Carbon dioxide and water vapor exchange in 
a warm temperate grassland
Authors: K. A. Novick, P. C. Stoy, G. G. Katul, D. S. 
Ellsworth, M. B. S. Siqueira, J. Juang, and R. Oren
The final publication is available at Springer via https://dx.doi.org/10.1007/s00442-003-1388-z.
Novick KA, Stoy PC, Katul GG, Ellsworth DE, Siqueira MBS, Juang J-Y, Oren R (2004) Carbon 
dioxide and water vapor exchange in a warm temperate grassland. Oecologia 138: 259-274. DOI: 
10.1007/s00442-003-1388-z.
Made available through Montana State University’s ScholarWorks 
scholarworks.montana.edu 
K. A. Novick . P. C. Stoy . G. G. Katul .
D. S. Ellsworth . M. B. S. Siqueira . J. Juang . R. Oren
Carbon dioxide and water vapor exchange in a warm
temperate grassland
Received: 24 March 2003 / Accepted: 18 August 2003 / Published online: 20 November 2003
Abstract Grasslands cover about 40% of the ice-free
global terrestrial surface, but their contribution to local and
regional water and carbon fluxes and sensitivity to climatic
perturbations such as drought remains uncertain. Here, we
assess the direction and magnitude of net ecosystem
carbon exchange (NEE) and its components, ecosystem
carbon assimilation (Ac) and ecosystem respiration (RE), in
a southeastern United States grassland ecosystem subject
to periodic drought and harvest using a combination of
eddy-covariance measurements and model calculations.
We modeled Ac and evapotranspiration (ET) using a big-
leaf canopy scheme in conjunction with ecophysiological
and radiative transfer principles, and applied the model to
assess the sensitivity of NEE and ET to soil moisture
dynamics and rapid excursions in leaf area index (LAI)
following grass harvesting. Model results closely match
eddy-covariance flux estimations on daily, and longer,
time steps. Both model calculations and eddy-covariance
estimates suggest that the grassland became a net source of
carbon to the atmosphere immediately following the
harvest, but a rapid recovery in LAI maintained a marginal
carbon sink during summer. However, when integrated
over the year, this grassland ecosystem was a net C source
(97 g C m−2 a−1) due to a minor imbalance between large
Ac (−1,202 g C m−2 a−1) and RE (1,299 g C m−2 a−1)
fluxes. Mild drought conditions during the measurement
period resulted in many instances of low soil moisture
(θ<0.2 m3m−3), which influenced Ac and thereby NEE by
decreasing stomatal conductance. For this experiment, low
θ had minor impact on RE. Thus, stomatal limitations to Ac
were the primary reason that this grassland was a net C
source. In the absence of soil moisture limitations, model
calculations suggest a net C sink of −65 g C m−2 a−1
assuming the LAI dynamics and physiological properties
are unaltered. These results, and the results of other
studies, suggest that perturbations to the hydrologic cycle
are key determinants of C cycling in grassland ecosystems.
Keywords Net ecosystem exchange . Ecosystem
modeling . Evapotranspiration . Eddy-covariance .
Grassland ecosystems
Introduction
Understanding the mechanisms that control ecosystem
carbon balance is a critical research priority given the
sensitivity of the carbon cycle to the biogeochemical and
hydrologic cycles of the terrestrial biosphere (Sarmiento
and Wofsy 1999; Houghton et al. 2001). Long-term CO2
and H2O flux monitoring initiatives such as FLUXNET
have arisen to understand how environmental variables
drive carbon cycling in ecosystems across the globe
(Baldocchi et al. 2001). However, most long-term CO2
flux research focuses on forests, with a consequent
shortage of CO2 flux data for grassland ecosystems
worldwide (Falge et al. 2001a, 2001b; Valentini et al.
2000; Baldocchi et al. 2001). In addition, many studies
and summaries of grassland C dynamics historically
concentrated on net primary productivity (NPP), not on
net ecosystem exchange of carbon (NEE) (e.g., Long et al.
1992; Scurlock et al. 2002).
Forested ecosystems in the southeastern United States
are characterized by long, warm, and mesic growing
seasons that favor high carbon assimilation rates (Clark et
al. 1999; Baldocchi and Wilson 2001; Oren et al. 2001;
Wilson and Baldocchi 2001). The grassland under study
here is typical of abandoned agricultural sites in the
southeastern United States, and is warmer and wetter than
most grassland ecosystems. Grassland ecosystems com-
prise approximately 40.5% of the Earth’s terrestrial land
area, excluding areas of permanent ice (White et al. 2000).
Large uncertainties remain in resolving whether grassland
ecosystems function as CO2 sources or sinks (Ojima et al.
1993; Parton et al. 1993; Baldocchi et al. 2001): annual
grassland NEE estimates based on eddy-covariance
measurements and Bowen Ratio Energy Balance techni-
ques vary from a net source of +400 g C m−2 a−1 to a net
sink of −800 g C m−2 a−1(Table 1). This uncertainty is
primarily attributable to the sensitivity of grasslands to
interannual variability in climate and associated biomass
dynamics (Knapp and Smith 2001; Meyers 2001;
Flanagan et al. 2002; Jackson et al. 2002; Scurlock et al.
2002), and incomplete understanding of the regulation of
ecosystem respiration (Raich and Potter 1995; Knapp et al.
1998; Wagai et al. 1998).
The objective of this investigation is to assess the
magnitude and direction of NEE and its components (Ac
and RE) in a southeastern United States warm-temperate
grassland ecosystem, and to assess NEE and ET responses
to episodic droughts and harvests. To this end, we use a
big-leaf process-based model that combines ecophysiolo-
gical and radiative transfer principles. The model is
calibrated with both leaf-level gas exchange and ecosys-
tem-level eddy-covariance measurements from an aban-
doned agricultural field at the Duke Forest C-H2O
Research Site near Durham, N. C. The model is then
used to conduct a sensitivity analysis to drought and leaf
area perturbations, and results are evaluated with respect to
grassland water and carbon balance studies to date.
Methods
Model
Net ecosystem carbon exchange (NEE) is defined as the difference
between ecosystem carbon assimilation (Ac) and ecosystem respi-
ration (RE).
NEE ¼ Ac þ RE (1)
We adopt the micrometeorological convention in which fluxes
from the biosphere to the atmosphere are positive. We employed the
biochemical photosynthesis model of Farquhar et al. (1980) as given
in Appendix A to compute leaf-level assimilation (An) for C3 grass
species, and coupled An to a big-leaf canopy scheme (Kim and
Verma 1991) to scale from leaf to canopy. Grassland foliage is
concentrated in short canopy heights, and the big-leaf approximation
can be used to scale photosynthesis from leaf to canopy to a first
approximation (Kim and Verma 1991) using:
Ac 
 An  fLAI  LAIð Þ (2)
where LAI is leaf area index (m2m−2) and fLAI is the fraction of
LAI that absorbs incident photosynthetically active radiation (PAR).
Here, fLAI is modeled after Campbell and Norman (1998):
fLAI ¼ exp Kb  ð Þ  LAIð Þ (3)
where Kb, the light extinction coefficient, is a function of the sun
zenith angle ψ. Kb is estimated from the Campbell and Norman
(1998) model with a leaf angle distribution parameter (x) of 0.7,
appropriate for erect grass leaves.
An is related to stomatal conductance to CO2 (gs) using a variant
on Fick’s law (e.g., Cowan 1977):
An ¼ gs  Ca CiCa  1

 
(4)
where Ci is the CO2 concentration in the intercellular spaces in the
leaf and Ca is the ambient atmospheric CO2 concentration
(~385 ppm). Numerous empirical and semi-empirical models for
gs and bulk canopy conductance (gc) have been proposed (e.g.,
Leuning 1995; Katul et al. 2000), with the simplest being a variant
on the Jarvis (1976) model, given by:
gc ¼ gref PARð Þ  f1 VPDð Þ  f2 
ð Þ
¼ fLAILAIð Þ  gs
(5)
where gref is the conductance at a reference vapor pressure deficit
(VPD) of 1 kPa for well-watered conditions (Oren et al. 1999),
f1(VPD) is a reduction function for vapor pressure deficit and f2(θ) is
a reduction function for soil moisture. The function gref is assumed
to vary with PAR and was determined from eddy-covariance
measured water vapor flux when θ>θR, where θ is the volumetric
root-zone soil moisture content and θR is the soil moisture content at
which gc is limited.
The function f1(VPD) is given by Oren et al. (1999):
f1 VPDð Þ ¼ 1 m ln VPDið Þ (6)
where i is PAR level (Appendix B) and the sensitivity parameter
m (~0.5–0.6) is determined in Appendix B using the boundary line
analysis proposed by Oren et al. (1999).
A standard soil moisture reduction function of the form:
f2ð
Þ ¼
1 ; 

R > 1
1 
R

R
  
; 

R  1
(
(7)
was chosen to account for drought effects on canopy conductance
(Campbell and Norman 1998). Following nonlinear optimization
using the Gauss-Newton algorithm (Dennis 1977), ν=0.6 and
θR=0.20. The θR is consistent with an earlier modeling study by Lai
and Katul (2000) in which actual and potential evapotranspiration
were shown to diverge at θR=0.19 for the same grassland.
To measure and model night-time respiration, we used a different
approach than the standard methodology of only accepting data if
the friction velocity (u*) exceeds a certain threshold, u*t. Commonly,
u*t is taken to be between 0.1 and 0.2 m s
−1 (Goulden et al. 1996;
Aubinet et al. 2000; Barford et al. 2001). Our respiration model is
based on night-time CO2 eddy-covariance measurements collected
for both u*>0.12 m s
−1 and for near-neutral atmospheric stability
conditions (|(z−d)/L|<0.1, see Appendix C). Here, z is instrument
height (3.0 m), L is the Obukhov length (Brutsaert 1982 p65), and h
and d are the mean canopy height and zero-plane displacement (~2/3
h), respectively. The addition of the atmospheric stability constraint
to the usual night-time friction velocity threshold ensures that the
flow is a fully developed turbulent flow that is near-neutral and not
“contaminated” by large-scale phenomena such as gravity waves or
meandering, and is critical for constraining the night-time flux
footprint (see Appendix C). In fact, from Appendix C, accepting u*t
as the only threshold with no atmospheric stability consideration can
T
ab
le
1
A
nn
ua
l
an
d
gr
ow
in
g
se
as
on
/s
um
m
er
ne
t
ec
os
ys
te
m
ca
rb
on
ex
ch
an
ge
(N
E
E
),
an
nu
al
ec
os
ys
te
m
ca
rb
on
as
si
m
ila
tio
n
(A
c)
,
an
nu
al
ec
os
ys
te
m
re
sp
ir
at
io
n
(R
E
),
an
d
m
ax
im
um
da
ily
N
E
E
an
d
A
c
re
po
rt
ed
fr
om
va
ri
ou
s
gr
as
sl
an
d
an
d
sa
ge
br
us
h
ec
os
ys
te
m
s.
T
he
si
gn
fo
llo
w
s
th
e
m
ic
ro
m
et
eo
ro
lo
gi
ca
l
co
nv
en
tio
n
th
at
po
si
tiv
e
fl
ux
es
ar
e
fr
om
th
e
bi
os
ph
er
e
to
th
e
at
m
os
ph
er
e.
A
ll
m
ea
su
re
m
en
ts
ar
e
m
ad
e
by
ed
dy
-c
ov
ar
ia
nc
e
or
B
ow
en
ra
tio
en
er
gy
ba
la
nc
e
(B
R
E
B
)
a .
M
ea
su
re
m
en
ts
ob
ta
in
ed
us
in
g
B
R
E
B
ar
e
ita
lic
iz
ed
.
S
ai
gu
sa
et
al
.
(1
99
8)
em
pl
oy
an
ae
ro
dy
na
m
ic
m
et
ho
d
si
m
ila
r
to
B
R
E
B
S
ite
E
co
sy
st
em
ty
pe
Y
ea
r
N
E
E
(g
C
m
−2
a−
1
)
G
ro
w
in
g
se
as
on
(s
um
m
er
)
N
E
E
(g
C
m
−2
se
as
on
−1
)
A
c
(g
C
m
−2
a−
1
)
R
E
(g
C
m
−2
a−
1
)
M
ax
N
E
E
(g
C
m
−2
da
y−
1
)
M
ax
A
c
(g
C
m
−2
da
y−
1
)
R
ef
er
en
ce
s
M
at
ad
or
,
SK
,
C
an
ad
a
M
ix
ed
-g
ra
ss
pr
ai
ri
e
19
70
−
2.
7
−
4.
9
R
ip
le
y
an
d
Sa
ug
ie
r
(1
97
4,
19
78
)
19
71
−
4.
1
R
ed
m
an
n
(1
97
8)
L
et
hb
ri
dg
e,
A
L
,
C
an
ad
a
M
oi
st
m
ix
ed
gr
as
sl
an
d
19
98
−1
09
−3
73
26
4
−5
.0
−
9
F
la
na
ga
n
et
al
.
(2
00
2)
19
99
−2
1
−2
87
26
7
−3
.2
−
5.
5
20
00
+
18
−2
72
29
0
−2
.4
−
4.
5
M
an
da
n,
N
.D
.,
U
SA
N
or
th
er
n
m
ix
ed
gr
as
s
pr
ai
ri
e
19
95
−
11
8
−
14
7
b
—
F
ra
nk
et
al
.
(2
00
1)
19
96
−
85
−
11
3
b
[−
10
8]
−
4.
9
F
ra
nk
an
d
D
ug
as
(2
00
1)
19
97
−
55
−
85
b
[−
78
]
−
4.
6
19
98
—
−4
9
b
−
3.
7
19
99
−
34
c
−
12
9
b
−
4.
9
D
ub
oi
s,
Id
ah
o,
U
SA
Sa
ge
br
us
h-
st
ep
pe
19
96
−
10
0
(a
ve
ra
ge
)
−
77
d
G
ilm
an
ov
et
al
.
(2
00
3a
)
19
97
−
11
2
d
19
98
−
20
3d
19
99
−
30
1d
M
an
ha
tta
n,
K
an
.,
U
SA
Ta
llg
ra
ss
pr
ai
ri
e
19
96
−
2.
2
−
9.
5
H
am
an
d
K
na
pp
(1
99
8)
Ja
sp
er
R
id
ge
,
C
al
if
.,
U
S
A
S
er
pe
nt
in
e
gr
as
sl
an
d
7/
90
–7
/9
1
−1
33
−1
.9
V
al
en
tin
i
et
al
.
(1
99
5)
S
hi
dl
er
,
O
kl
a.
,
U
S
A
Ta
llg
ra
ss
pr
ai
ri
e
19
87
−2
05
e
−7
.6
K
im
an
d
V
er
m
a
(1
99
1)
19
89
−3
18
to
−3
83
f
−6
8e
K
im
et
al
.
(1
99
2)
P
on
ca
,
O
kl
a.
,
U
S
A
Ta
llg
ra
ss
P
ra
ir
ie
19
97
−2
74
g
−4
46
−8
12
h
53
8
i
−9
.5
S
uy
ke
r
an
d
V
er
m
a
(2
00
1)
19
98
−4
6g
−2
04
−5
48
h
50
2
i
−7
.5
S
uy
ke
r
et
al
.
(2
00
3)
19
99
−1
24
g
−3
00
−6
34
h
51
0
i
−8
.5
W
oo
dw
ar
d,
O
kl
a.
,
U
SA
M
ix
ed
gr
as
s
pr
ai
ri
e
19
95
−
97
[−
10
1]
−
11
8
−
4.
9
Si
m
s
an
d
B
ra
df
or
d
(2
00
1)
j
[F
ra
nk
et
al
.
(2
00
1)
]
19
96
+
46
[+
19
]
−
13
−
3.
3
19
97
−
15
9
[−
17
9]
−
19
9
−
5.
5
W
oo
dw
ar
d,
O
kl
a.
,
U
SA
Sa
ge
br
us
h/
m
ix
ed
gr
as
sl
an
d
19
95
−
55
−
74
−
2.
7
Si
m
s
an
d
B
ra
df
or
d
(2
00
1)
19
96
+
9
+
31
−
2.
2
19
97
+
27
+
26
−
4.
6
C
hi
ck
as
ha
,
O
kl
a.
,
U
S
A
M
ix
ed
-g
ra
ss
ra
ng
el
an
d
19
95
−1
96
k
−4
.5
M
ey
er
s
(2
00
1)
k
[F
al
ge
et
al
.
(2
00
1a
)]
19
96
−4
1
k
−4
19
97
+
41
to
+
15
0
f
−1
88
k
−5
19
98
+
41
9
to
+
52
1
+
15
5
k
−1
S
ite
E
co
sy
st
em
ty
pe
Y
ea
r
N
E
E
(g
C
m
−2
a−
1
)
G
ro
w
in
g
se
as
on
(s
um
m
er
)
N
E
E
(g
C
m
−2
se
as
on
−1
)
A
c
(g
C
m
−2
a−
1
)
R
E
(g
C
m
−2
a−
1
)
M
ax
N
E
E
(g
C
m
−2
da
y−
1
)
M
ax
A
c
(g
C
m
−2
da
y−
1
)
R
ef
er
en
ce
s
Te
m
pl
e,
Te
x.
,
U
SA
Ta
llg
ra
ss
pr
ai
ri
e
19
93
−
50
l
−
5
D
ug
as
et
al
.
(1
99
9)
19
94
−
80
−
6.
5
F
ra
nk
et
al
.
(2
00
1)
19
95
−
77
2
−
83
1
19
96
−
22
3
−
28
6
19
97
−
38
6
−
47
6
Te
m
pl
e,
Te
x.
,
U
SA
C
yn
od
on
da
ct
yl
on
19
93
+
10
0
−
3
D
ug
as
et
al
.
(1
99
9)
19
94
−
80
0
−
14
W
al
nu
t
G
ul
ch
W
at
er
sh
ed
,
A
ri
z.
,
U
SA
(L
uc
ky
H
ill
s)
M
ix
ed
sh
ru
b
19
97
+
13
0
−
0.
8
E
m
m
er
ic
h
(2
00
3)
19
98
+
14
0
−
2.
7
19
99
+
15
5
−
2.
7
20
00
+
15
0
−
2.
7
W
al
nu
t
G
ul
ch
W
at
er
sh
ed
,
A
ri
z.
,
U
SA
(K
en
da
ll)
M
ix
ed
gr
as
s
19
97
+
13
0
−
3.
8
E
m
m
er
ic
h
(2
00
3)
19
98
+
21
0
−
4.
6
19
99
+
11
0
−
5.
5
20
00
+
60
−
3.
8
Sh
or
ta
nd
y,
K
az
ak
hs
ta
n
Sh
or
tg
ra
ss
st
ep
pe
19
98
–2
00
1
−
35
7
to
−
53
4
G
ilm
an
ov
et
al
.
(2
00
3b
)
K
as
be
gi
ra
ng
e,
G
eo
rg
ia
P
as
tu
re
19
96
−
2.
1
−
2.
5
Ta
pp
ei
ne
r
an
d
C
er
nu
sc
a
(1
99
6)
W
et
m
ea
do
w
−
5.
3
−
5.
9
H
ay
−
5.
9
−
7.
0
Ta
ll
he
rb
−
8.
5
−
9.
1
Ts
uk
ub
a,
Ja
pa
n
G
ra
ss
fie
ld
19
93
−
12
.5
Sa
ig
us
a
et
al
.
(1
99
8)
19
94
−
14
.2
R
od
on
ia
,
B
ra
zi
l
C
4
pa
st
ur
e
5/
93
−1
.9
m
−
8.
0
m
G
ra
ce
et
al
.
(1
99
8)
M
ac
ke
nz
ie
B
as
in
,
N
Z
T
us
so
ck
gr
as
sl
an
d
19
98
–1
99
9
−1
.9
−
3.
9
H
un
t
et
al
.
(2
00
2)
D
uk
e
F
or
es
t,
N
.C
.,
U
S
A
M
ix
ed
-g
ra
ss
fi
el
d
4/
00
–4
/0
1
+
97
+
23
n
(−
2
k
)
−1
,2
02
1,
29
9
−2
.8
−
7.
6
P
re
se
nt
st
ud
y
T
ab
le
1
(c
on
tin
ue
d)
S
ite
E
co
sy
st
em
ty
pe
Y
ea
r
N
E
E
(g
C
m
−2
a−
1
)
G
ro
w
in
g
se
as
on
(s
um
m
er
)
N
E
E
(g
C
m
−2
se
as
on
−1
)
A
c
(g
C
m
−2
a−
1
)
R
E
(g
C
m
−2
a−
1
)
M
ax
N
E
E
(g
C
m
−2
da
y−
1
)
M
ax
A
c
(g
C
m
−2
da
y−
1
)
R
ef
er
en
ce
s
D
uk
e
F
or
es
t,
N
.C
.,
U
SA
M
ix
ed
-g
ra
ss
fie
ld
4/
00
–4
/0
1
−
65
−
17
9
n
(−
94
k )
−
1,
35
6
1,
29
1
M
od
el
ed
an
nu
al
C
flu
x
w
ith
ou
t
θ
lim
ita
tio
ns
a B
R
E
B
sy
st
em
s
ca
n
be
em
pl
oy
ed
at
si
te
s
in
w
hi
ch
gr
ad
ie
nt
m
ea
su
re
m
en
ts
ar
e
co
nd
uc
te
d
in
th
e
at
m
os
ph
er
ic
su
rf
ac
e
la
ye
r
(e
.g
.,
D
ug
as
et
al
.1
99
9)
.T
he
te
ch
ni
qu
e
ha
s
se
ve
ra
ll
im
ita
tio
ns
:(
1)
er
ro
rs
du
e
to
sp
at
ia
lly
va
ri
ab
le
so
il
he
at
fl
ux
,
(2
)
th
e
as
su
m
pt
io
n
th
at
ed
dy
-d
if
fu
si
vi
tie
s
fo
r
he
at
,
w
at
er
va
po
r,
an
d
C
O
2
ar
e
eq
ua
l,
an
d
(3
)
un
re
lia
bi
lit
y
w
he
n
L
E
is
sm
al
l.
D
es
pi
te
th
es
e
lim
ita
tio
ns
th
e
m
et
ho
d
st
ill
of
fe
rs
a
va
lu
ab
le
es
tim
at
e
of
lo
ng
-t
er
m
fl
ux
es
w
he
n
da
ta
is
pr
op
er
ly
po
st
-p
ro
ce
ss
ed
(B
ru
st
ae
rt
19
82
)
b
A
pr
il
24
to
O
ct
ob
er
26
c −
57
g
C
m
−2
a−
1
w
he
n
so
il
fl
ux
,
no
t
B
R
E
B
,
is
us
ed
to
m
ea
su
re
do
rm
an
t-
se
as
on
fl
ux
es
d
L
en
gt
h
of
gr
ow
in
g
se
as
on
va
ri
es
be
tw
ee
n
10
5
an
d
14
0
da
ys
e M
ay
–O
ct
ob
er
19
87
,
Ju
ne
–A
ug
us
t
19
89
f V
al
ue
de
pe
nd
s
on
ga
pf
ill
in
g
an
d
u*
co
rr
ec
tio
n,
se
e
F
al
ge
et
al
.
(2
00
1a
)
g
N
E
E
~0
g
C
m
−2
a−
1
w
he
n
pr
es
cr
ib
ed
pr
ai
ri
e
bu
rn
is
in
cl
ud
ed
h
D
ay
tim
e
N
E
E
i N
ig
ht
-t
im
e
N
E
E
j B
ra
ck
et
ed
va
lu
es
ar
e
fr
om
S
im
s
an
d
B
ra
df
or
d
(2
00
1)
.
D
if
fe
re
nt
re
sp
ir
at
io
n
va
lu
es
ar
e
us
ed
fo
r
es
tim
at
es
of
an
nu
al
fl
ux
es
k
30
M
ay
to
28
A
ug
us
t
l B
ur
ne
d
pr
io
r
to
gr
ow
in
g
se
as
on
m
N
et
da
ily
fl
ux
fo
r
an
11
-d
ay
pe
ri
od
in
M
ay
19
93
n
11
A
pr
il
to
6
D
ec
em
be
r
T
ab
le
1
(c
on
tin
ue
d)
give flux source areas in excess of 5 km, an order of magnitude
larger than the dimensions of our field.
Ecosystem respiration was modeled as a function of temperature
with the widely used van’t Hoff (1898) equation:
RE ¼ R10ð Þ Q10ð Þ T10ð Þ=10 (8)
where R10 is the reference respiration rate at 10°C and Q10 is the
ecosystem respiration sensitivity to temperature. From regression
analysis on night-time fluxes of the entire data set, we computed an
effective Q10 of 1.55 and R10 of 2.54 μmol CO2 m
−2s−1. All model
parameters are summarized in Table 2.
Seasonal variations in respiration parameters may be important
components of the error in annual C budget estimates generated
from eddy-covariance measurements that Goulden et al. (1996)
termed “sampling uncertainty”. There is evidence that ecosystem
Q10 and R10 vary throughout the course of a year, but also that single
annual respiration parameters may be sufficient to calculate annual
RE estimates (Janssens and Pilegaard 2003). To test this observation
using eddy-covariance measurements, we estimate annual fluxes
using both annually averaged and seasonally generated Q10 and R10
parameters (Tables 2, 3). To quantify respiration parameters that
vary throughout the course of the year, Q10 and R10 are calculated
for summer (May–August) and winter (November–February) per
unit leaf area, and the actual Q10 and R10 are generated using a
simple interpolation:
Q10 ¼ Q10;w  Q10;sð Þ
 LAI LAImin
LAImin  LAImax

 
þ Q10;w
(9)
where Q10,w and Q10,s are Q10 values calculated for winter and
summer, respectively, and LAImax and LAImin are maximum and
minimum measured LAI. The interpolation for R10 follows the same
model.
RE was unrelated to soil moisture for the mild drought conditions
encountered during the measurement period. We tested whether the
residuals from equation 8 (i.e., the difference between measured and
modeled RE fluxes) depend on soil moisture and found weak
correlation (r2=0.09). Other studies have also shown an insignificant
relationship between respiration and soil moisture (Fang and
Moncreiff 2001), and this observation appears to hold for the mild
drought encountered here though a stronger dependence on soil
moisture may result from more severe droughts.
During the study period, the grass was cut on 29 June 2001,
necessitating a dynamic LAI growth model immediately following
this perturbation. We chose a mathematical model whose canonical
form resembles a logistic growth equation. Such a model is
approximated by a sequence of cubic splines to the four discrete LAI
measurements (Fig. 1). The cubic spline technique to fit discrete
data is described in Press et al. (1992 p108).
To calibrate our conductance model and to explore water fluxes
from the canopy, we modeled latent heat exchange (LE) as:
LE ¼ Lv  1:6gc  VPD (10)
where Lv is the latent heat of vaporization of water, and the 1.6
factor is needed to correct canopy conductance for differences in
binary diffusivity between CO2 and H2O.
In the analysis, Ac is computed as the difference between NEE
and modeled RE for daytime runs. RE is the eddy-covariance-
measured NEE value at night when u*>0.12 m s
−1 and |(z−d)/L|<0.1,
and is modeled RE at all other times. Daily, seasonal, and annual
sums of fluxes are called “estimates” because they depend on both
directly measured fluxes and model results (e.g., equation 8) that fill
gaps in the data record.
Table 2 Model parameters were measured via gas exchange or estimated from eddy-covariance through nonlinear optimization or
boundary-line analysis (BLA). Parameters not directly measured are taken from the cited literature
Parameter Description Value Units Source
Photosynthesis
Vcmax Maximum Rubisco carboxylation capacity 81.3 μmol CO2 m
−2s−1 Gas exhange measurements taken during the
growing season
α Leaf absorptivity for PAR 0.83 mol mol−1 Campbell and Norman (1998)
em Maximum quantum efficiency 0.08 — Campbell and Norman (1998)
Ci/Ca Mean ratio of intercellular to ambient CO2 0.75 — Gas exhange measurements taken during the
growing season
[O2] Oxygen mole fraction (mmol mol
−1) 210 mmol mol−1 Campbell and Norman (1998)
τ Ratio describing CO2/O2 partitioning
by Rubisco
1.3 — Campbell and Norman (1998)
Kc25 Michaelis constant for CO2 fixation 300 μmol mol
−1 Campbell and Norman (1998)
γ Kc parameter 0.074 —
KO2, 25 Michaelis constant for O2 inhibition 300 mmol mol
−1 Campbell and Norman (1998)
γ KO2 parameter 0.018 —
x Leaf angle distribution parameter 0.7 — Campbell and Norman (1998)
Canopy conductance
ν Soil parameter 0.6 — Nonlinear optimization
θR Moisture content below which gc is reduced 0.2 m
3m−3 Nonlinear optimization and Lai and Katul (2000)
m Sensitivity of canopy conductance to VPD 0.6 — Eddy-covariance measured LE using BLA.
Oren et al. (1999)
Respiration
R10 Scale parameter 2.54 μmol m
−2s−1 Eddy-covariance measurements for
u*>0.12 m s
−1and |(z−d)/L|<0.1
Q10 Slope parameter 1.55 — Same as R10
Measurements
The experimental site is a grass-covered field in the Blackwood
Division of the Duke Forest in Orange County, near Durham, North
Carolina (35.971°N, 79.09°W, elevation 163 m). The long-term
mean annual temperature and precipitation are 15.5°C and
1,145 mm, respectively. The field is approximately 480×305 m,
dominated by the C3 grass Festuca arundinaria Shreb., and
surrounded by loblolly pine (Pinus taeda L.) forest. The vegetation
includes minor components of C3 herbs and the C4 grass
Schizachyrium scoparium (Michx.) Nash, not considered here.
The site was burned in 1979 and is mowed annually during the
summer for hay according to local practices. For this investigation,
we consider data collected between 11 April 2001 and 11 April
2002.
Scalar fluxes were measured using an eddy-covariance system
comprised of a triaxial sonic anemometer (CSAT3, Campbell
Scientific, Logan, Utah) and an open-path gas analyzer (LI-7500,
Li-Cor, Lincoln, Neb.), positioned 3.0 m above the canopy. The gas
analyzer was tilted 35° from the vertical to avoid direct sunlight
contamination and to minimize water accumulation on the absorbing
lens surface. The LI-7500 was separated from the CSAT3 by 10 cm,
a distance comparable to the sonic path averaging length. The time
series of all three velocity components, temperature, and scalar
concentrations were sampled using a 23X data logger (Campbell
Scientific, Logan, Utah) at 10 Hz. All covariances were then
computed over a 30-minute period using the procedures described in
Katul et al. (1997). The Webb-Pearman-Leuning correction (Webb
et al. 1980) was subsequently applied to the computed scalar
covariances.
The tower was located in the middle of the field with
approximately 250 m fetch to the southwest, the predominant
direction of flow during summer. The peak of the source weight
function (xp) describes the peak of the maximum source area that
contributes to fluxes, and was estimated using the footprint model in
Hsieh et al. (2000) to be smaller than 150 m (at z=3 m) for most
stability runs, except for stable conditions (Appendix C). The source
weight function describes the relative scalar flux contribution to a
measurement location for various scalar source areas upwind. The
term “footprint” is the distance at which 90% of the scalar flux
contributes to the measurement location as determined from the
integrated source weight function. RE data collected during stable
atmospheric conditions at night or when xp exceeded the size of the
field were discarded and replaced with the output of the Q10
respiration function.
PAR, air temperature (Ta), relative air humidity (RH), net
radiation (Rnet) and θ were sampled every second and averaged
every half-hour. Rnet was measured with a Fritschen-type net
radiometer (Q7, REBS, Seattle, WA) and incident PAR with a
quantum sensor (LI-190SA Li-Cor, Lincoln, NE). Ta and RH were
measured with a HMP35C temperature/RH probe (Campbell
Scientific, Logan, Utah). θ was measured using ThetaProbe soil
moisture sensors Type ML1 (Delta-T Devices, Cambridge, UK)
positioned at 10 cm and 25 cm depths at six locations north and
south of the eddy-covariance tower, and at a depth of 10 cm at
locations east and west of the tower.
Gas exchange measurements were used to estimate the apparent
maximum rubisco carboxylation capacity (Vcmax) and were
performed in May, June and August 2001 using an open-flow LI-
6400 portable photosynthesis system (Li-Cor, Lincoln, Neb.). For
leaf-level photosynthesis, Vcmax was directly fitted to in situ
responses of An to CO2 supply under controlled conditions
(following the approach of Medlyn et al. 2002). The response of
An to intercellular CO2 concentration (Ci) was measured in the field
within 50 m of the tower in mid-morning on sunny days by
controlling chamber conditions to light saturation (1,800 μmol m−2
s−1 quantum flux density) and 29–32°C leaf temperature with
VPD=1.8 kPa for 9–10 different Ci steps. Vcmax was then fitted to
the initial slope comprising at least five Ci levels <250 μmol mol
−1,
with r2 exceeding 98% (Table 2).
We calculated LAI from PAR transmission data collected along
three 50 m transects in a 120° swath to the south of the eddy-
covariance tower. The PAR transmission data were measured with
an 80-sensor series of quantum sensors (AccuPAR model PAR-80
Ceptometer, Decagon Instruments, Pullman, Wash.) and used to
calculate gap fractions, which were inverted to provide LAI
estimates after Norman and Campbell (1989). Time series of
environmental drivers for the measurement period 11 April 2001 to
11 April 2002 are presented in Fig. 2.
The eddy-covariance methodology is prone to missing data points
that can occur due to precipitation, extreme weather events, sensor
malfunction, or power outage. Raw flux data coverage for this site
over the period of examination (11 April 2001 to 11 April 2002) was
92.9% (16,281 of 17,520 potential data points). After filtering out
night-time CO2 flux data using the criteria in Appendix C, 45.1% of
potential CO2 flux data points remain. A variety of methods exist to
‘gapfill’ missing CO2 and LE data, as summarized in Falge et al.
(2001a, 2001b). NEE data gaps when PAR exceeded the light
compensation point (213 μmol photons m−2 s−1) were filled by
fitting a nonlinear regression of measured CO2 flux about PAR and
replacing missing data points with the results of the regression
(Fig. 3a). NEE data gaps for night-time and low PAR periods were
filled using the Q10 respiration equation. Gaps in the LE record were
Table 3 Modeled and estimated annual carbon and water budgets in a southeastern United States grassland. ‘Modeled—no drought’ refers
to the modeling analysis where soil moisture was parameterized to have no influence on stomatal conductance
Variable Estimated: single annual
respiration paramters
Estimated: seasonal
respiration paramters
Modeled Modeled—
no drought
Modeled—
drought, no harvest
Modeled—no drought,
no harvest
ET (mm a−1) 568 568 547 738 570 767
Ac (g C m
−2a−1) −1,202 −1,304 −1,207 −1,356 −1,230 −1,381
RE (g C m
−2 a−1) 1,299 1,433 1,291 1,291 1,291 1,291
NEE (g C m−2 a−1) 97 129 84 −65 61 −90
Fig. 1 Modeled leaf area index (LAI) from 11 April 2001 to 11
April 2002. LAI measurements are shown as large circles
filled by fitting a linear regression of LE about Rnet (Fig. 3b), and
replacing missing data points with the results of the regression
(Brutsaert and Sugita 1992).
Temporal data coverage for PAR was 87.5%, for Rnet 87.0%, for
RH and Ta 80.5% and for θ 67.6%. To gapfill missing environmental
data, a linear relationship was derived between measured data points
at the grass site and adjacent pine and hardwood eddy-covariance
tower sites under identical climatic and edaphic conditions. Missing
data were gapfilled with the results of the regression.
Results and discussion
To address the study objectives, we first discuss estimated
and modeled ET and NEE and compare these results with
other studies. We then proceed to assess the impact of
harvesting on ET and NEE and summarize the contribu-
tions of this study to current understanding of grassland
water and carbon cycling.
Evapotranspiration
Maximum estimated daily evapotranspiration (ET, 4.1 mm
day−1) is within the range of values estimated for other
grasslands based on eddy-covariance, which range from
~3 mm day−1to 5.5 mm day−1 (Ripley and Saugier 1978;
Meyers 2001; Dugas et al. 1999; Hunt et al. 2002; Wever
et al. 2002). Values estimated by other methods range from
4.2 mm day−1to 6.2 mm day−1 as summarized by Kelliher
Fig. 2 Time series of maxi-
mum daily measurements for
key ecophysiological drivers
from 11 April 2001 to 11 April
2002: a photosynthetically ac-
tive radiation (PAR), b Rnet, c
maximum and minimum Ta, d
vapor pressure deficit, and e
volumetric soil moisture content
(θ). Measured data are black
points; gray points indicate
gapfilled data. The point at
which soil moisture limits con-
ductance (θ=0.2 m3m−3) is in-
dicated by a dashed line
et al. (1993). We repeated the model analysis assuming
that soil moisture does not limit conductance for the entire
record, yet retaining the same Ta and VPD for a ‘drought-
free’ scenario. In the absence of soil moisture limitations,
modeled maximum daily ET is 5.3 mm, near the
maximum of eddy-covariance estimated values for grasses
(5.5 mm in Dugas et al. 1999).
Total summer evapotranspiration [ETs, for day of year
(DOY) 150–240 based on Meyers (2001)] was 239.2 mm.
Our value is close to ETs over an Oklahoma rangeland
during three non-drought seasons (1995–1997), which
averaged ~253 mm (Meyers 2001); ETs in southeastern
United States grassland during a dry year resembled ETs
during average precipitation years over the United States
Great Plains.
Annual precipitation (Pa) for the measurement period 11
April 2001 to 11 April 2002 was 821.5 mm. This is 72%
of the long-term average of 1,145 mm, indicating mild
drought conditions. Annual evapotranspiration (ETa) for
the study period was 568 mm, slightly more than modeled
ETa (547 mm; Fig. 4; Table 3). ETa represented 69.1% of
Pa, which is intermediate between published ETa/Pa for
grasslands that range from <50% to over 100% (Meyers
2001; Paz et al. 1996; Bellot et al. 1999; Nouvellon et al.
2000; Everson 2001; Wever et al. 2002). ETa/Pa here is
consistent with a natural grassland catchment in Natal
Drakensberg, South Africa during two low precipitation
years (~70%; Everson 2001), and two mixed-grassland
sites in northwestern Spain during a low precipitation year
(~69%; Paz et al. 1996), suggesting similarities in the
water balance of warm-temperate/Mediterranean grassland
ecosystems during dry years.
Soil moisture limited gc (meaning θ<θR, θR=0.2 m
3m−3)
52% of the time for the entire year and 66% of the time
during non-winter periods (Figs. 5a, b). We consider non-
winter periods to be DOY 101–340, when Ac is not
seasonally suppressed by dead vegetation. Re-parameter-
izing the model to simulate the ‘drought-free’ scenario
results in modeled ETa of 738 mm or ETa/Pa of 89.9%
(Table 3). The drought-free modeled ETa/Pa is consistent
Fig. 3 a The relationship be-
tween net ecosystem carbon
exchange (NEE) and PAR for
the one-year measurement
record. The solid nonlinear re-
gression line is used to estimate
NEE for PAR above the C
compensation point
(PAR=213 μmol m−2 s−1) for
gaps in the measurement record.
b Latent heat exchange (LE)
plotted against net radiation
(Rnet). The solid linear regres-
sion line is used to estimate LE
for gaps in the measurement
record
Fig. 4 Estimated and modeled cumulative annual evapotranspira-
tion
with eddy-covariance and Bowen Ratio Energy Balance
studies in grasslands during years with normal to above-
average precipitation, although infiltration and soil storage
and lower VPD during very wet years decrease this ratio
(Meyers 2001; Nouvellon et al. 2000; Wever et al. 2002).
Eddy-covariance studies are typically performed on flat
terrain. Thus, surface runoff is a small component of the
water budget and is not likely to affect the water balance.
Net ecosystem exchange
Maximum daily Ac (−7.6 g C m−2day−1) is at the upper
end of the range of reported values for other grasslands
(−2.5 g C m−2day−1 to −9.1 g C m−2day−1; Table 1).
Maximum daily NEE was low because of high daily RE
but comparable to other studies with drought impacts
(Table 1). High maximum daily Ac and low maximum
daily NEE suggest that the magnitude of RE played a
major role in determining the magnitude of net fluxes
during both drought and non-drought periods.
Previous studies have demonstrated that RE decreases
with θ (e.g., Reichstein et al. 2002). However, in our study,
the magnitude of RE was primarily dependent on temper-
ature and was insensitive to θ for the measurement period,
consistent with Fang and Moncreiff (2001). Perhaps a
more prolonged drought, especially when coupled with
high temperatures, would generate a stronger dependence
of RE on θ.
Positive NEE throughout non-winter periods indicated a
net return of CO2 to the atmosphere during this time
(Table 1). However, model results show that, in a drought-
free year with no soil moisture limitation on conductance,
growing season NEE in our grassland is comparable to
other studies (Table 1). NEE is near zero during summer,
in contrast to a south-central United States rangeland
which experienced negative net C fluxes except in the case
of severe drought (Meyers 2001). Model results suggest
that even in the absence of drought, summer C fluxes in
our grassland only approach 50% of summer values
observed by Meyers (2001) (Table 1), indicating that our
grassland is unable to sequester appreciable C during the
summer. In the absence of drought, growing season NEE
values approach those of other grasslands not because of a
large summertime sink, but because of a longer growing
season in the southeastern United States.
Annual NEE (+97 g C m−2 a−1) was an order of
magnitude smaller than estimated annual Ac (−1,202 g C
m−2 a−1) and RE (+1,299 g C m
−2 a−1; Tables 1, 3; Fig. 6).
Annual Ac and RE were 1.5 times larger and over 2.4 times
larger, respectively, than other grasslands (Table 1). These
large fluxes reflect the longer growing season in the
southeast United States that is warmer and wetter than in
the grassland biome. These annual flux estimates were
generated using static respiration parameters for the entire
year (Tables 2, 3). The “effective” Q10 for the entire year,
1.55, is at the low end of previously reported values for
ecosystems (Raich and Schlesinger 1992; Kirschbaum
2000). Q10 estimates are highly dependent on the reference
temperature used (Lloyd and Taylor 1994; Tjoelker et al.
2001), whether air or soil temperatures are employed in
the calculations, and how the data is pooled or ensemble-
averaged.
Flux estimates using respiration parameters that are
scaled by leaf area resulted in Ac=−1,304 g C m−2 a−1 (i.e.,
102 g C m−2 a−1 more negative) and increased RE to
1,433 g C m−2 a−1 (i.e., 134 g C m−2 a−1 more positive) for
a modified NEE estimate of 129 g C m−2 a−1 (Table 3).
Despite the large variation in LAI, the annual exchange
rates appear to be robust to seasonally dynamic leaf-area
scaled respiration parameters. A chamber-based study of
belowground respiration by Janssens and Pilegaard (2003)
suggested that single annual respiration parameters are
adequate for estimating annual belowground respiration.
In this eddy-covariance-based study, the magnitude of
annual NEE was impacted little by varying respiration
parameters throughout the year, although component
fluxes (Ac and RE) increased in magnitude by over 100 g
C m−2 a−1.
Modeled Ac, RE, and NEE with and without soil
moisture limitations on gc are contrasted in Table 3.
Estimated and modeled Ac, RE, and NEE are presented in
Fig. 6 as cumulative carbon exchange throughout the
measurement period. Root mean square error between
modeled and estimated NEE is 0.057 g C m−2 a−1.
Estimated annual NEE represents a non-trivial C flux to
the atmosphere compared to many other grassland eddy-
covariance studies (Table 1; but see Falge et al. 2001a).
Model calculations suggest that this grassland ecosystem
switches from a net annual C source to a net annual C sink
depending on soil moisture conditions (Table 3). Ac
increases when soil moisture limitations are removed due
to gc enhancement, while RE is insensitive to θ under mild
drought and does not change in the ‘drought-free’ model
analysis since no soil moisture limitation was employed.
Fig. 5 a Probability density function (pdf) of measured root-zone
soil moisture content (θ) from 11 April 2001 to 10 April 2002. θ<θR
for 52% of the year. The dotted line indicates the point at which θ
suppresses canopy conductance. b pdf of soil moisture content (θ)
measurements for non-winter periods (11 April 2001 to 6 December
2002, defined as the period when Ac is not suppressed). θ<θR for
66% of the season
The sign shift observed in drought-free modeled NEE is
consistent with results from a northern Great Plains mixed
grassland (Flanagan et al. 2002), a southern Great Plains
rangeland (Frank et al. 2001), and a southern Great Plains
tallgrass prairie (Suyker et al. 2003) (Table 1). In all three
studies, the sign shift or increase in NEE sink strength was
attributed to an increase in Ac and not a reduction in annual
RE, consistent with our model calculations (Table 3).
These studies, although limited in number, suggest that
NEE variability is driven by drought impacts on assim-
ilation and subsequent growth (Knapp and Smith 2001).
This contrasts with observations made across a range of
European forests, which found that RE, not Ac, was the
primary contributor of variation in the carbon balance
across sites (Valentini et al. 2000).
Positive annual NEE values commonly occur as a result
of drought (Table 1), highlighting the strong coupling of
the carbon and water cycles in grasslands and further
suggesting that perturbations in the hydrologic cycle
disrupts the C balance in grasslands (e.g., Knapp and
Smith 2001). However, even in the absence of soil
moisture limitations, NEE modeled for our site is low
compared to the range of non-drought annual grassland
NEE (Tables 1, 3), further suggesting that this southeastern
United States grassland is unable to sequester large
amounts of carbon under the management protocol (i.e.,
mowing) required to check woody encroachment.
Our grassland was a net daily C sink before the harvest
on DOY 179 (28 June 2001; Fig. 7), but became a net
source immediately thereafter due to the combined
impacts of low LAI (Fig. 1) and low θ (Figs. 2, 5).
Unlike Bremer et al. (1998) and Bremer and Ham (2002),
we found little reduction in RE after harvesting. Leaf
biomass is a small component of the total respiring
biomass, and leaf area is able to quickly respond to
mechanical destruction to balance RE within days (Dugas
et al. 1999). Thus, the grassland turned into a net daily
sink (on a 24-h basis) as soon as 6 days after harvesting
(Fig. 7). These results are similar to a harvested
Bermudagrass [Cynodon dactylon (L.)] field (Dugas et
al. 1999). In the C. dactylon field, negative daily C fluxes
were nearly restored 5 days after the cut, and fully restored
after 11 days. In both harvesting studies, restoration of
negative daily NEE was impelled by the rapid recovery of
LAI after the cut (Fig. 1; see also Dugas et al. 1999).
However, perturbations on longer time scales, such as
drought or nutrient limitation, may have long-lasting
effects on LAI and NEE (Meyers 2001; Flanagan et al.
2002).
To evaluate the effect of the management protocol on
NEE, we re-parameterized the model to simulate a
Fig. 6 Estimated and modeled
cumulative An (negative fluxes),
RE (positive fluxes), and NEE
Fig. 7 Comparison between estimated (points) and modeled
(circles) daily NEE estimates before and after grass harvesting. The
vertical shaded line represents the duration of the tractor harvest and
is omitted from the record
scenario without a harvest. In these simulations, the
physiological properties (leaf and respiration parameters)
and maximum LAI were not modified after the harvest.
Under these idealized conditions, harvesting had a weak
and transient effect on NEE fluxes at seasonal and annual
time scales, increasing annual NEE by an additional ~
−24 g C m−2 a−1 for both drought and non-drought
scenarios (Table 3). Thus, the model sensitivity analysis
suggests that attempting to manage for C sequestration by
ending the annual harvest over the course of the
measurement period would have resulted in little addi-
tional C sequestration. In reality, harvesting impacts other
processes, including nutrient content and physiological
properties, community composition, and soil compaction
and below-ground dynamics, all of which are likely to
impact both Ac and RE, but whose effects cannot be
assessed with our approach.
It has been suggested that grasslands in warmer and
wetter climates will act as a large carbon sink in the future
to help mitigate greenhouse warming (Ojima et al. 1993).
Although the low estimated and modeled potential C sink
strength for this grassland does not support this notion,
more long-term monitoring must be undertaken before the
role of warm, moist grasslands in C sequestration is
ascertained (e.g., Miranda et al. 1997; Wilsey et al. 2002).
The emerging picture of net ecosystem C cycling in
grassland ecosystems based on this study and similar ones
(Table 1) suggests a characteristic NEE that is close to zero
—but comprised of large assimilatory and respiratory
fluxes—that can readily switch between C source and sink
depending primarily on hydrologic perturbations (Kim et
al. 1992; Bruce et al. 1999; Frank et al. 2001; Flanagan et
al. 2002). A recent study has suggested that future changes
in elevated atmospheric CO2 can have an adverse effect on
grass ecosystem NPP when combined with other likely
global changes including increased N deposition, temper-
ature, and precipitation (Shaw et al. 2002). Another study
has suggested that past increases in atmospheric CO2 have
played a more important role in C sequestration in
grasslands than will projected future CO2 enrichment
(Gill et al. 2002). Thus, if future increases in atmospheric
CO2 may have little effect on future grassland NEE
dynamics, the effects of global changes on the variability
in the hydrological cycle (Vörösmarty and Sahagian 2000;
Jackson et al. 2001; Houghton et al. 2001; Rosenzweig et
al. 2002) may be the key driver of future NEE responses in
grasslands.
Conclusions
This study investigated ET and the direction and
magnitude of NEE and its components (Ac and RE) in a
southeastern grassland ecosystem under drought, under
simulated drought-free conditions, and with rapid changes
in LAI through harvesting.
We found that the relationship between annual ET and
annual precipitation was similar to drought-impacted
warm-temperate and Mediterranean grasslands, and that
‘drought-free’ modeled ET resembled North American
Great Plains grassland ecosystems studied during non-
drought years. The impact of soil moisture limitation on gc
and consequently Ac, and not on the variability of RE, was
the dominant control on NEE at time scales from days to
seasons during a year with mild drought. This contrasts
with results from eddy-covariance studies in European
forests, which have suggested that variability in RE exerts
dominant control on C exchange (Valentini et al. 2000).
Low soil moisture (0.1<θ<0.2) reduced stomatal conduc-
tance for over half of the year, resulting in a grass
ecosystem that was a net source of CO2. In the absence of
soil moisture limitations, model calculations suggest that
this grassland ecosystem would become a small net annual
C sink. The synthesis of results on grassland studies
further implies that interannual variability in NEE is large
(Table 1), and additional sources of climatic and
hydrologic sensitivity should be explored (Knapp and
Smith 2001) to improve predictability of grassland
ecosystem C cycling for the future.
Acknowledgements Support was provided by the National
Science Foundation (NSF-EAR and NSF-DMS), the Biological
and Environmental Research (BER) Program, United States
Department of Energy, through the Southeast Regional Center
(SERC) of the National Institute for Global Environmental Change
(NIGEC), and through the Terrestrial Carbon Processes Program
(TCP) and the FACE project. The authors appreciate the contribu-
tions of data collection from Ben Poulter and Heather McCarthy.
The footprint model of Hsieh et al. (2000) [in Matlab] is available
upon request.
Appendix A: leaf-level assimilation model
According to Farquhar et al. (1980), as later modified by
Collatz et al. (1991) and Campbell and Norman (1998),
the net photosynthetic rate at the leaf scale depends on
light, CO2 concentration, and leaf temperature (Tl) and can
be described as:
An ¼ min JEJC

 
 Rd
where JE and JC are the assimilation rates restricted by
light-driven electron transport processes and ribulose
bisphosphate (RuBP) carboxylase-oxygenase activity (Ru-
bisco), respectively, and Rd is dark respiration. For leaf-
level processes we adopt the ecophysiological convention
of positive fluxes into the leaf. When these fluxes are
scaled to the canopy we revert to the micrometeorological
convention. JE is given by:
JE ¼  em  Qp  Ci  

Ci þ 2

where α is the leaf absorptivity [not to be confused with
the apparent quantum efficiency (αa)] for photosyntheti-
cally active radiation (PAR), em is the maximum quantum
efficiency for leaf CO2 uptake, Qp is PAR irradiance on the
leaf, and Ci is the mean intercellular CO2 concentration.
The values of all parameters are listed in Table 3. The
photosynthetic CO2 compensation point, Γ*, is given by:

 ¼ O2½ 2
where [O2] is the oxygen concentration in air
(210 mmol mol−1), and τ is a ratio of kinetic parameters
describing the partitioning of RuBP to the carboxylase or
oxygenase reactions of Rubisco. Jc is computed from
Jc ¼ VcmaxðCi  
Þ
Ci þ Kc 1þ O2½ =KO2ð Þ
where Vcmax is the maximum catalytic capacity of
Rubisco per unit leaf area (μmol m−2s−1), and Kc and KO2
are the Michaelis constants for CO2 fixation and O2
inhibition with respect to CO2, respectively. Jc increases
linearly with increasing Ci, but approaches a maximum
under a high CO2 concentration state rarely encountered
under present conditions, though likely under future
climate scenarios.
Temperature dependence of kinetic variables is com-
puted following the equations in Campbell and Norman
(1998). Five kinetic parameters are needed to adjust for
temperature: Kc, KO2 , τ, Vcmax and Rd. For the first two
parameters, a modified exponential temperature function
of the form:
k ¼ k25  exp ðTL  25Þ½ 
is employed, where k is defined at the leaf surface
temperature or Tl, k25 is the value of the parameter at 25°C,
and γ is the temperature coefficient for that parameter. τ is
assumed to be 1.3.
Vcmax and Rd are adjusted by:
Vcmax ¼ Vcmax;25 exp :088 TL  25ð Þ½ 1þ exp :29 TL  41ð Þ½ 
and
Rd ¼ Rd;25 exp :069 TL  25ð Þ½ 1þ exp 1:3 TL  55ð Þ½ 
where Vcmax, 25 and Rd, 25 are values of Vcmax and Rd at
25°C, respectively (Campbell and Norman 1998).
Following Collatz et al. (1991), the dark respiration rate
at 25°C (Rd, 25) can be estimated using
Rd;25 ¼ 0:015 Vcmax;25
Appendix B: the boundary line analysis
Stomatal conductance was modeled according to Oren et
al. (1999), with the parameters m and gref generated from a
boundary line analysis. The boundary line analysis sorts
the measured conductance data into 10 bins characterized
by increasing mean light levels. A logarithmic function
relating conductance to VPD is generated for each light
level (i) using data points falling above the mean plus one
standard deviation, after removing outliers at each light
level. The function is given by:
gs;i ¼ ai  ln VPDið Þ þ bi
where i=1–10 (for ten light levels). The slope (ai) and
intercept (bi) for each i were computed via regression
analysis, and the parameter m is the ratio of these two
vectors:
m ¼ a
b
in this study, m=0.64, which is consistent with the
theoretical value of m=0.6 from Oren et al. (1999). We use
the latter value in the model.
The parameter gref is a light-dependent function derived
from fitting the intercept vector b as a logarithmic function
of PAR. Here, we found that gref is
gref ¼0:0922 log PARð Þ
 0:3985 mol m2 s1
Appendix C: night-time atmospheric stability
considerations
Correcting night-time eddy-covariance fluxes under con-
ditions of low u* with respiration models parameterized
using night-time fluxes with high u* is standard eddy-
covariance methodology (Goulden et al. 1996; Aubinet et
al. 2000; Falge et al. 2001a). We conducted a sensitivity
analysis on the annual NEE estimate by varying u*t
between 0 and 0.3, and found that NEE did not vary
appreciably for u*t between 0.12 and 0.18. Hence, we first
filtered the data with u*t=0.12 m s
−1. NEE is highly
sensitive to the u* threshold value chosen (u*t; Barford et
al. 2001), but the exclusive use of u*t has not been
examined, and we propose additional meteorological
constraints to filters used for night-time eddy-covariance
data. Namely, we propose two additional constraints that
only accept fluxes when atmospheric stability conditions
are near-neutral and when the peak of the source-weight
function (xp) lies within the dimensions of the study site
(here 150 m). The atmospheric stability parameter in the
atmospheric surface layer is defined as ς=(z−d)/L, and
near-neutral conditions are defined as |ς|<0.1. We define
the near-neutral atmospheric stability threshold of 0.1 to be
ςn.
The importance of adding ςn, to model night-time
respiration is illustrated by considering the flux footprint at
night for all atmospheric conditions, which exceeds 5 km
(Fig. 8a). Adding u*t alone results in a flux footprint that
exceeds 2 km, an order of magnitude larger than the
dimensions of our field (Fig. 8b). Filtering with both u*t
and ςn (Fig. 8c) reduces the night-time flux footprint to
~1,000 m, which still exceeds field dimensions, so we
further filter night-time flux measurements when the peak
of the source-weight function (xp) exceeds 150 m,
guaranteeing that measured night-time fluxes originate
from our field in a probabilistic sense.
References
Aubinet M, Grelle A, Ibron A, Rannik Ü, Moncrieff J, Foken T,
Kowalski AS, Martin PH, Berbigier P, Bernhofer Ch, Clement
R, Elbers J, Granier A, Grünwald T, Morgenstern K, Pilegaard
K, Rebmann C, Snijders W, Valentini R, Vesala T (2000)
Estimates of the annual net carbon and water exchange of
forests: the EUROFLUX methodology. Adv Ecol Res 30:113–
175
Baldocchi DD, Wilson KB (2001) Modeling CO2 and water vapor
exchange of a temperate broadleaf forest across hourly to
decadal time scales. Ecol Modell 142:155–184
Baldocchi D, Falge E, Gu L, Olson R, Hollinger D, Running S,
Anthoni P, Bernhofer Ch, Davis K, Evans R, Fuentes J,
Goldstein A, Katul G, Law B, Lee X, Malhi Y, Meyers T,
Munger W, Oechel W, Paw UKT, Peilgaard K, Schmid HP,
Valentini R, Verma S, Vesala T, Wilson K, Wofsy S (2001)
FLUXNET: a new tool to study the temporal and spatial
variability of ecosystem-scale carbon dioxide, water vapor, and
energy flux densities. Bull Am Meteorol Soc 82:2415–2434
Barford CC, Wofsy SC, Goulden ML, Munger JW, Pyle EH,
Urbanski SP, Hutyra L, Saleska SR, Fitzjarrald D, Moore K
(2001) Factors controlling long- and short-term sequestration of
atmospheric CO2 in a mid-latitude forest. Science 294:1688–
1691
Bellot J, Sanchez JR, Chirinu E, Hernandez N, Abdelli F, Martinez
JM (1999) Effect of different vegetation type cover on the soil
water balance in semi-arid areas of southeastern Spain. Phys
Chem Earth Part B – Hydrol Oceans Atmos 4:353–357
Bremer DJ, Ham JM (2002) Measurement and modeling of soil CO2
flux in a temperate grassland under mowed and burned regimes.
Ecol Appl 12:1318–1328
Bremer DJ, Ham JM, Owensby CE, Knapp AK (1998) Responses in
soil respiration to clipping and grazing in a tallgrass prairie. J
Environ Qual 27:1539–1548
Bruce JP, Frome M, Haites E, Janzen H, Lal R, Paustian K (1999)
Carbon sequestration in soils. J Soil Water Conserv 54:382–389
Brutsaert W (1982) Evaporation into the atmosphere: theory, history
and applications. Reidel, Boston
Brutsaert W, Sugita M (1992) Application of self-preservation in the
diurnal evolution of the surface-energy budget to determine
daily evaporation. J Geophys Res 97:18377–18382
Campbell GS, Norman JM (1998) An introduction to environmental
biophysics, 2nd edn. Springer, Berlin Heidelberg New York
Clark KL, Gholz HL, Moncrieff JB, Cropley F, Loescher HW
(1999) Environmental controls over net exchanges of carbon
dioxide from contrasting Florida ecosystems. Ecol Appl 9:936–
948
Collatz CJ, Ball JT, Grivet C, Berry JA (1991) Physiological and
environmental regulation of stomatal conductance, photosyn-
thesis, and transpiration: a model that includes a laminar
boundary layer. Agric For Meteorol 54:107–136
Cowan IR (1977) Stomatal behaviour and environment. Adv Bot
Res 4:117–228
Dennis JE Jr (1977) Nonlinear least-squares. In: Jacobs D (ed) State
of the art in numerical analysis. Academic Press, pp269–312
Dugas WA, Heuer ML, Mayeux HS (1999) Carbon dioxide fluxes
over bermudagrass, native prairie, and sorghum. Agric For
Meteorol 93:121–139
Emmerich WE (2003) Carbon dioxide fluxes in a semiarid
environment with high carbonate soils. Agric For Meteorol
116:91–102
Everson CS (2001) The water balance of a first order catchment in
the montane grasslands of South Africa. J Hydrol 241:110–123
Falge E, Baldocchi D, Olson R, Anthoni P, Aubinet M, Bernhofer C,
Burba G, Ceulemans R, Clement R, Dolman H, Granier A,
Gross P, Grunwald T, Hollinger D, Jensen NO, Katul G,
Keronen P, Kowalski A, Lai CT, Law BE, Meyers T, Moncrieff
H, Moors E, Munger JW, Pilegaard K, Rannik Ü, Rebmann C,
Suyker A, Tenhunen J, Tu K, Verma S, Vesala T, Wilson K,
Wofsy S (2001a) Gap filling strategies for defensible annual
sums of net ecosystem exchange. Agric For Meteorol 107:43–
69
Falge E, Baldocchi D, Olson R, Anthoni P, Aubinet M, Bernhofer C,
Burba G, Ceulemans G, Clement R, Dolman H, Granier A,
Gross P, Grunwald T, Hollinger D, Jensen NO, Katul G,
Keronen P, Kowalski A, Lai CT, Law BE, Meyers T, Moncrieff
J, Moors E, Munger JW, Pilegaard K, Rannik Ü, Rebmann C,
Suyker A, Tenhunen J, Tu K, Verma S, Vesala T, Wilson K,
Wofsy S (2001b) Gap filling strategies for long term energy
flux data sets. Agric For Meteorol 107:71–77
Fig. 8a–c The effect of u* and
atmospheric stability on the
calculated night-time CO2 flux
footprint. The measurement
tower is at the center of the polar
plot. Radial lines represent wind
directions (0°=North) and con-
centric lines represent radial
distances from tower (m). a
Points represent footprint dis-
tances from the tower for all
night-time 30-minute runs. b
same as a but implementing the
friction velocity threshold (i.e.,
selecting only runs with u*>u*t,
u*t=0.12). c Same as a but
implementing both friction ve-
locity and atmospheric stability
thresholds (i.e., selecting runs
with u*>0.12 and |(z−d)/L|<ςn,
ςn=0.1)
Fang C, Moncreiff JB (2001) The dependence of soil CO2 efflux on
temperature. Soil Biol Biochem 33:155–165
Farquhar GD, von Caemmerer S, Berry JA (1980) A biochemical
model of photosynthetic CO2 assimilation in leaves of C3
species. Planta 149:78–90
Flanagan LB, Wever LA, Carlson PJ (2002) Seasonal and
interannual variation in carbon dioxide exchange and carbon
balance in a northern temperate grassland. Glob Change Biol
8:599–615
Frank AB, Dugas WA (2001) Carbon dioxide fluxes over a northern,
semiarid, mixed-grass prairie. Agric For Meteorol 108:317–326
Frank AB, Sims PL, Bradford JA, Mielnick PC, Dugas WA,
Mayeux HS (2001) Carbon dioxide fluxes over three great
plains grasslands. In: Follett RF, Kimble JM, Lal R (eds) The
potential of US grazing lands to sequester carbon and mitigate
the greenhouse effect. Lewis, Boca Raton, pp167–187
Gill RA, Polley HW, Johnson HB, Anderson LJ, Maherali H,
Jackson RB (2002) Nonlinear grassland responses to past and
future atmospheric CO2. Nature 417:279–282
Gilmanov TG, Johnson DA, Saliendra NZ (2003a) Growing season
CO2 fluxes in a sagebrush-steppe ecosystem in Idaho: Bowen
ratio/energy balance measurements and modeling. Basic Appl
Ecol 4:167–183
Gilmanov TG, Johnson DA, Saliendra NZ, Akshalov K, Wylie BK
(2003b) Gross primary productivity of the true steppe in central
Asia in relation to NDVI: scaling up of CO2 fluxes. Environ
Manage (in press)
Goulden ML, Munger JW, Fan SM, Daube BC, Wofsy SC (1996)
Measurements of carbon sequestration by long-term eddy
covariance: methods and a critical evaluation of accuracy. Glob
Change Biol 2:169–182
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). Aust J Plant Physiol 25:519–530
Ham JM, Knapp AK (1998) Fluxes of CO2, water vapor, and energy
from a prairie ecosystem during the seasonal transition from
carbon sink to carbon source. Agric For Meteorol 89:1–14
Hoff JH van’t (1898) Lectures on theoretical and physical chemistry.
In: Chemical dynamics part I. Arnold, London, pp224–229
Houghton JT, Ding Y, Griggs DJ, Nouger M, van der Linden PJ,
Day X, Maskell K, Johnson CA (eds) (2001) Climate change
2001: the scientific basis. Third assessment report of Working
Group I. Cambridge University Press, Cambridge
Hsieh C-I, Katul GG, Chi T (2000) An approximate analytical
model for footprint estimation of scalar fluxes in thermally
stratified atmospheric flows. Adv Water Resour 23:765–772
Hunt JE, Kelliher FM, McSeveny TM, Byers JN (2002) Evaporation
and carbon dioxide exchange between the atmosphere and a
tussock grassland during a summer drought. Agric For
Meteorol 111:65–82
Jackson RB, Carpenter SR, Dahm CN, McKnight DM, Naiman RJ,
Postel SL, Running SW (2001) Water in a changing world. Ecol
Appl 11:1027–1045
Jackson RB, Banner JL, Jobbágy EG, Pockman WT, Wall DH
(2002) Ecosystem carbon loss with woody plant invasion of
grasslands. Nature 418:623–626
Janssens IA, Pilegaard K (2003) Large seasonal changes in Q(10) of
soil respiration in a beech forest. Glob Change Biol 9:911–918
Jarvis PG (1976) The interpretation of the variations in leaf water
potential and stomatal conductance found in canopies in the
field. Philos Trans R Soc Lond B Biol Sci 273:593–610
Katul GG, Oren R, Ellsworth D, Hsieh CI, Phillips N, Lewin K
(1997) A lagrangian dispersion model for predicting CO2
sources, sinks, and fluxes in a uniform loblolly pine (Pinus
taeda L.) stand. J Geophys Res 102:9309–9321
Katul GG, Ellsworth DS, Lai C-T (2000) Modeling assimilation and
intercellular CO2 from measured conductance: a synthesis of
approaches. Plant Cell Environ 23:1313–1328
Kelliher FM, Leuning R, Schulze ED (1993) Evaporation and
canopy characteristics of coniferous forests and grasslands.
Oecologia 95:153–163
Kim J, Verma SB (1991) Modeling canopy stomatal conductance in
a temperate grassland ecosystem. Agric For Meteorol 55:149–
166
Kim J, Verma SB, Clement RJ (1992) Carbon dioxide budget in a
temperate grassland ecosystem. J Geophys Res 97:6057–6063
Kirschbaum MUF (2000) Will changes in soil organic carbon act as
a positive or negative feedback on global warming? Biogeo-
chemistry 48:21–51
Knapp AK, Smith MD (2001) Variations among biomes in temporal
dynamics of aboveground primary productivity. Science
291:481–484
Knapp AK, Conard SL, Blair JM (1998) Determinants of soil CO2
flux from a sub-humid grassland: effects of fire and fire history.
Ecol Appl 8:760–770
Lai C-T, Katul GG (2000) The dynamic role of root-water uptake in
coupling potential to actual transpiration. Adv Water Res
23:427–439
Leuning R (1995) A critical appraisal of a combined stomatal-
photosynthesis model for C3 plants. Plant Cell Environ 18:339–
355
Lloyd J, Taylor JA (1994) On the temperature dependence of soil
respiration. Funct Ecol 8:315–323
Long SP, Jones MB, Roberts MJ (eds) (1992) Primary productivity
of grass ecosystems of the tropics and sub-tropics. Chapman
and Hall, London
Medlyn BE, Dreyer E, Ellsworth D, Forstreuter M, Harley PC,
Kirschbaum MUF, Le Roux X, Loustau D, Montpied P,
Strassemeyer J, Walcroft A, Wang K (2002) Temperature
response of parameters of a biochemically-based model of
photosynthesis. II. A review of experimental data. Plant Cell
Environ 25:1167–1175
Meyers TP (2001) A comparison of summertime water and CO2
fluxes over rangeland for well watered and drought conditions.
Agric For Meteorol 106:205–214
Miranda AC, Miranda HS, Lloyd J, Grace J, Francey RJ, McIntyre
JA, Meir P, Riggan P, Lockwood R, Brass J (1997) Fluxes of
carbon, water and energy over Brazilian cerrado: an analysis
using eddy covariance and stable isotopes. Plant Cell Environ
20:315–328
Norman JM, Campbell GS (1989) Canopy structure. In: Pearcy RW,
Ehleringer J, Mooney HA, Rundel PW (eds) Plant physiolog-
ical ecology field methods and instrumentation. Chapman and
Hall, New York, pp 301–325
Nouvellon Y, Rambal S, Lo Seen D, Moran MS, Lhomme JP, Bégué
A, Chehbouni AG, Kerr Y (2000) Modelling of daily fluxes of
water and carbon from shortgrass steppes. Agric For Meteorol
100:137–153
Ojima DS, Dirks BOM, Glenn EP, Owensby CE, Scurlock JO
(1993) Assessment of C budget for grasslands and drylands of
the world. Water Air Soil Pollut 70:95-109
Oren R, Sperry JS, Katul GG, Pataki DE, Ewers BE, Phillips N,
Schäfer KVR (1999) Survey and synthesis of intra- and
interspecific variation in stomatal sensitivity to vapour pressure
deficit. Plant Cell Environ 22:515–1526
Oren R, Ellsworth DS, Johnsen KH, Phillips N, Ewers BE, Maier C,
Schäfer KVR, McCarthy H, Hendrey G, McNulty SG, Katul
GG (2001) Soil fertility limits carbon sequestration by forest
ecosystems in a CO2-enriched atmosphere. Nature 411:469–
472
Parton WJ, Scurlock JMO, Ojima DS, Gilmanov TG, Scholes RJ,
Schimel DS, Kirchner T, Menaut JC, Seastedt T, Moya EG,
Kamnalrut A, Kinyamario JI (1993) Observations and model-
ing of biomass and soil organic matter dynamics for the
grassland biome worldwide. Glob Biogeochem Cycles 7:785–
809
Paz A, Neira A, Castelao A (1996) Soil water regime under pasture
in the humid zone of Spain: validation of an empirical model
and prediction of irrigation requirements. Agric Water Manage
29:147–161
Press WH, Teukolsky SA, Vetterling WT, Flannery BP (1992)
Numerical recipes in Fortran. Cambridge University Press,
Cambridge
Raich JW, Potter CS (1995) Global patterns of carbon dioxide
emissions from soils. Glob Biogeochem Cycles 9:23–36
Raich JW, Schlesinger WH (1992) The global carbon dioxide flux in
soil respiration and its relationship to vegetation and climate.
Tellus 44B:81–99
Redmann RE (1978) Seasonal dynamics of carbon dioxide exchange
in a mixed grassland ecosystem. Can J Bot 56:1999–2005
Reichstein M, Tenhunen JD, Roupsard O, Ourcival JM, Rambal S,
Dore S, Valentini R (2002) Ecosystem respiration in two
Mediterranean evergreen holm oak forests: drought effects and
decomposition dynamics. Funct Ecol 16:27–39
Ripley EA, Saugier B (1974) Microclimate and production of a
native grassland: a micrometeorological study. Oecol Plant
9:333–363
Ripley EA, Saugier B (1978) Biophysics of a natural grassland:
evaporation. J Appl Ecol 15:459–479
Rosenzweig C, Tubiello FN, Goldberg R, Mills E, Bloomfield J
(2002) Increased crop damage in the US from excess precip-
itation under climate change. Glob Environ Change 12:197–
202
Saigusa N, Oikawa T, Liu S (1998) Seasonal variations of the
exchange of CO2 and H2O between a grassland and the
atmosphere: an experimental study. Agric For Meteorol
89:131–139
Sarmiento JL, Wofsy SC (co-chairs) (1999) A US carbon cycle
science plan: a report of the Carbon and Climate Working
Group. US Global Change Research Program, Washington, D.
C.
Scurlock JMO, Johnson K, Olson RJ (2002) Estimating net primary
productivity from grassland biomass dynamics measurements.
Glob Change Biol 8:736–753
Shaw MR, Zavaleta ES, Chiariello NR, Cleland EE, Mooney HA,
Field CB (2002) Grassland responses to global environmental
changes suppressed by elevated CO2. Science 298:1987–1990
Sims PL, Bradford JA (2001) Carbon dioxide fluxes in a southern
plains prairie. Agric For Meteorol 109:117–134
Suyker AE, Verma SB (2001) Year-round observations of the net
ecosystem exchange of carbon dioxide in a native tallgrass
prairie. Glob Change Biol 7:279–289
Suyker AE, Verma SB, Burba GG (2003) Interannual variability in
net CO2 exchange of a native tallgrass prairie. Glob Change
Biol 9:255–265
Tappeiner U, Cernusca A (1996) Microclimate and fluxes of water
vapour, sensible heat and carbon dioxide in structurally
differing subalpine plant communities in the central Caucasus.
Plant Cell Environ 19:403–417
Tjoelker MG, Oleksyn J, Reich PB (2001) Modelling respiration of
vegetation: evidence for a general temperature-dependent Q10.
Glob Change Biol 7:223–230
Valentini R, Gamon JA, Field CB (1995) Ecosystem gas exchange
in a California grassland: seasonal patterns and implications for
scaling. Ecology 76:1940–1952
Valentini R, Matteucci G, Dolman AJ, Schulze E-D, Rebmann C,
Moors EJ, Granier A, Gross P, Jensen NO, Pilegaard K,
Lindroth A, Grelle A, Bernhofer C, Grünwald T, Aubinet M,
Ceulemans R, Kowalski AS, Vesala T, Rannik Ü, Berbigier P,
Loustau D, Gumundsson J, Thorgeirsson H, Ibrom A,
Morgenstern K, Clement R, Moncrieff J, Montagnani L,
Minerbi S, Jarvis PJ (2000) Respiration as the main determinant
of carbon balance in European forests. Nature 404:861–865
Vörösmarty CJ, Sahagian D (2000) Anthropogenic disturbance of
the terrestrial water cycle. Bioscience 50:753–765
Wagai R, Brye KR, Gower ST, Norman JM, Bundy LG (1998) Land
use and environmental factors influencing soil surface CO2 flux
and microbial biomass in natural and managed ecosystems in
southern Wisconsin. Soil Biol Biochem 30:1501–1509
Webb EK, Pearman GI, Leuning R (1980) Correction of flux
measurements for density effects due to heat and water vapour
transfer. Q J R Meteorol Soc 106:85–100
Wever LA, Flanagan LB, Carlson PJ (2002) Seasonal and
interannual variation in evapotranspiration, energy balance
and surface conductance in a northern temperate grassland.
Agric For Meteorol 112:31–49
White R, Murray S, Rohweder M (2000) Pilot Analysis of Global
Ecosystems (PAGE): grassland ecosystems. World Resources
Institute, Washington, D.C.
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. Ecol
Lett 5:367–376
Wilson KB, Baldocchi DD (2001) Comparing independent
estimates of carbon dioxide exchange over 5 years at a
deciduous forest in the southeastern United States. J Geophys
Res 106:34167–34178