
measurement uncertainties or carbon export, such as
by livestock or DOC. However, Waterloo et al. (2006)
found that export of DOC played only a minor role in
the carbon budget of a tropical forest with similar
annual rainfall in Amazonia.
The seven-year-old Sardinilla afforestation was a
larger net carbon sink in 2008 than reported for most
tropical forests (Table 5). On the other hand, Carswell
et al. (2002) and Malhi et al. (1998, 1999) found larger
carbon sequestration while Saleska et al. (2003) and
Hutyra et al. (2007) found carbon losses in tropical
forests in Amazonia (Table 5). Consequently, the young
afforestation is sequestering substantial amounts of
carbon following its establishment phase that exceeds
other, mature tropical forests. It can be expected that the
carbon sink strength of the afforestation will continue
and might even increase until the trees reach maturity
(Canadell et al., 2007), provided that there are no dis-
turbances like fires, storms or harvesting. A long-term
(55–61 years) inventory-based study of Silver et al.
(2004) reported that carbon sequestration in a tropical
afforestation did not slow down with aging trees,
indicating significant carbon uptake (140 g C m
–2
yr
–1
)
even after the establishment phase.
Only few studies assessed comparative EC and
biometric carbon budgets in the tropics (Table 5). In
general, EC derived carbon budgets were lower than
biometric field estimates as observed by, for example,
Malhi et al. (1998, 1999) in Manaus (Amazonia), Saleska
et al. (2003) and Hutyra et al. (2007) in Santarem
(Amazonia), and Tan et al. (2010) in China (Table 5).
The only tropical study that reported a larger carbon
uptake with the EC method compared with biometric
– – 2002 (10 mo) 2.5–5.0 3.9 3.1 – – –
Priante-Filho
et al. (2004)
Emendadas
(Brazil) 151330S, 471360W 1556 S 1993–1994 0.4–1.0 4.8 – – 250 –
Miranda et al.
(1996, 1997)
– – 1999–2000 – 7.6 1.3 – – –
von Randow
et al. (2004)
All studies except the one from Xishuangbanna (China) were located in the neotropics. The vegetation types are tropical rainforest (TRF), Young forest (YF, o10 years),
Transitional tropical forest (TF), Tropical forest with monsoonal climate (TMF), Pasture (P), Pasture with starting afforestation (P-A) and Savanna (S). Further listed are leaf area
index (LAI), mean of nighttime total ecosystem respiration (TER
night
), daily mean net ecosystem exchange (NEE
day
), annual total gross primary production (GPP
annual
), annual
total net ecosystem exchange (NEE
annual
) and total above and belowground biomass increment; except Guyaflux: only aboveground biomass increment. Missing information on
location and climate was supplemented from the FLUXNET database (http://www.fluxnet.ornl.gov).
*
Mean of 1998–2000.
wValues reported from 2008.
zReported by Gilmanov et al. (2010).
§Aboveground biomass of 70 g C m
–2
yr
–1
reported only, adding belowground estimate (45% of aboveground, (Malhi & Grace 2000) yields 102 g C m
–2
yr
–1
.
}Reported by Saleska et al. (2003).
kDerived from Table 1 in Malhi et al. (1998); modelled annual budget.
**
Selectively logged in Sept. 2001 (reported measurements were done before).
wwEstimated from Fig. 7f in Gilmanov et al. (2010).
CARBON SEQUESTRATION OF PASTURE AND AFFORESTATION 2777
r2011 Blackwell Publishing Ltd, Global Change Biology,17, 2763–2780
data was Bonal et al. (2006, 2008) in French Guinea.
Including the topsoil carbon losses with the biometric
carbon uptake yields a similar result in Sardinilla, with
a larger carbon uptake measured by EC compared with
inventory data in 2008 (–335 g C m
–2
yr
–1
, Fig. 5). Only
the study by San Jose
´et al. (2008) reported close agree-
ment of EC and biometric-derived carbon budgets for a
tall-grass Andropogon site and a savanna–woodland
continuum in the Orinoco lowlands. However, and as
emphasized by Saleska et al. (2003), large uncertainties
are associated with both methods.
The change in topsoil d
13
C from 2001 to 2009 at the
afforestation clearly indicates the increased inputs of
organic matter (litter) by the dominating C
3
-vegetation
in combination with the rapid carbon turnover in this
tropical ecosystem. Besides the considerable reductions
in topsoil carbon pools at both sites, it should be noted
that topsoil generally constitutes only a small amount of
the total soil carbon pool in the tropics. In Sardinilla, the
topsoil carbon pool represents about 30% of the total
carbon pool down to 1 m depth, which is more or less
consistent with about 25% found for tropical forest at
Barro Colorado Island (B. Turner, unpublished data).
Measurement uncertainties in the EC carbon budgets
are largely related to ecosystem respiration and its
consideration in data quality filtering and gap filling.
We observed only a weak temperature sensitivity of
nighttime ecosystem respiration, which is consistent
with other tropical studies (Hutyra et al., 2007) and
the relatively low temperature sensitivity of tropical
forest soils as reported by Davidson & Janssens (2006).
Our results indicate that using weak temperature sen-
sitivities to gap-fill nighttime ecosystem respiration
could result in large biases of carbon budgets in tropical
ecosystems. Hence, alternative running mean
approaches should be considered more comprehen-
sively. Further bias in carbon budgets can originate
from advection (see e.g. Kruijt et al., 2004). However,
nighttime advection is probably only small at our
Sardinilla. The u
*
-filter and stationarity criterion used
in data quality filtering are assumed to account for
advection effects already (Aubinet, 2008; Etzold et al.,
2010). Further evidence for this is given by the energy
balance closure, which was found to be comparable
with other flux tower sites globally, with 84% and 81%
for pasture and afforestation, respectively (Wolf et al.,
2011), and the close agreement with independently
measured inventories.
Overall, the reported uncertainties in annual EC
budgets range from less than 50 g C m
–2
yr
–1
for
nearly ideal sites (relatively flat terrain) to 130–
180 g C m
–2
yr
–1
for non-ideal sites with hilly topogra-
phy (Baldocchi, 2003). Considering that both sites in
Sardinilla are nearly ideal from a topographic perspec-
tive, we conservatively estimate that our annual budget
uncertainties should be below 100 g C m
–2
yr
–1
.
Management impact on carbon cycling. Both ecosystems
exhibited strong responses to management, with
understory thinning at the afforestation and grazing at
the pasture site. Understory thinning considerably
reduced GPP and gave evidence that understory
vegetation accounts for a significant amount of GPP at
the young afforestation 6–7 years after establishment.
At the pasture site, periodical overgrazing during 2008
and persistently high stocking densities in 2009 were
the major cause for carbon losses. Vegetation recovery
was swift after periods of overgrazing during the wet
season but was inhibited by soil moisture during the
dry season. Hence, overgrazing during the dry season
reduced aboveground biomass without the potential of
recovery before the beginning of the wet season.
Similar rates of forage consumption by livestock like
at the Sardinilla pasture were found by Dias-Filho et al.
(2000) in the Amazon basin (0.74 g C m
–2
d
–1
). Wilsey
et al. (2002) reported lower forage consumption at
adjacent pastures in Sardinilla (0.61 g C m
–2
d
–1
) and
found that grazing significantly reduced ecosystem
respiration, but not carbon uptake. However, they also
emphasized that grazing does not necessarily increase
carbon losses from tropical pastures as the reduction in
aboveground biomass lowers ecosystem respiration
whereas grazing enhances aboveground productivity.
Kirkman et al. (2002) reported a decrease in carrying
capacity of about 50% from a cattle ranch in Brazil from
1992 to 2000, indicating a strong impact of grazing on
carbon and nutrient cycling of tropical pastures. If the
current losses in soil carbon and nitrogen continue, the
pasture in Sardinilla seems at high risk of irreversible
degradation. Consequently, a reduction of stocking
densities to a maximum of 1 LU ha
–1
appears crucial
for mitigation efforts to decrease carbon losses in this
highly seasonal climate.
Conclusions
We conclude that tropical afforestation can sequester
large amounts of carbon, reduce the intra-annual varia-
bility of GPP, and enhance the ecosystem resilience to
seasonal drought. High stocking densities in combina-
tion with seasonal drought can result in reduced
productivity and carbon losses from tropical pasture.
Projected changes in precipitation (reduction and
increased variability) for Central America might affect
the carbon balance of these tropical ecosystems in different
ways, i.e., the carbon source strength of pastures might
increase while the sink strength of afforestations might be
reduced. Furthermore, the conversion from pasture to
2778 S. WOLF et al.
r2011 Blackwell Publishing Ltd, Global Change Biology,17, 2763–2780
afforestation may become more relevant for Panama and
other Latin American countries in the future, within the
carbon accounting of the Kyoto protocol.
Acknowledgements
Funding for this project was provided by the North-South Centre
(former Swiss Centre for International Agriculture) of ETH
Zurich. Catherine Potvin acknowledges the support from a
Discovery grant of the Natural Sciences and Engineering Council
of Canada. We are grateful to the Smithsonian Tropical Research
Institute (STRI) for support with the Sardinilla site and the
Meteorology and Hydrology Branch of the Panama Canal
Authority (ACP), Republic of Panama, for providing meteoro-
logical data. We thank Timothy Seipel for identifying grass
species of the pasture, Jefferson Hall and his project funded by
SENACYT for providing information on rooting depth at the
afforestation site, Dayana Agudo and Tania Romero (STRI) for
laboratory support, Nicolas Gruber for inspiring discussions,
Rodrigo Vargas for valuable comments on the manuscript, and
Jose
´Monteza for tower maintenance and assistance in the field.
References
Abraham M (2004) Spatial variation in soil organic carbon and stable carbon isotope
signature in a pasture and a primary forest in Central Panama. Master, McGill
University, Montreal, 91 pp.
ACP (2010). Panama Canal Authority (ACP), Meteorology and Hydrology Branch
(Republic of Panama), Data provided by Mike Hart.
Alves DS, Morton DC, Batistella M, Roberts DA, Souza JRC (2009) The changing rates
and patterns of deforestation and land use in Brazilian Amazonia. In: Amazonia and
Global Change (eds Keller M, Bustamante M, Gash J, Dias PS), pp 11–23. American
Geophysical Union, Washington, DC.
ANAM (2010). Autoridad Nacional del Ambiente of Panama (ANAM).
Aubinet M (2008) Eddy covariance CO
2
flux measurements in nocturnal conditions: an
analysis of the problem. Ecological Applications,18, 1368–1378.
Baldocchi DD (2003) Assessing the eddy covariance technique for evaluating carbon
dioxide exchange rates of ecosystems: past, present and future. Global Change
Biology,9, 479–492.
Beer C, Reichstein M, Tomelleri E et al. (2010) Terrestrial gross carbon dioxide uptake:
Global distribution and covariation with climate. Science,329, 834–838.
Bonal D, Bosc A, Ponton S et al. (2006) The influence of severe seasonal soil drought on
ecosystem functioning in the tropical rainforest of French Guiana. Poster at Guyaflux
Summer School. Available at: http://www.ecofog.gf/fr/fonctionnement/guyaflux/
Bonal_etal_Poster%20Guyaflux-Summer%20School%20Eddy%202006.pdf (accessed
21 August 2010)
Bonal D, Bosc A, Ponton S et al. (2008) Impact of severe dry season on net ecosystem
exchange in the Neotropical rainforest of French Guiana. Global Change Biology,14,
1917–1933.
Canadell J, Pataki D, Gifford R et al. (2007) Saturation of the Terrestrial Carbon Sink. In:
Terrestrial Ecosystems in a Changing World (eds Canadell JG, Pataki DE, Pitelka LF),
pp 59–78. Springer, Berlin, Heidelberg.
Carswell FE, Costa AL, Palheta M et al. (2002) Seasonality in CO
2
and H
2
O flux at an
eastern Amazonian rain forest. Journal of Geophysical Research-Atmospheres,107,
doi:10.1029/2000JD000284.
Chilonda P, Otte J (2006) Indicators to monitor trends in livestock production at
national, regional and international levels. Livestock Research for Rural Development,
18, Article #117, Available at: http://www.lrrd.org/lrrd18/8/chil18117.htm,
(accessed 6 August 2010).
Davidson EA, Janssens IA (2006) Temperaturesensitivity of soil carbon decomposition
and feedbacks to climate change. Nature,440, 165–173.
Dias-Filho MB, Davidson EA, de Carvalho CJR (2000) Linking Biogeochemical cycles
to cattle pasture management and sustainability in the Amazon basin. In: The
Biogeochemistry of the Amazon Basin (eds McClain ME, Victoria RL, Richey JE),
pp 84–105. Oxford University Press, New York.
Don A, Rebmann C, Kolle O, Scherer-Lorenzen M, Schulze ED (2009) Impact of
afforestation-associated management changes on the carbon balance of grassland.
Global Change Biology,15, 1990–2002.
Etzold S., Buchmann N., Eugster W. (2010) Contribution of advection to the carbon
budget measured by eddy covariance at a steep mountain slope forest in Switzer-
land. Biogeosciences,7, 2461–2475.
Eugster W, Moffat AM, Ceschia E et al. (2010) Management effects on European
cropland respiration. Agriculture, Ecosystems & Environment,139, 346–362.
Eugster W, Senn W (1995) A cospectral correction model for measurement of turbulent
NO
2
flux. Boundary-Layer Meteorology,74, 321–340.
Fan SM, Wofsy SC, Bakwin PS, Jacob DJ, Fitzjarrald DR (1990) Atmosphere-biosphere
exchange of CO
2
and O
3
in the Central Amazon forest. Journal of Geophysical
Research-Atmospheres,95, 16851–16864.
FAO (2009), Food and Agriculture Organization of the United Nations (FAO), Rome.
168 pp
Fearnside PM (2005) Deforestation in Brazilian Amazonia: History, rates, and
consequences. Conservation Biology,19, 680–688.
Foken T, Wichura B (1996) Tools for quality assessment of surface-based flux
measurements. Agricultural and Forest Meteorology,78, 83–105.
Gilmanov TG, Aires L, Barcza Z et al. (2010) Productivity, Respiration, and Light-
Response Parameters of World Grassland and Agroecosystems Derived From Flux-
Tower Measurements. Rangeland Ecology & Management,63, 16–39.
Gilmanov TG, Johnson DA, Saliendra NZ (2003) Growing season CO
2
fluxes in a
sagebrush-steppe ecosystem in Idaho: bowen ratio/energy balance measurements
and modeling. Basic and Applied Ecology,4, 167–183.
Grace J, Lloyd J, McIntyre J et al. (1995) Carbon-dioxide uptake by an undisturbed
tropical rain-forest in Southwest Amazonia, 1992 to 1993. Science,270, 778–780.
Grace J, Lloyd J, Miranda AC, Miranda H, Gash JHC (1998) Fluxes of carbon dioxide
and water vapour over a C
4
pasture in south-western Amazonia (Brazil). Australian
Journal of Plant Physiology,25, 519–530.
Grace J, Mahli Y,Higuchi N, Meir P (2001) Productivity of tropical forests. In: Terrestrial
Global Productivity (eds Roy J, Saugier B, Mooney H), pp 401–426. Academic Press,
San Diego, CA, USA.
Grace J, Malhi Y, Lloyd J, McIntyre J, Miranda AC, Meir P, Miranda HS (1996) The use
of eddy covariance to infer the net carbon dioxide uptake of Brazilian rain forest.
Global Change Biology,2, 209–217.
Graham NE, Georgakakos KP, Vargas C, Echevers M (2006) Simulating the value
of El Nino forecasts for the Panama Canal. Advances in Water Resources,29,
1665–1677.
Gu LH, Falge EM, Boden T et al. (2005) Objective threshold determination for night-
time eddy flux filtering. Agricultural and Forest Meteorology,128, 179–197.
Hutyra LR, Munger JW, Saleska SR et al. (2007) Seasonal controls on the exchange of
carbon and water in an Amazonian rain forest. Journal of Geophysical Research-
Biogeosciences,112.
IPCC (2007) Climate Change 2007: Mitigation. Contribution of Working Group III to the
Fourth Assessment Report of the Intergovernmental Panel on Climate Change. Cambridge
University Press, Cambridge, United Kingdom.
Kirkman GA, Gut A, Ammann C et al. (2002) Surface exchange of nitric oxide, nitrogen
dioxide, and ozone at a cattle pasture in Rondonia, Brazil. Journal of Geophysical
Research-Atmospheres,107, doi:10.1029/2001JD000523.
Kruijt B, Elbers JA, von Randow C et al. (2004) The robustness of eddy correlation
fluxes for Amazon rain forest conditions. Ecological Applications,14, 101–113.
Lachniet MS (2009) Sea surface temperature control on the stable isotopic composition
of rainfall in Panama. Geophysical Research Letters,36, doi:10.1029/2008GL036625.
Lambers H, Chapin FS III, Pons TL (2008) Plant Physiological Ecology. Springer-Verlag
New York Inc., New York.
Loescher HW, Oberbauer SF, Gholz HL, Clark DB (2003) Environmental controls on
net ecosystem-level carbon exchange and productivity in a Central American
tropical wet forest. Global Change Biology,9, 396–412.
Luyssaert S, Inglima I, Jung M et al. (2007) CO
2
balance of boreal, temperate, and
tropical forests derived from a global database. Global Change Biology,13, 2509–2537.
Malhi Y, Baldocchi DD, Jarvis PG (1999) The carbon balance of tropical, temperate and
boreal forests. Plant Cell and Environment,22, 715–740.
Malhi Y, Davidson EA (2009) Biogeochemistry and ecology of terrestrial ecosystems in
Amazonia. In: Amazonia and global change (eds Keller M, Bustamante M, Gash J, Dias
PS), pp 293–298. American Geophysical Union, Washington, DC.
Malhi Y, Grace J (2000) Tropical forests and atmospheric carbon dioxide. Trends in
Ecology & Evolution,15, 332–337.
Malhi Y, Meir P, Brown S (2002) Forests, carbon and global climate. Philosophical
Transactions of the Royal Society of London Series A - Mathematical Physical and
Engineering Sciences,360, 1567–1591.
CARBON SEQUESTRATION OF PASTURE AND AFFORESTATION 2779
r2011 Blackwell Publishing Ltd, Global Change Biology,17, 2763–2780
Malhi Y, Nobre AD, Grace J, Kruijt B, Pereira MGP, Culf A, Scott S (1998) Carbon
dioxide transfer over a Central Amazonian rain forest. Journal of Geophysical
Research-Atmospheres,103, 31593–31612.
Mauder M, Foken T, Clement R et al. (2008) Quality control of CarboEurope flux data -
Part 2: Inter-comparison of eddy-covariance software. Biogeosciences,5, 451–462.
Miranda AC, Miranda HS, Lloyd J et al. (1996) Carbon dioxide fluxes over a cerrado
sensu stricto in central Brazil. In: Amazonian Deforestation and Climate (eds Gash JHC,
Nobre CA, Roberts JM, Victoria RL), pp 353–362. Wiley, Chichester.
Miranda AC, Miranda HS, Lloyd J et al. (1997) Fluxes of carbon, water and energy over
Brazilian cerrado: an analysis using eddy covariance and stable isotopes. Plant Cell
and Environment,20, 315–328.
Moffat AM (unpublished) A new methodology to interpret high resolution measurements of
net carbon fluxes between terrestrial ecosystems and the atmosphere. PhD thesis, Friedrich
Schiller University, Jena.
Moureaux C, Debacq A, Bodson B, Heinesch B, Aubinet M (2006) Annual net
ecosystem carbon exchange by a sugar beet crop. Agricultural and Forest Meteorology,
139, 25–39.
Potvin C, Manzilla L, Buchmann N et al. (2011) An ecosystem approach to biodiversity
effects: carbon pools in a tropical tree plantation. Forest Ecology and Management,
261, 1614–1624.
Priante-Filho N, Vourlitis GL, Hayashi MMS et al. (2004) Comparison of the mass and
energy exchange of a pasture and a mature transitional tropical forest of the
southern Amazon Basin during a seasonal transition. Global Change Biology,10,
863–876.
Saleska S, da Rocha H, Kruijt B, Nobre AD (2009) The changing rates and patterns of
deforestation and land use in Brazilian Amazonia. In: Amazonia and Global Change
(eds Keller M, Bustamante M, Gash J, Dias PS), pp 389–408. American Geophysical
Union, Washington, DC.
Saleska SR, Miller SD, Matross DM et al. (2003) Carbon in Amazon forests: unexpected
seasonal fluxes and disturbance-induced losses. Science,302, 1554–1557.
San Jose
´J, Montes R, Grace J, Nikonova N (2008) Land-use changes alter CO
2
flux
patterns of a tall-grass Andropogon field and a savanna–woodland continuum in
the Orinoco lowlands. Tree Physiology,28, 437–450.
Seitlinger T (2008) Einfluss der Artenzahl auf die Phosphorgehalte der oberirdischen Biomasse
und auf die Phosphorverfu
¨gbarkeit im Boden. University of Zurich, Zurich, 58 pp.
Serrano-Ortiz P, Roland M, Sanchez-Moral S, Janssens IA, Domingo F, Godderis Y,
Kowalski AS (2010) Hidden, abiotic CO
2
flows and gaseous reservoirs in the
terrestrial carbon cycle: review and perspectives. Agricultural and Forest Meteorology,
150, 321–329.
Silver WL, Kueppers LM, Lugo AE, Ostertag R, Matzek V (2004) Carbon sequestration
and plant community dynamics following reforestation of tropical pasture. Ecolo-
gical Applications,14, 1115–1127.
STRI (2010). Physical Monitoring Program. Smithsonian Tropical Research Institute
(STRI). Available at: http://striweb.si.edu/esp/physical_monitoring/index_phy_-
mon.htm (accessed 15 January 2010).
Tan ZH, Zhang YP, Yu GR, Sha LQ, TangJW, Deng XB, Song QH (2010) Carbon balance
of a primary tropical seasonal rain forest. Journal of Geophysical Research-Atmo-
spheres,115, doi:10.1029/2009JD012913.
von Randow C, Manzi AO, Kruijt B et al. (2004) Comparative measurements and
seasonal variations in energy and carbon exchange over forest and pasture in South
West Amazonia. Theoretical and Applied Climatology,78, 5–26.
Vourlitis GL, Nogueira JDS, Priante N et al. (2005) The sensitivity of Diel CO
2
and H
2
O
vapor exchange of a tropical transitional forest to seasonal variation in meteorology
and water availability. Earth Interactions,9, 1–23.
Vourlitis GL, Priante N, Hayashi MMS, Nogueira JD, Caseiro FT, Campelo JH (2001)
Seasonal variations in the net ecosystem CO
2
exchange of a mature Amazonian
transitional tropical forest (cerradao). Functional Ecology,15, 388–395.
Vourlitis GL, Priante N, Hayashi MMS, Nogueira JD, Raiter F, Hoegel W, Campelo JH
(2004) Effects of meteorological variations on the CO
2
exchange of a Brazilian
transitional tropical forest. Ecological Applications,14, 89–100.
Wassenaar T, Gerber P, Verburg PH, Rosales M, Ibrahim M, Steinfeld H (2007)
Projecting land use changes in the Neotropics: the geography of pasture expansion
into forest. Global Environmental Change,17, 86–104.
Waterloo MJ, Oliveira SM, Drucker DP et al. (2006) Export of organic carbon in run-off
from an Amazonian rainforest blackwater catchment. Hydrological Processes,20,
2581–2597.
Webb EK, Pearman GI, Leuning R (1980) Correction of flux measurements for density
effects due to heat and water-vapor transfer. Quarterly Journal of the Royal Meteor-
ological Society,106, 85–100.
Wilsey BJ, Parent G, Roulet NT, Moore TR, Potvin C (2002) Tropical pasture carbon
cycling: relationships between C source/sink strength, above-ground biomass and
grazing. Ecology Letters,5, 367–376.
Wolf S, Eugster W, Potvin C, Buchmann N (2011) Strong seasonal variations in net
ecosystem CO
2
exchange of a tropical pasture and afforestation in Panama.
Agricultural and Forest Meteorology,151, 1139–1151.
2780 S. WOLF et al.
r2011 Blackwell Publishing Ltd, Global Change Biology,17, 2763–2780
