288 A. Miyata et al. Agricultural and Forest Meteorology 102 2000 287–303
development of fast response CO
2
analyzers enabled us to measure CO
2
fluxes over a rice canopy by the eddy covariance method Ohtaki and Matsui, 1982;
Ohtaki, 1984, which gave us more reliable flux esti- mates than before. However, the mechanism of CO
2
exchange between rice paddies and the atmosphere is not fully understood. For example, using eddy covari-
ance measurements, Tsukamoto 1993 found a sig- nificantly smaller net CO
2
flux from the atmosphere to a rice canopy when the field was drained com-
pared to when it was flooded, but the reason for the difference was not clear. The existence of floodwater,
anaerobic soil or changes in the micrometeorological environment with flooding will influence root activity,
photosynthesis and respiration of rice plants. Activity of aquatic plants such as algae in the floodwater may
also affect CO
2
exchange between rice paddies and the atmosphere. Many of the data obtained so far are
not sufficiently detailed to examine the influence of these factors on the CO
2
exchange in rice paddies. Paddy fields are also one of largest sources in the
global budget of CH
4
. Based on incubation experi- ments in a laboratory, Koyama 1963 first estimated
the CH
4
production rate by world rice production to be 190 Tg per year. Since the 1980s there have been
numerous field measurements of CH
4
fluxes in various rice paddies over the world e.g. Cicerone and Shet-
ter, 1981; Holzapfel-Pschorn and Seiler, 1986; Schütz et al., 1989; Sass et al., 1990; Yagi and Minami, 1990;
Khalil et al., 1998, leading to revised estimates of the global CH
4
emission from rice paddies of 60 Tg per year, but with uncertainty ranging from 20 to 100 Tg
per year IPCC, 1995. Most estimates of CH
4
fluxes have used chambers placed over plants, soil and paddy water, but chambers
disturb the environment during measurement. Several pioneering studies on net CH
4
fluxes over rice pad- dies using non-disturbing micrometeorological meth-
ods have now been made Denmead, 1991; Simpson et al., 1995; Harazono et al., 1996. The flux-gradient
approach was used in these studies rather than the eddy covariance technique because, unlike for CO
2
, fast-response gas analyzers for CH
4
have not been available until very recently. Eddy covariance mea-
surements using tunable diode laser absorption spec- troscopy are now becoming available Verma et al.,
1992; Shurpali et al., 1993; Edwards et al., 1994; Kim et al., 1998a, b.
Net exchanges of CO
2
and CH
4
between rice pad- dies and the atmosphere are controlled by several bi-
ological and physical processes. During the daytime plant photosynthesis leads to uptake of CO
2
from both the atmosphere and from respired CO
2
emit- ted by the soil and floodwater. Respiration at night
leads to an efflux of CO
2
to the atmosphere. CH
4
is released to the atmosphere by ebullition, diffusion across the water-air interface and by transport through
aerenchyma, well-developed intracellular air spaces which supply atmospheric oxygen from pores in the
leaves, through the plant stems, and to the roots of the rice plants Nouchi, 1994. Up to 90 of CH
4
emis- sion occurs through the aerenchyma in undisturbed
paddy fields Minami and Neue, 1994. To improve understanding of the process control-
ling CO
2
and CH
4
exchanges in rice paddies, an intensive field experiment called IREX96 the 1996
International Rice Experiment was conducted in Japan during August 1996. In this paper, we present
measurements of CO
2
and CH
4
fluxes over a rice canopy obtained using micrometeorological tech-
niques; eddy covariance for CO
2
and flux-gradient methods for CH
4
. The measurements were used to assess the role of floodwater in controlling the ex-
changes of CO
2
and CH
4
from the soil, the floodwater and the plant canopy. Factors controlling exchange
processes are examined further in a companion pa- per Leuning et al., 2000, where we estimate source
strength distributions for CO
2
and CH
4
within the rice using an analysis of turbulent dispersion and
measured concentration profiles.
2. CH
4 4
4
flux measurement using flux-gradient theory
Methane fluxes over the rice canopy were measured using two methods based on flux-gradient theory; an
aerodynamic method and a gradient technique which uses the eddy covariance flux of CO
2
, a reference scalar tracer. These methods have been used conven-
tionally for the measurement of fluxes of gases as well as sensible heat and latent heat e.g. Inoue et al., 1958,
1969. In this study we used a modified aerodynamic method Harazono and Miyata, 1997, and there-
fore it is instructive to describe here the methods in detail.
A. Miyata et al. Agricultural and Forest Meteorology 102 2000 287–303 289
We assume horizontal uniformity in surface fluxes, statistically stationary turbulence, and that the pro-
duction and destruction of the gas within the surface layer can be neglected. Following the custom, an over-
bar represents time-averaged quantities, and a prime does deviation from the time-averaged value. From
Monin–Obukhov similarity theory, the vertical flux of the gas F is related to the mean vertical gradient of the
gas mass mixing ratio s as follows e.g. Fowler and Duyzer, 1989; Denmead, 1994:
F = −ρ
a
K
g
z ∂s
∂z = −
ρ
a
κu
∗
z − d φ
g
ζ ∂s
∂z 1
where K
g
z is the eddy diffusivity at a height z, d is the zero-plane displacement, u
∗
is the friction velocity, ρ
a
is the density of dry air and κ is von Karman’s constant 0.4. The term φ
g
ζ in Eq. 1 provides the correction to the eddy diffusivity as a function of the
Monin–Obukhov stability parameter ζ , defined as ζ =
z − d L
= − κgz − dw
′
θ
′ v
u
3 ∗
θ
′ v
2 where, L is the Monin–Obukhov length, w
′
θ
′ v
is the covariance between the vertical wind component w
and the virtual potential temperature θ
v
. This is given by, θ
v
= θ 1 + 0.61q ≈ θ + 0.61θ q, where θ is the
potential temperature and q is the specific humidity. By definition, w
′
θ
′ v
= H ρC
p
+ 0.61 θρ E, where
H is the sensible heat flux, E is the water vapour flux, ρ is the density of moist air and C
p
is the specific heat capacity of air at constant pressure.
The friction velocity in Eq. 2 is measured by the eddy covariance method using a three-dimensional
sonic anemometer. The covariance between w and the temperature from the sonic signal closely approxi-
mates w
′
θ
′ v
in Eq. 2 because of the humidity depen- dence of the sound speed Kaimal and Gaynor, 1991;
Hignett, 1992. By using u
∗
and w
′
θ
′ v
obtained from the sonic signal, the stability parameter ζ and therefore
φ
g
ζ can be determined. Once φ
g
is known, we can calculate F from the mean vertical gradient of the gas
mixing ratio using Eq. 1. In this paper, we call this the aerodynamic AD method, and use ‘AD–F
CH4
’ to represent the CH
4
flux determined by this method. An advantage of this method over the conventional
aerodynamic approach e.g. Monteith and Unsworth, 1990 is that u
∗
and ζ determined from the eddy covariance method are more reliable than those from
the profiles of windspeed and temperature. From a practical point of view, we must approxi-
mate the vertical gradient of the mixing ratio using measurements made at two heights above the canopy,
z
1
and z
2
z
1
z
2
. By integrating Eq. 1 from z
1
to z
2
and assuming that F and u
∗
are constant between the two heights, we obtain
F = −Dρ
a
1s = −κu
∗
Z
ζ
2
ζ
1
φ
g
ζ ζ
dζ
− 1
ρ
a
1s 3 where subscripts 1 and 2 represent the values at z
1
and z
2
, respectively, and 1 denotes the difference of the quantity at z
2
from that at z
1
. D is referred to as the diffusion velocity or conductance. Eq. 3 was used
to calculate the CH
4
flux from the mean difference of the gas mixing ratio between two heights above the
canopy. Based on the results of previous field studies over
short vegetation, we assume that φ
g
is equal to the dimensionless gradient of the potential temperature φ
h
Denmead, 1994. Following Dyer and Hicks 1970 and Webb 1970 we used:
φ
h
= 1 − 16ζ
− 12
ζ ≤ 0 4a
φ
h
= 1 + 5ζ
ζ ≥ 0 4b
The zero-plane displacement d, which is required for the determination of ζ and K
g
z, was assumed to be 0.7 times the mean plant height.
A second method which uses CO
2
as a tracer to calculate CH
4
fluxes was also used in this study. This method assumes that φ
g
for CO
2
and CH
4
in Eq. 1 are equal and hence no stability corrections to the flux
estimates are required. With these assumptions F
CH4
= 1s
CH4
1s
CO2
F
CO2
= M
CH4
M
CO2
1c
CH4
1c
CO2
F
CO2
5 where c is the volume mixing ratio, M is the molecular
mass, and subscript CH4 and CO2 represent CH
4
and CO
2
, respectively. F
CH4
is determined using Eq. 5 from F
CO2
measured by the eddy covariance method and the ratio of the mean vertical gradient of methane
mixing ratio to that of CO
2
. In this paper, we call this method the ‘K
CO2
method’, and use ‘K
CO2
–F
CH4
’ to represent the CH
4
flux determined by this method. Potential temperature or water vapor can also be
290 A. Miyata et al. Agricultural and Forest Meteorology 102 2000 287–303
used as a reference scalar instead of CO
2
. These tracer methods are advisable particularly when the
nighttime fluxes are being investigated because calm periods may invalidate the aerodynamic approach
Denmead, 1994. In this study, we choose CO
2
as a tracer because the nighttime vertical gradients of CO
2
mixing ratio can be measured more accurately than those of potential temperature or water vapor.
3. Experimental
