Tree Physiology 28, 499–508
© 2008 Heron Publishing—Victoria, Canada
Variation in photosynthetic light-use efficiency in a mountainous
tropical rain forest in Indonesia
ANDREAS IBROM,1–3 ALEXANDER OLTCHEV,1,4 TANIA JUNE,5,6 HEINER KREILEIN,1
GOLAM RAKKIBU,1 THOMAS ROSS,1 OLEG PANFEROV1 and GODE GRAVENHORST1
1
2
3
4
5
6
Institute of Bioclimatology, Georg-August-University, Büsgenweg 1, D-37077 Göttingen, Germany
Present address: Biosystems Department, Risø National Laboratory, Technical University of Denmark, Building 309, Frederiksborgvej 399,
DK-4000 Roskilde, Denmark
Corresponding author ([email protected])
Present address: A.N. Severtsov Institute of Ecology and Evolution of RAS, Moscow, Russia
SEAMEO BIOTROP, Bogor, Indonesia
Institute Pertanian Bogor, IPB, Bogor, Indonesia
Received May 8, 2007; accepted August 2, 2007; published online February 1, 2008
Summary Photosynthetically active radiation (Q)-use efficiency (ε) is an important parameter for deriving carbon fluxes
between forest canopies and the atmosphere from meteorological ground and remote sensing data. A common approach is to
assume gross primary production (Pg ) and net primary production (Pn ) are proportional to Q absorbed by vegetation (Qabs ) by
defining the proportionality constants ε Pg and ε Pn (for Pg and
Pn , respectively). Although remote sensing and climate monitoring provide Qabs and other meteorological data at the global
scale, information on ε is particularly scarce in remote tropical
areas. We used a 16-month continuous CO2 flux and meteorological dataset from a mountainous tropical rain forest in
central Sulawesi, Indonesia to derive values of ε Pg and to investigate the relationship between Pg and Qabs. Absorption was
estimated with a 1D SVAT model from measured canopy structure and short wave radiation. The half-hourly Pg data showed a
saturation response to Qabs. The amount of Qabs required to
saturate Pg was reduced when water vapor saturation deficit
(D) was high. Light saturation of Pg was still evident when
shifting from half-hourly to daily and monthly time scales.
Thus, for a majority of observations, Pg was insensitive to
changes in Qabs. A large proportion of the observed seasonal
variability in Pg could not be attributed to changes in Qabs or D.
Values of ε Pg varied little around the long-term mean of 0.0179
mol CO2 (mol photon) –1 or 0.99 g C MJ –1 (the standard deviations were ± 0.006 and ± 0.0018 mol CO2 (mol photon) – 1 for
daily and monthly means, respectively). In both cases, ε Pg values were more sensitive to Qabs than to daytime D. These findings show that the current ε-approaches fail to predict Pg at our
tropical rain forest site for two reasons: (1) they neglect saturation of Pg when Qabs is high; and (2) they do not include factors,
other than Qabs and D, that determine seasonality and annual
sums of Pg.
Keywords: canopy photosynthesis, eddy correlation, gross primary production, light use efficiency, remote sensing.
Introduction
The global carbon cycle is usually studied by a monitoring approach that includes tower flux measurements, flask gas concentration sampling, satellite observations and atmospheric
and ecosystem modeling (Running et al. 2000). Gross primary
production (Pg ) and net primary production (Pn ) at global
scales can be estimated with ecosystem models driven with
meteorlogical data and remotely sensed data on vegetation
properties (Running et al. 2004). However, model estimates of
global Pg and Pn vary significantly, reflecting the need to better
understand the processes influencing carbon dynamics
(Cramer et al. 1999).
A relatively simple and often used approach for estimating
carbon accumulation by terrestrial vegetation was initially
suggested by Monteith (1977). It assumes that rates of primary
production are proportional to rates of solar radiation intercepted by vegetation, with the proportionality factor photosynthetically active radiation (Q)-use efficiency (ε), such that:
Pg = ε Pg Qabs
(1)
Pn = ε Pn Qabs
(2)
where Qabs is amount of Q that is absorbed by vegetation and
ε Pg and ε Pn are Q-use efficiencies for Pg and Pn, respectively.
Linear Q-use efficiency approaches (ε approaches) are currently used in regional- and global-scale studies based on remotely sensed vegetation structure data and regional meteoro-
500
IBROM ET AL.
logical ground data (e.g., Field et al. 1995, Ruimy et al. 1996,
Running et al. 2004, Xiao et al. 2004, Heinsch et al. 2006).
The fraction of incident Q that is absorbed by vegetation can
be estimated as a linear or nonlinear function of the normalized difference vegetation index (NDVI) or the enhanced vegetation index (EVI) (Knyazikhin et al. 1998, Running et al.
2000, Xiao et al. 2004). In some studies, Qabs is directly derived as a function of LAI and the light extinction coefficient
(Ruimy et al. 1999, Ito and Oikawa 2004). If used for the calculation of Pg, Xiao et al. (2004) suggests that only the fraction
of Qabs that is being absorbed by photosynthetic tissue (e.g.,
green leaves) be used and that Q absorption by non-assimilating structures (e.g., senescent or dry foliage, stems and
branches) be ignored. In the case of homogeneous vegetation,
Qabs can be calculated from bidirectional (up and down) Q
measurements above and below a canopy. To estimate Qabs in
heterogeneous vegetation (e.g., tropical rain forests), extensive, time-consuming measurements are needed to capture the
heterogeneity of the radiation field appropriately. Alternatively, Qabs can be estimated with adequately parameterized
one- or three-dimensional radiative transfer models (Wang
and Jarvis 1990, Knyazikhin et al. 1997, Gravenhorst et al.
1999, Ibrom et al. 2006).
Values of ε for different vegetation types have been summarized in several publications (Ruimy et al. 1999, Heinsch et al.
2003, Ito and Oikawa 2004, Still et al. 2004), and differ significantly both among different geographical regions and across
vegetation types of the same natural zone. To adapt ε to especially adverse conditions or phenological phases, it must be
expressed as a function of maximal ε (εmax), which is then multiplied by functions that include the effects of high D, air or
vegetation temperature and low soil water content (Heinsch et
al. 2003, Xiao et al. 2004).
Compared with other vegetation types, information on the
spatial and temporal variability in ε in tropical forests is relatively scarce, despite being of major importance to the global
terrestrial CO2 sink (Grace 2004). Field studies of net ecosystem CO2 exchange (Fc), Pg and Pn are mainly confined to tropical rain forests located in South America (Grace et al. 1995,
Saleska et al. 2003) and should be extended to other tropical
areas.
Recent attempts have been made to validate and improve the
ε-approach (Zhao et al. 2005, Turner et al. 2006a). Of particular importance are those that describe the biological determinants of ε Pg , its maximum value and the functions that reduce
this maximum according to environmental conditions and
physiological traits. As yet, the approach as a whole has not
been scrutinized.
It has been suggested that the relationship between Pg and
Qabs could be described with hyperbolic functions (e.g., Ruimy
et al. 1996, Choudhury 2001, Turner et al. 2003). However
Ruimy et al. (1996) concluded from their sensitivity analysis
that estimation of daily Pg and Pn of different vegetation types
at the global scale with a linear model provides more accurate
predictions than with a nonlinear model. Turner et al. (2003)
reported that the proposed relationships between ε Pg and D or
air temperature were not obvious in the measured data from
four different biomes. Instead they found more or less pronounced negative relationships between ε Pg and Qabs even in
dense forest canopies. Their findings appear relevant to an assessment of the current ε-approach algorithms, because the
usual linear approaches will not yield accurate results for situations where ε Pg depends on Qabs.
In this study we investigated relationships between Pg and
Qabs in a mountainous tropical rain forest in South East Asia
and compared those results with those of the current ε-approach. We used a continuous 16-month dataset of turbulent
CO2 fluxes and radiation measurements together with biometric observations that allow simulation of Qabs with a canopy
model. We tested the major assumption of the ε-approach, i.e.,
that the proposed linear relationship between Pg and Qabs exists
at our study site.
Materials and methods
Study area
The experimental site is part of the Lore Lindu National Park,
a natural tropical rain forest area located near Bariri in central
Sulawesi, Indonesia (01°39′ S, 120°10′ E, 1427 m a.s.l.). The
vegetation is uniform and the terrain is characterized by a
smooth relief with a small slope of less than 5° within a radius
of 1000 m around the flux tower. The investigated forest canopy includes up to 90 tree species ha –1 and trees of different
ages and structures. Dominant tree species are Castanopsis
accuminatissima (Bl.) A. DC. (29%), Canarium vulgare
Leenh. (18%) and Ficus spp. (9.5%). Total projected LAI is
about 7.2, and mean tree density is about 567 trees ha – 1. Although the maximum height of the trees in the stand approaches 36 m, mean tree height is 21 m (Ibrom et al. 2007).
The climate of the study area is tropical with a distinct monsoonal wet season and a drier season. The wet season lasts
from about November to April and the dry season from May to
October. Mean annual temperature is 19.6 °C. The annual precipitation of 1700 mm year – 1 is relatively moderate for tropical mountainous areas.
Eddy covariance flux measurements
Fluxes of CO2 were measured with eddy covariance equipment installed at a height of 48 m (about 12 m above the forest
stand) on a 70-m tower. The equipment consisted of a 3D ultrasonic anemometer-thermometer USA-1 (Metek, Germany)
and an open-path CO2 /H2O gas analyzer LI-7500 (Li-Cor,
Lincoln, NE) linked to a field computer that logged raw data at
10 Hz. Data were processed with the software rcpm (Morgenstern 2000, Ibrom 2001), employing spike detection
(Højstrup 1993), 2D rotation (Moore 1986), linear detrending
(Rannik and Vesala 1999) and consideration of density fluctuations (Webb et al. 1980). The method requires an appropriately extended flat and homogeneous surface or fetch, and
builds on the similarity of the CO2 flux at the surface with the
sum of the turbulent CO2 flux through the atmospheric boundary layer and CO2 storage change beneath the sensor. These
conditions are met at the site during daytime and on windy
TREE PHYSIOLOGY VOLUME 28, 2008
LIGHT-USE EFFICIENCY IN A TROPICAL FOREST
nights. The eddy covariance method is currently the chief empirical method to estimate scalar fluxes between forests and
the atmosphere (Valentini et al. 2000, Baldocchi et al. 2001).
Merits and limitations of the method have been described in
many publications (e.g., Aubinet et al. 2000, Lee et al. 2004)
and are further discussed for our particular application.
Net CO2 fluxes, ecosystem respiration and gross
photosynthetic rates from eddy covariance data
Estimation of Fc included consideration of the CO2 storage
term using the approach suggested by Hollinger et al. (1994),
i.e., approximating the course of CO2 concentration in the air
column beneath the sensor by the CO2 concentration at the
sampling height. This approach is suitable in many studies
provided that periods of low turbulence are excluded from the
dataset.
A key problem in the interpretation of CO2 flux measurements in the tropics is the scarcity of nighttime flux data, because of weak nocturnal turbulence conditions (e.g., Loescher
et al. 2003, Falk 2004, Miller et al. 2004). Weak turbulence at
night may result in underestimation of nocturnal CO2 fluxes
and ecosystem respiration (RE ) and in errors in estimating of
Pg from Fc. Friction velocity (u* ) is currently used as a criterion to identify and exclude periods of low turbulent mixing.
Using the approach suggested by Aubinet et al. (2000), our
data showed that nighttime fluxes, i.e., RE , became independent of u* when u* ≥ 0.3 m s –1. Therefore only the flux data collected when u* ≥ 0.3 m s –1 were used to calculate two monthly
means of RE . We calculated Pg as the difference between Fc
and calculated RE. To comply with current ε-approaches, CO2
uptake from the atmosphere by the canopy is expressed as a
positive flux. The errors associated with the estimation of RE
were usually small, because daytime data were used that were
collected when unstable conditions and high wind speeds
guaranteed sufficiently strong turbulence. Even if RE was underestimated, which is unlikely because of the pronounced saturation function of RE versus u*, the effect on Pg would be
comparably small and independent of radiation. Thus, the saturation phenomenon that we investigated is unconnected with
the estimation of Pg. We are aware, however, that turbulent
flux measurements are prone to random errors in the order of
at least 20% resulting from natural variability in the atmospheric boundary layer (Wesely and Hart 1985). This can be
seen in the comparison of measured and modeled Pg reported
by Ibrom et al. (2006) and in daytime data from multi-tower
studies (Richardson et al. 2006). Given the random nature of
this error and the large quantity of data, the mean behavior of
the measured CO2 flux can be deemed accurate (Moncrieff et
al. 1996).
Calculation of Q absorbed by a forest stand
We calculated Qabs with the 1D multi-layer SVAT model,
MixFor-SVAT, that was developed to describe energy, water
and CO2 exchange between multi-species forest ecosystems
and the atmosphere (Oltchev et al. 2002, Ibrom et al. 2007).
Mixfor-SVAT simulates H2O and CO2 fluxes for both an entire
501
ecosystem and for individual tree species at the same time,
thereby taking into account the individual biophysical properties of the tree species as well as the variety of their responses
to environmental changes.
Absorption of Q was modeled for both active (green leaves)
and inactive (dead foliage, branches and stems) vegetation
components. The modeling algorithm uses the two-stream approximation (Dickinson 1983, Pinty et al. 2006) both for direct and diffuse solar radiation. The fraction of sunlit and
shaded leaves in the canopy was calculated according to Dai et
al. (2004). Reflection and transmission coefficients for each
sub-layer were calculated based on individual optical properties of tree species, vertical distributions of leaf area density
for each tree species and fractions of leaf area density of the
plant area index. The main input for the simulations is measured global shortwave radiation (0.3–2.5 µm wavelengths;
CM11 pyranometer, Kipp and Zonen, Delft, The Netherlands). From these observations, both incident Q and cloudiness were calculated with empirical functions (Ross 1981).
Q response of gross canopy photosynthesis
The empirical relationship between measured half-hourly Pg
and simulated Qabs was described by a non-rectangular hyperbola (Thornley 1976):
Pg =
1
(α Qabs + Pmax −
2θ
(α Qabs + Pmax )
2
− 4θα Qabs Pmax ⎞⎟
⎠
(3)
where α is quantum-use efficiency, Pmax is maximum gross
photosynthetic rate and θ is the curvature parameter. In a second approach Pmax was allowed to depend on D as:
b
⎛
a− 2 ⎞
Pmax = ⎜⎜1 − e D ⎟⎟ Pmax, 0
⎝
⎠
(4)
where a and b are empirical parameters, Pmax,0 is Pmax at D →
0 Pa. The right-hand side term in brackets is referred to in the
text as the normalized reduction function of D, f(D).
We calculated ε Pg from Pg and Qabs by rearranging Equation 1. To characterize the processes at different time scales,
the data were analyzed as half-hourly data, daytime means,
and daily and monthly sums.
Results
The study period was characterized by varying environmental
conditions including clear sky and overcast periods as well as
wet and dry conditions. Mean daily temperature ranged between 17.1 and 21.6 °C, mean daily relative humidity between
61 and 98% and daily solar radiation between 2.7 and 11.6 MJ
m – 2 day – 1. Monthly precipitation ranged between 20 mm (in
August 2004) and 321 mm (in April 2005).
TREE PHYSIOLOGY ONLINE at http://heronpublishing.com
502
IBROM ET AL.
Absorbed Q versus incoming global radiation
Simulated Qabs was linearly related to incoming global radiation (r 2 > 0.99). The root mean square error was 14.5 µmol m – 2
s – 1. Scattering around the regression line was mainly caused
by differing proportions of beam and diffuse radiation and the
effects of solar elevation angles. The slope of the linear relationship was 1.82 µmol J –1, which accords well with experimental data (e.g., Wang et al. 2007). On average about 90% of
incident Q was absorbed by the canopy. The remainder was either reflected (about 6%) or transmitted (about 4%) toward the
forest floor. The dark understory conditions resulted in sparse
forest-floor vegetation. The canopy base was about 10 m
above the forest floor.
Distribution of Q beneath the canopy was not accurately
measured because the number of available Q sensors was limited; however, we measured the shortwave albedo and the net
radiation budget (Ibrom et al. 2007). The model MixFor-SVAT
matched these results closely (r 2 > 0.99) (Oltchev et al. 2007).
Thus the only remaining error might be an under- or overestimation of transmission. Given the high LAI of the broadleaved canopy and the sparse below-canopy vegetation, the estimated value was reasonable. A potential error in Qabs would
be systematic and would not lead to major changes in the relationship between Pg and Qabs. Even a conservative estimate of
the inaccuracy of the simulated Q transmission (50% relative
error) would correspond to a small error of ± 2% relative to incident Q flux density.
Q response of gross photosynthesis
The general relationship between Pg and Qabs at the half-hourly
timescale showed a hyperbolic increase in Pg with increasing
Qabs (Figure 1a), which is similar to results from other CO2
flux studies (Goulden et al. 1996, Valentini et al. 1996, Jarvis
et al. 1997, Pilegaard et al. 2001, Ibrom et al. 2006). A linear
dependency of Pg on Qabs was found only when Qabs was less
than 600 µmol m – 2 s –1. Gross photosynthesis saturated
quickly at Qabs > 600 µmol m – 2 s –1 and was, therefore, almost
independent of Qabs for a large portion of flux observations.
Based on a Blackman response function (Blackman 1905),
i.e., constraining Equation 3 to a θ value of 1, we calculated
that Pg saturated (Pmax ) at Qabs = 770 µmol m – 2 s –1 (Qsat ):
Qsat =
Pmax, θ =1
α θ =1
(5)
The relationship between measured Pg and simulated Qabs is
characterized by considerable scatter: 68% of the variability in
Pg could be explained by the non-rectangular hyperbola model
driven by Qabs only (Equation 3). Inclusion of the sensitivity of
Pmax to D, using Equation 4, increased the explained variability
to 72%. The remaining scatter is caused by the random error in
turbulent flux measurements and seasonal variability in Pg
which is is unrelated to Qabs or D. The parameters and regression statistics are listed in Table 1.
The lower curve in Figure 1a shows the effect of extreme
D—the 99% percentile of D is 1953 Pa for this site—indicat-
ing a substantial sensitivity of Pg to climatic conditions with
high D. From the regression parameters shown in Table 1, it
can be concluded that high D reduced Qsat on average by more
than 40% of the hypothetical value that was estimated for humid conditions. The relative D-dependency of Pmax is depicted
in Figure 1b. The combined regression model, i.e., Equations 3
and 4, shows a decrease in Pg when D exceeded 500 Pa. The
combined model with a Blackman type of photosynthetic Q
response, allows calculation of the dependency of Qsat on D
(Figure 1c). The maximum Qsat reached about 900 µmol m – 2
s – 1 and decreased quickly to 2/3 of this value over the range of
the D that was observed at the study site.
The residual variability of Pg was unrelated to the other observed meteorological or biological parameters. However using monthly data subsets to fit the combined Q and D response
curves improved the model representation by another 14%,
i.e., the r 2 value reached 86.3%. The slope and offset of the regression between observed and predicted Pg were 0.996 and
–0.0065 µmol m – 2 s – 1, respectively.
To characterize daily and monthly time scales and to facilitate comparison with the remote sensing literature, we converted the units for absorbed radiation from mean daytime Qabs
(µmol m – 2 s – 1) to daily sums (MJ m – 2 day – 1). The conversion
factor is 1.8783 10 – 2 MJ s µmol – 1 day – 1, i.e., 1 µmol m – 2 s – 1
corresponds to 1.8783 10 – 2 MJ day – 1. Considering a 12-h
photoperiod at the site, a mean daytime Qabs value of 600 µmol
m – 2 s – 1 corresponds to a daily sum of 5.6 MJ m – 2 day – 1. At the
daily time scale, the dependency of daytime Pg on Qabs was
similar to the relationships obtained for half-hourly data.
However, days with mean daytime Qabs < 5.6 MJ m – 2 day – 1
were rare and thus the linear portion of the curve is small (Figure 2a). Low irradiances occurred only during periods of
cloudy and rainy weather. For days with mostly sunny
weather, when Qabs was usually > 5.6 MJ m – 2 day – 1, Pg was independent of Qabs. Saturation led to a more or less constant Pg
value of 7.32 ± 1.1 g C m – 2 day – 1 (mean ± standard deviation).
Only a small part of the variability in daily Pg could be attributed to the seasonal development of maximum Pg. Mean Pg
(± standard deviation) for a dry month (September 2004)
reached 6.7 ± 0.7 g C m – 2 day – 1, whereas it reached 8.3 ± 0.9 g
C m – 2 s – 1 for the most productive month of the study (November 2003).
Variability in mean monthly daytime Pg and Qabs was small,
ranging between 6.5 and 8.5 g C m – 2 day – 1, and 7.48 and
9.35 MJ m – 2 day – 1, respectively. As with the daily data, the results indicated little dependency of Pg on Qabs for the entire period of measurements (Figure 2c).
Figure 3 shows the seasonal development in measured
monthly Pg together with the seasonal pattern of precipitation
and predicted Pg values from four different regression models.
The solid symbols are predicted values from the regression
analyses based on the entire data set (single fit), and the open
symbols are predicted values from monthly fits. In both cases,
the pure Qabs response and the combined response with D are
shown. With one exception, the single fit approaches failed to
capture the observed seasonality in Pg. The low Pg value in
April 2004 was represented by both types of single fits, indi-
TREE PHYSIOLOGY VOLUME 28, 2008
LIGHT-USE EFFICIENCY IN A TROPICAL FOREST
503
Figure 1. (a) Measured half-hourly canopy
gross primary production rates (Pg ) of the
tropical rain forest versus simulated absorbed photosynthetic active radiation
(Q abs ) from October 2003 to February
2005. Broken curves are predicted Pg at
vapor pressure saturation deficits (D) =
0 Pa (upper curve) and D = 2000 Pa
(lower curve). The solid line shows predicted Pg without consideration of any D
dependency. (b) Reducing effect of increasing D on saturated Pg (see Equation 4) and (c) the effect of increasing D
on saturating Q abs (Q sat ) using Equation 5.
The parameter values of the regression
models are listed in Table 1.
cating that low Qabs was responsible for the low Pg. April 2004
was also the wettest month, when clouds reduced Q. Conversely, during the drier season from August to October 2004,
the sole Qabs-response model overestimated Pg. Overestimation could be reduced by including D in the combined model.
The combined model reduced the predictions in two out of
three cases to values that were close to the measured values
during this period. However, in other cases with more extreme
monthly Pg values, the single fit regressions failed to capture
seasonality, whereas both of the monthly fits came close to the
measured value.
These results indicate that seasonality of Qabs and D contributed only slightly to the seasonal course of Pg, and likely also
to the interannual variability of the annual sums. Among possible factors that may be related to the seasonality of Pg are soil
water availability and the fraction of direct solar radiation. Unfortunately, these factors have yet to be measured at the site.
The monthly regression analyses showed that seasonality affected both α and Pmax with relatively low values in February
and September 2004. The low monthly Pg in September 2004
coincided with a relatively dry period lasting from the second
half of July until the first half of September 2004. After rainfall
in September Pg recovered to previous rates.
The relatively small contribution of Qabs to the variability in
daily and monthly Pg is emphasized by the averaged diurnal
courses of Qabs, D and Pg (Figure 4). The values of Pg and Qabs
started to increase at the same time of day, but the relative increase with time was much steeper for Pg than for Qabs. Gross
photosynthesis reached maximal values at 1000 h, 2 h before
Qabs, which peaked around noon. Although D started increasing with Qabs, it reached its maximum about 1–2 h after Qabs.
The strong increase in Pg weakened soon after 1000 h, whereas
Qabs continued to increase strongly. At this time, D exceeded
500 Pa, which is the value when half hourly Pg tended to be reduced by D according to the combined regression model (see
Figure 1b). The broken line indicates the mean diurnal course
of Qsat, calculated from Equation 5. The noon and early afternoon depression in Pg followed the course of Qsat as long as
Qabs > Qsat, i.e., from 0900 to 1500 h. Thus Pg was saturated on
average during half of the daytime hours.
TREE PHYSIOLOGY ONLINE at http://heronpublishing.com
504
IBROM ET AL.
Table 1. Regression statistics of the relationship between measured
canopy gross primary production (Pg ) and absorbed Q (Q abs ). All parameter values are highly significant (P < 0.0001). The parameters α,
θ and Pmax are the initial slope, a curvature parameter and the saturated photosynthetic rate, respectively. Subscript c stands for the combined Q-response model. Parameters a and b are empirical parameters
for the vapor pressure saturation deficit (D) dependency of Pmax,0, i.e.,
the Q saturated canopy photosynthesis rate at D = 0 Pa. The coefficient of determination (r 2 ) indicates the proportion of the variability
of observed Pg that can be explained by the model predictions. Abbreviation: SE = standard error.
Parameter
Value
SE
Equation 3
α (mol CO2 mol –1 photons)
θ
Pmax (µmol m – 2 s –1)
r2
Number of observations
0.028
0.98
21.17
0.68
11,639
0.0004
0.004
0.14
Equations 3 and 4 combined
αc (mol CO2 mol –1 photons)
θc
a
b (Pa2 )
Pmax,0 (µmol m – 2 s –1)
r2
Number of observations
0.031
0.88
–0.46
1,413,400
28.80
0.72
11,595
0.0005
0.02
0.03
97,200
0.67
Q-use efficiency of gross photosynthesis
The observed Q-responses at different time scales indicate that
ε Pg is not a constant entity that describes the physiological capacity of transforming absorbed quanta into assimilated CO2.
Contrary to the current ε-approaches, but consistent with other
findings (e.g., Turner et al. 2003), the effects of D or temperature are small compared with the effect of Qabs. That the Q-response of Pg can be accurately described by a Blackman response curve indicates that the traditional ε-approach can be
used only as long as Qabs < Qsat. Above this level, ε Pg depends
strictly hyperbolically on Qabs and to a certain degree on other
factors like D:
⎧ ε Pg max f ( D)
⎪
ε Pg = ⎨
Qsat
⎪ ε Pg max f ( D) Q
abs
⎩
Qabs ≤ Qsat
Qabs > Qsat
(6)
This key relationship was valid when considering half
hourly, daily or monthly means (Figures 2b and 2d). In all
cases, ε Pg values decreased with increasing Qabs. Given saturation of Pg and the definition of ε Pg , it is not surprising that daily
ε Pg followed a hyperbolic relationship (r 2 = 0.74, broken line in
Figure 2b). Estimated Qsat was 4.66 MJ day – 1, which is somewhat lower than the Qsat value for half-hourly Pg. The estimate
for ε Pg was 1.57 g C MJ – 1. These daily values illustrate the
max
Figure 2. Measured (a) daily and (c)
monthly mean daily gross primary production (Pg ) versus simulated daily absorbed
Q (Q abs ), and (b) daily and (d) monthly
Q-use efficiency of Pg (εPg ) versus simulated daily Q abs. The broken line in (b) is a
fit of Equation 6 to the data. The solid
lines are linear regression lines.
TREE PHYSIOLOGY VOLUME 28, 2008
LIGHT-USE EFFICIENCY IN A TROPICAL FOREST
Figure 3. Seasonal course of measured gross primary production (Pg ),
(stars and broken line), and predictions of four regression approaches:
pure Q-response approach (䉮,䉲) and a combined Q- and water-vapor-saturation-deficit-response approach (䉫,䉬). Open symbols represent monthly parameters, solid symbols represent predictions based
on a single parameter set for the whole period as presented in Figure 1.
The lower part shows monthly precipitation sums.
hyperbolic nature of the relationship between ε Pg and Qabs
more clearly, but the order of magnitude and the slope
(–0.00214 g C MJ – 2 day –1; r 2 = 0.66) of the monthly values
505
show that they represent the same relationship as the daily values. Because of the simultaneous occurrence of high Qabs and
high D, monthly ε Pg was also linearly related to daytime D, but
this relationship was weaker (r 2 = 0.43) than the ε Pg –Qabs relationship. These single effects cannot be discriminated from the
empirical data at the monthly timescale. The half-hourly data
suggest that both effects take place concurrently.
From the foregoing, it is apparent that ε Pg has a diurnal pattern (lower panel of Figure 4). During the early morning and
late afternoon ε Pg was nearly constant and maximal. The few
larger values during the night–day and day–night transitions
are probably artefacts caused by imperfections of the radiation
sensors. During the rest of the daytime, ε Pg followed a parabolic temporal course according to the hyperbolic relationship
with Qabs, and including a relatively small D effect that caused
the ε Pg curve to be slightly skewed in the early afternoon. The
ε Pg minimum thus occurred about 1 h later than the mean Qabs
maximum.
These analyses require strict synchronization of the CO2
flux and radiation measurements. In our study, these measurements were taken with different data logging systems and thus
a lag of one or two minutes between instruments may have occurred. To achieve equal ε Pg values for Qabs < Qsat, the original
Pg data have been shifted by one minute against the radiation
simulations.
Figure 4. (a) Mean diurnal courses of absorbed Q (Q abs ) or predicted saturating
Q abs (Q sat ) water vapor pressure saturation
deficits (D), (b) measured gross primary
production rates (Pg ) and (c) Q use efficiencies of Pg (εPg ). The error bars represent standard deviations.
TREE PHYSIOLOGY ONLINE at http://heronpublishing.com
506
IBROM ET AL.
Discussion
Half-hourly time scale
We demonstrated that current ε-approaches, which are normally used at daily and longer time scales, cannot be used to
predict half-hourly Pg values accurately from Qabs at the study
site with a single and constant ε Pg . Our analysis showed that Pg
saturation occurs at the daily time scale, as also noted in other
studies, and is even indicated at the monthly time scale. The
ε-approach assumes that the effect of saturation becomes negligible when averaging data over longer intervals. Larger temporal averaging can, however, only mitigate the unfavorable
effects of saturation, if conditions with Pg saturation are rare
compared with unsaturated conditions. Figures 1 and 2 demonstrate that this was not the case at our site. The median daytime Qabs was 683 µmol m – 2 s –1 and the mean was 783 µmol
m – 2 s –1, both of which were close to the observed Q for Pg saturation.
The ε Pg value can be calculated easily from Pg and Qabs at
any time scale, but our findings indicate that ε Pg lacks prognostic value, as assumed by ε-approaches. If Qabs changes from
year to year or from season to season, predictions of the effect
of altered Qabs on Pg cannot be made without testing whether
ε Pg changes with the new conditions.
Daily and monthly time scales
The remote sensing community expresses Q-use efficiency in
units of g C MJ –1. The mean value for ε Pg at the study site over
the observed 16 months was 0.99 g C MJ –1, and ranged from
0.77 g C MJ –1 during drier seasons to 1.10 g C MJ –1 during the
rainy season. The maximum value of ε Pg for evergreen broadleaved forests, which is used in the computation of the current
MODIS products (MOD17) at the global scale is 1.16 g C
MJ –1 (Table 2.1 in Heinsch et al. 2003). Based on measured D
and reduction function of Heinsch et al. (2003), ε Pg is reduced
a little to 1.15 g C MJ –1. Both values are of the same order of
magnitude as those we observed, however, and if used to calculate Pg for the investigated rain forest site, Pg would be overestimated by 16%, which would lead to a substantial overestimation of net CO2 flux, because this is relatively small compared with Pg.
Ito and Oikawa (2004) reported ε Pg values between 0.9 and
1.2 g C MJ –1, a range that embraces both the higher value obtained by Heinsch et al. (2003) and our value. Turner (2006a)
used a model to simulate ε Pg for a tropical rain forest site in
Brazil (Tapajos; see Saleska et al. 2003) and found that the estimates varied seasonally around mean values that were generally larger than the maximum ε Pg value reported by Heinsch et
al. (2003) and which is currently used for the MOD17 calculations.
Daily ε Pg varied more than the long-term values discussed
above. Turner et al. (2003) showed that values from four sites
varied from somewhat lower than 1.4 g C MJ –1 to more than
4 g C MJ –1, about twice the maxima we found. Their study is
directly comparable with our study, because it used the same
flux estimation method, eddy correlation or eddy accumulation, and also a similar approach to calculate Q absorption. Al-
though the four investigated sites were all from temperate or
boreal zones, the ranges of daily Qabs in the Turner et al. (2003)
study were similar to those in our study (from about 1 to 14 MJ
m – 2 day – 1). However, these daily sums occurred during a
shorter photoperiod at the tropical site and thus reflected
higher Q at the half-hourly time scale compared with the vegetation periods in temperate and boreal climates. Even under
non-tropical conditions, Turner et al. (2003) observed saturation of daily Pg with Qabs in three out of four investigated sites
(two forests and one agricultural site), whereas counter intuitively the site with the lowest LAI did not show saturation.
The observed tight hyperbolic relationship between ε Pg and
Qabs at the half-hourly time scale (Figure 1b) cannot easily be
compensated by time averaging on a daily time scale. Saturation does not lead to a constant ε Pg as assumed with the common ε-approaches. However, because the seasonal changes in
Qabs are relatively small at the tropical rain forest site, the ε-approaches yield correct results—provided the choice of a longterm value of ε Pg is appropriate. Neglecting saturation will result in overestimation of Pg at high Qabs and to underestimation
at low Qabs. Using an appropriate long-term value of ε Pg might
lead to an accurate long-term estimate of Pg even if the seasonal or interannual variability is potentially biased. The problem remains how to determine the appropriate long-term ε Pg ,
because the mean distribution of hours with Q-saturated Pg
needs to be known.
The strength of the ε-approach is its simplicity and the use
of a linear relationship between Pg and Qabs. Our results as well
as results from other investigations have shown that this relationship is not general. Saturation of Pg can be introduced in
the ε-approach as described in Equation 6. However, using this
formulation to calculate Pg is an unnecessary complication
compared with calculating it directly with a simplified Qabs-response function, such as:
⎧ ε P f ( D)Qabs
Pg = ⎨ g max
⎩ Pg sat f ( D )
Qabs ≤ Qsat
Qabs > Qsat
(7)
where Pg sat is the saturated rate of Pg, i.e., Pg sat = ε Pg Qsat .
max
Equation 7 provides a simple and more general representation
of the Qabs and D responses of Pg and can be used to replace the
current ε-approach.
However, Equation 7 neglects the problem that seasonality
of Pg might not be driven by Qabs, or D but by other factors. The
monthly regressions showed that it is variability in Pmax that
drives seasonality in Pg at the study site. Potential drivers that
might affect seasonality, e.g., soil water availability, or the way
radiation is distributed and used in canopies as a consequence
of differing proportions of beam and diffuse radiation, should
be investigated to define the value of light-saturated Pg, the
value of Qsat and the ε Pg of the linear part of the Q-response
function of Pg. We advocate replacing the ε-approach by a
more general Q-response approach that includes common response features of vegetation canopies to environmental conditions.
TREE PHYSIOLOGY VOLUME 28, 2008
LIGHT-USE EFFICIENCY IN A TROPICAL FOREST
The parameters of such Q-response functions can be retrieved from measurements, but as we show here, they cannot
easily be extrapolated from one vegetation type to another. A
feasible and reasonable alternative is to use existing prognostic canopy models to calculate the parameters of the Q-response to areas and times where measurements are unavailable, as shown by Turner et al. (2006b) at the regional scale.
In conclusion, detailed analysis of the photosynthetic light
response showed that current ε-approaches cannot be used to
predict seasonality and interannual variability in Pg of the investigated tropical rain forest. Short-term variability in Pg
(~30 min) could be predicted from Qabs and D. It was shown
that Pg saturated at Q values that were in many cases exceeded
at the site. Our empirical analysis also indicated that Qsat decreased at high D. Both effects led to frequent saturation of Pg
which contradicts the assumption of the current ε-approaches
that ε Pg is independent of Qabs. We suggest replacing the linear
ε-approach by a Blackman-type Q response function that can
be adapted to Pg saturating weather conditions. It can be assumed that the saturation threshold depends on LAI and leaf
physiology and thus, direct extrapolation of the relationship to
canopies with differing absorbances will most probably yield
unrealistic results. In these cases, prognostic models should be
used that are able to predict the parameters of the Q-response
functions from available stand information for typical situations and canopies in certain biomes. Future research will focus on the determinants of saturated Pg, Qsat and the linear relationship at Qabs < Qsat.
Acknowledgments
This study was finacially supported by the German Research Foundation (DFG) within the framework of the interdisciplinary research
program “Stability of Rain forest Margins in Indonesia” (STORMA,
SFB 552). AI was funded by a Marie-Curie Intra European Fellowship (BCP-TEF, Proposal no. 008354). We are grateful for the help of
staff from the Tadulaku University in Palu, especially Pa M.Sc. Ir. Abdul Rauf, during the establishment of the flux measurements at the
Bariri site. We thank Leon Gareth Linden, Risø (DTU), Denmark, for
discussing the manuscript.
References
Aubinet, M., A. Grelle, A. Ibrom et al. 2000. Estimates of the annual
net carbon and water exchange of European forests: the EUROFLUX Methodology. Adv. Ecol. Res. 30:113–175.
Baldocchi, D., E. Falge, G. Liahong et al. 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. 82:2415–2435.
Blackman, F.F. 1905. Optima and limiting factors. Ann Bot. 19:
281–295.
Choudhury, B.J. 2001. Estimating gross photosynthesis using satellite
and ancillary data: approach and preliminary results. Remote Sensing Environ. 75:1–25.
Cramer, W., D.W. Kicklighter, A. Bondeau et al. 1999. Comparing
global models of terrestrial net primary productivity (NPP): overview and key results. Global Change Biol. 5:1–15.
Dai, Y., R.E. Dickinson and Y.P. Wang. 2004. A two big-leaf model
for canopy temperature, photosynthesis and stomatal conductance.
J. Climate 17:2281–2299.
507
Dickinson, R.E. 1983. Land surface processes and climate-surface
albedos and energy balance. Adv. Geophys. 25:305–353.
Falk, U. 2004. Turbulent fluxes of CO2, H2O and energy in the atmospheric boundary layer above tropical vegetation investigated by
eddy-covariance measurements. Institute Bioclimatology, GeorgAugust University, Göttingen, 116 p.
Field, C.B., J.T. Randerson and C.M. Malmstroem. 1995. Global net
primary production: combining ecology and remote sensing. Remote Sensing Environ. 51:74–88.
Goulden, M.L., J.W. Munger, S.-M. Fan, C.D. Daube and S.C. Wofsy.
1996. Exchange of carbon dioxide by a deciduous forest: response
to interannual climate variability. Science 271:1576–1578.
Grace, J. 2004. Understanding and managing the global carbon cycle.
J. Ecol. 92:189–202.
Grace, J., J. Lloyd, J. McIntyre et al. 1995. Carbon dioxide uptake by
an undisturbed tropical rain forest in southwest Amazonia, 1992 to
1993. Science 270:778–780.
Gravenhorst, G., Y. Knyazikhin, J. Kranigk, G. Mießn, O. Panfyorov
and K.-G. Schnitzler. 1999. Is forest albedo measured correctly?
Meteorol. Zeitschrift. 8:107–114.
Heinsch, F.A., M. Reeves, P. Votava et al. 2003. User’s guide GPP and
NPP (MOD17A2/A3) products NASA MODIS land algorithm.
NASA Goddard Space Flight Center, Greenbelt, MD, 57 p.
Heinsch, F.A., M. Zhao, S.W. Running et al. 2006. Evaluation of remote sensing based terrestrial productivity from MODIS using
AmeriFlux tower eddy flux network observations. IEEE Trans.
Geosci. Remote Sensing 44:1908–1925.
Højstrup, J. 1993. A statistical data screening procedure. Sci.
Technol. 4:153–157.
Hollinger, D.Y., F.M. Kelliher, J.N. Byers, J.E. Hunt, T.M. McSeveny
and P.L. Weir. 1994. Carbon dioxide exchange between an undisturbed old-growth temperate forest and the atmosphere. Ecology
75:134–150.
Ibrom, A. 2001. Die biophysikalische Steuerung der Kohlenstoffbilanz in einem Fichtenbestand im Solling. Ber. FZW Reihe B.
173:209.
Ibrom, A., P.G. Jarvis, R.B. Clement et al. 2006. A comparative analysis of simulated and observed photosynthetic CO2 uptake in two
coniferous forest canopies. Tree Physiol. 26:845–864.
Ibrom, A., A. Oltchev, T. June et al. 2007. Effects of land use change
on matter and energy exchange between a tropical rain forest margin and the atmosphere. In The Stability of Tropical Rain forest
Margins: Linking Ecological, Economic and Social Constraints
Eds. T. Tscharntke, C. Leuschner, M. Zeller, E. Guhardja and
A. Bidin. Springer-Verlag, Berlin, pp 463–492.
Ito, A. and T. Oikawa. 2004. Global mapping of terrestrial primary
productivity and light-use efficiency with a process-based model.
In Global Environmental Change in the Ocean and on Land.
Ed. M. Shiyomi. Terrapublishing, Tokyo, pp 343–358.
Jarvis, P.G., J.M. Massheder, S.E. Hale, J.B. Moncrieff, M. Rayment
and S.L. Scott. 1997. Seasonal variation of carbon dioxide, water
vapour and energy exchanges of a boreal black spruce forest.
J. Geophys. Res. 102:28,923–28,966.
Knyazikhin, Y., G. Miessen, O. Panfyorov and G. Gravenhorst. 1997.
Small-scale study of three dimensional distribution of photosynthetically active radiation in a forest. Agric. For. Meteorol. 88:
215–239.
Knyazikhin, Y., J.V. Martonchik, D.J. Diner, R. Myneni, M. Verstraete, B. Pinty and N. Gobron. 1998. Estimation of vegetation
canopy leaf area index and fraction of absorbed photosynthetically active radiation from atmosphere-corrected MISR data.
J. Geophys. Res. 103, No. D24:32,239–32,256.
TREE PHYSIOLOGY ONLINE at http://heronpublishing.com
508
IBROM ET AL.
Lee, X., W. Massman and B. Law. 2004. Handbook of Micrometeorology: a guide for surface flux measurements. In Atmospheric and Oceanographic Sciences Library. Eds. L.A. Mysak and
K. Hamilton. Kluwer Academic Publishers, Dordrecht, 250 p.
Loescher, H., S.H. Oberbauer, H.L. Gholz and D.A. Clark. 2003. Environmental control on net ecosystem-level carbon exchange and
productivity in a Central American tropical wet forest. Global
Change Biol. 9:396–412.
Miller, S.D., M.L. Goulden, M.C. Menton, H.R. da Rocha, H.C. de
Freitas, A.M.S. Figueira and C.A. Dias de Sousa. 2004. Biometric
and micrometeorological measurements of tropical forest Carbon
balance. Ecol. Appl. 14:114–126.
Moncrieff, J.B., Y. Malhi and R.L. Leuning. 1996. The propagation of
errors in long-term measurements of carbon and water. Global
Change Biol. 2:231–240.
Monteith, J.L. 1977. Climate and the efficiency of crop production in
Britain. Philos. Trans. R. Soc. Lond. B 281:277–294.
Moore, C.J. 1986. Frequency response corrections for eddy correlation systems. Boundary-Layer Meteorol. 37:17–35.
Morgenstern, K. 2000. Turbulent CO2-, H2O- and energy fluxes
above a Mediterranean oak and a mountainous spruce forest investigated by eddy-covariance measurements. Institut Bioklimatologie, Georg-Augusty University Göttingen, Göttingen, 104 p.
Oltchev, A., J. Èermak, N. Nadezhdina, F. Tatarinov, A. Tishenko,
A. Ibrom and G. Gravenhorst. 2002. Transpiration of a mixed forest stand: field measurements and simulation using SVAT models.
Boreal Environ. Res. 7:389–397.
Oltchev, A., A. Ibrom, T. Ross et al. 2008. A modeling approach for
simulation of water and carbon dioxide exchange between multispecies tropical rain forest and the atmosphere. Ecol. Model.
doi:10.1016/j.ecolmodel.2007.10.021.
Pilegaard, K., P. Hummelshøj, N.O. Jensen and Z. Chen. 2001. Two
years of continuous CO2 eddy-flux measurements over a Danish
beech forest. Agric. For. Meteorol. 107:29–41.
Pinty, B., T. , T. Lavergne, R.E. Dickinson, J.-L. Widlowski, N. Gobron and M.M. Verstraete. 2006. Simplifying the interaction of land
surfaces with radiation for relating remote sensing products to climate models. J. Geophys. Res. 111:D02116.
Rannik, Ü. and T. Vesala. 1999. Autoregressive filtering versus linear
detrending in estimation of fluxes by the eddy covariance method.
Boundary-Layer Meteorol. 91:259–280.
Richardson, A.D., D.Y. Hollinger, G.G. Burba et al. 2006. A multisite analysis of random error in tower-based measurements of carbon and energy fluxes. Agric. For. Meteorol. 136:1–18.
Ross, J. 1981. The radiation regime and architecture of plant stands.
Dr. W. Junk Publ., The Hague, 391 p.
Ruimy, A., G. Dedieu and B. Saugier. 1996. TURC: a diagnostic
model of continental gross primary productivity and net primary
productivity. Global Biogeochem. Cycles 10:269–285.
Ruimy, A., L. Kergoat, A. Bondeau and Participants of Potsdam NPP
Model Intercomparison. 1999. Comparing global models of terrestrial net primary productivity (NPP): analysis of differences in
light absorption and light-use efficiency. Global Change Biol.
5:56–64.
Running, S.W., P.E. Thornton, R. Nemani and J.M. Glassy. 2000.
Global terrestrial gross and net primary productivity from the Earth
Observing System. In Methods in Ecosystem Science. Eds.
O.E. Sala, R.B. Jackson, H.A. Mooney and R.W. Howarth.
Springer-Verlag, New York, pp 44–57.
Running, S.W., R.R. Nemani, F.A. Heinsch, M. Zhao, M. Reeves and
H. Hashimoto. 2004. A continuous satellite-derived measure of
global terrestrial production. BioScience 54:547–560.
Saleska, S.R., S.D. Miller, D.M. Matross et al. 2003. Carbon in amazon forests: unexpected seasonal fluxes and disturbance-induced
losses. Science 302:1554–1557.
Still, C.J., J.T. Randerson and I.Y. Fung. 2004. Large-scale plant
light-use efficiency inferred from the seasonal cycle of atmospheric
CO2. Global Change Biol. 10:1240–1252.
Thornley, J.H.M. 1976. Mathematical models in plant physiology.
Academic Press, New York, 318 p.
Turner, D.P., S. Urbanski, D. Bremer, S.C. Wofsy, T. Meayers,
S.T. Gower and M. Gregory. 2003. A cross-biome comparison of
daily light use efficiency for gross primary production. Global
Change Biol. 9:383–395.
Turner, D.P., W.D. Ritts, W.B. Cohen et al. 2006a. Evaluation of
MODIS NPP and GPP products across multiple biomes. Remote
Sensing Environ. 102:282–292.
Turner, D.P., W.D. Ritts, J.M. Styles, Z. Yang, W.B. Cohen, B.E. Law
and P.E. Thornton. 2006b. A diagnostic carbon flux model to monitor the effects of disturbance and interannual variation in climate on
regional NEP. Tellus B. 58:476–490.
Valentini, R., P. De Angelis, G. Matteucci, R. Monaco, S. Dore and
G.E. Scarascia Mugnosa. 1996. Seasonal net carbon dioxide exchange of a beech forest with the atmosphere. Global Change Biol.
2:199–207.
Valentini, R., G. Matteucci, A.J. Dolman et al. 2000. Respiration as
the main determinant of carbon balance in European Forests. Nature 404:861–865.
Wang, Y.P. and P.G. Jarvis. 1990. Description and validation of an array model - MAESTRO. Agric. For. Meteorol. 51:257–280.
Wang, Q., Y. Kakubari, M. Kubota and J. Tenhunen. 2007. Variation
on PAR to global solar radiation ratio along altitude gradient in
Naeba Mountain. Theor. Appl. Climatol. 87:239-235.
Webb, E.K., G.I. Pearman and R. Leuning. 1980. Correction of flux
measurements for density effects due to heat and water vapour
transfer. Quart. J. Roy. Meteorol Soc. 106:85–100.
Wesely, M.L. and R.L. Hart. 1985. Variability of short term eddy-correlation estimates of mass exchange. In The Forest–Atmosphere
Interaction. Eds. B.A. Hutchinson and B.B. Hicks. D. Reidel,
Dordrecht, pp 591–612.
Xiao, X., D. Hollinger, J. Aber, M. Goltz, E. Davidson, Q. Zhang and
B. Moore. 2004. Satellite-based modeling of gross primary production in an evergreen needle-leaf forest. Remote Sensing Environ.
89:519–534.
Zhao, M., F.A. Heinsch, R.R. Nemani and S.W. Running. 2005.
Improvements of the MODIS terrestrial gross and net primary production global data set. Remote Sensing Environ. 95:164–176.
TREE PHYSIOLOGY VOLUME 28, 2008