CO 2 uptake of a mature Acacia mangium plantation estimated from sap flow measurements and stable carbon isotope discrimination

A simple, nondestructive method for the estimation of canopy CO2 uptake is important for understanding the CO2 exchange between forest and atmosphere. Canopy CO2 uptake ( FCO2) of a subtropical mature A. mangiumplantation was estimated by combining sap flow measurements and stable carbon isotope discrimination ( 1) in Southern China from 2004 to 2007. The mechanistic relationship linking FCO2, 1 in leaf sap, and sap flow-based canopy stomatal conductance ( Gs) was applied in our study. No significant seasonal variations were observed in 1 or in the ratio of the intercellular and ambient CO 2 concentrations (Ci/Ca), although diurnalCi/Ca varied between sunlit and shaded leaves. A sensitivity analysis showed that estimates of FCO2 were more sensitive to dynamics in Gs than inCa and 1. By using seasonally and canopy averaged Ci/Ca values, we obtained an acceptable estimate of FCO2 compared to other estimates. FCO2 exhibited similar diurnal variation to that ofGs. Large seasonal variation in FCO2 was attributed to the responsiveness of Gs to vapor pressure deficit, photosynthetically active radiation, and soil moisture deficit. Our estimate of FCO2 for a matureA. mangium plantation (2.13± 0.40 gC m−2 d−1) approached the lower range of values for subtropical mixed forests, probably due to lower mean canopy stomatal conductance, higher Ci/Ca, and greater tree height than other measured forests. Our estimate was also lower than values determined by satellitebased modeling or carbon allocation studies, suggesting the necessity of stand level flux data for verification. Qualitatively, the sap flux/stable isotope results compared well with gas exchange results. Differences in results between the two approaches likely reflected variability due to leaf position and age, which should be reduced for the combined sap flux and isotope technique, as it uses canopy average values of Gs andCi/Ca.


Introduction
The continued increase in atmospheric carbon dioxide levels due to anthropogenic emissions has led to significant climate changes (Schneider, 1989;Pachauri and Reisinger, 2007).Rising atmospheric CO 2 is due to the imbalance between the rates that sources emit CO 2 into the atmosphere and the rates that sinks remove CO 2 from the atmosphere (Baldocchi et al., 1996).In this context, forests act as vital CO 2 sinks by storing carbon in woody biomass (Nowak et al., 2002).It has been reported that forests at northern midlatitude sites are large CO 2 sinks (Ciais et al., 1995).A recent study suggests that old-growth forests continue to accumulate carbon (Luyssaert et al., 2008).In China, since the mid-1970s, planted forests (afforestation and reforestation) have sequestered 0.45 petagram of carbon, while natural forests have lost an additional 0.14 petagram of carbon (Fang and Chen, 2001).The biomass increments vary with stand age and many other factors, primarily resource availability (Oren et al., 2001;Peng, 2003).However, estimation of canopy scale photosynthesis has not been common.Such uncertainties hinder our ability to estimate forest CO 2 sequestration for global carbon budgets, as well as to increase CO 2 sequestration through forest management.
The tree canopy scale is an important intermediate scale between the leaf and the ecosystem.Although established photosynthesis monitoring techniques are available for both leaf and ecosystem scales (Farquhar et al., 1980;Baldocchi et al., 1996), there are currently few accepted estimation methods for photosynthesis at the tree canopy scale.This limits our understanding of canopy photosynthesis compared with canopy water fluxes, which can be determined through sap flow techniques.Notably, comparisons between water and photosynthesis fluxes must account for the fact that water fluxes are generally much larger than CO 2 fluxes.A number of methods have been designed to address the dilemma, including (1) directly adopting branch bag techniques, or combining water use efficiency obtained with branch bag techniques with whole-tree transpiration (Rayment and Jarvis, 1999;Morén et al., 2001), (2) coupling sap flow-based canopy stomatal conductance and photosynthesis (Catovsky et al., 2002), and (3) using a canopy conductance constrained carbon assimilation scheme, which couples actual or potential canopy conductance with vertical gradients of light distribution, leaf-level conductance, maximum Rubisco capacity and maximum electron transport (Schäfer et al., 2003;Kim et al., 2008).However, application of these methods still requires extensive leaf gas exchange measurements, which are time-consuming and difficult, because access to tall canopies is required.Recently, a novel approach combining sap flow measurement and stable carbon isotope techniques (the SF/SI approach) was proposed to estimate forest CO 2 uptake (Zhao et al., 2005a;Hu et al., 2010).Firstly, it is ideal for plot-scale studies, since measurements are taken at the whole tree scale, rather than scaled up or down from the leaf or the ecosystem level (Zhao et al., 2005a).Secondly, it is suitable for field studies of forests in mountainous areas, especially for upper-canopy species such as the pure A. mangium stand in our study.This overcomes the restrictions of the eddy covariance method (EC) associated with surface heterogeneity and complex terrain (Moncrieff et al., 1996).Lastly, it has huge potential given the large number of sites with continuous sap flow measurement and the relatively easy sampling for carbon isotope analysis.
Previous application of the SF/SI approach has accounted for both within-canopy and seasonal variability to estimate correctly CO 2 uptake in needle leaf trees in a subalpine forest (Hu et al., 2010).Hu et al. (2010) tested for differences in the carbon isotope ratio of needle sugars and water-use efficiency between sun and shade needles, throughout the growing season, and found observed the highest values of wholetree CO 2 assimilation rates following snowmelt or summer rain events, and during weather with lower temperatures.In our study, we accounted for sub-daily and inter-annual variability, as well as within-canopy and seasonal variability for a broadleaf species in a subtropical climate.The instantaneous forest CO 2 uptake rate can be estimated based on the relationship between canopy stomatal conductance (G s ) and 13 C discrimination ( ) in leaf sap.Our SF/SI approach is unique in its continuity, representativeness, accuracy and ease of use: G s is derived from continuous measurement of sap flow (Zhao et al., 2005a); the relationship of net photosynthetic rate/canopy stomatal conductance (A net /G s ) derived from is very stable and representative on both temporal and spatial scales (Dawson et al, 2002); and sap flow measurement is nondestructive, easy, and relatively low cost (Granier, 1987;Zhao et al., 2005b).Given these features, we demonstrate in this paper that with appropriate sampling, CO 2 uptake obtained from the SF/SI approach is representative at the tree canopy scale, and is comparable with results from other methods, including the gas exchange method, carbon allocation studies, and modeling approaches.Hereby, the stem sap flux, carbon isotopic compositions of leaf sap, and environmental parameters in a subtropical A. mangium plantation in Southern China were measured from 2004 to 2007.Our aims were: (1) to provide an alternative sap flow-based method for estimating instantaneous CO 2 uptake at the canopy scale and verify it by comparing it with values obtained from other methods, (2) to investigate the diurnal, seasonal, annual and inter-annual changes of CO 2 uptake of the plantation and (3) to examine how CO 2 uptake of the plantation is regulated by stomatal conductance and environmental conditions.We expected that F CO 2 of a subtropical mature A. mangium plantation could be estimated by the SF/SI approach, by accounting for variability of sunlit vs. shaded leaves.Given that this forest is likely to be water limited, we also expected that F CO 2 would be higher in the wet season than in the dry season, and be highest in the mid-morning.Because CO 2 is assimilated through the stomatal pores, we expected that F CO 2 would be more sensitive to changes in G s than in C a and .

Site description and environmental conditions
This study was conducted at the Heshan Hilly Land Interdisciplinary Experimental Station of the Chinese Academy of Sciences ( 112• 54 E, 22 • 41 N) in Guangdong Province, China over the 2004-2007 period.The study site is dominated by a subtropical monsoonal climate, with the wet season occurring from April to September and the dry season from October to March.Mean annual temperature is 21.7 • C and average precipitation is 1400 mm, of which 78 % occurs in the wet season.
The study stand was situated at 60-70 m altitude, on a 20 • slope with a south-east aspect.The soil is an oxisol developed from sandstone with surface soil to a depth of 10 cm and subsoil for a further 50 cm (Ma et al., 2008a).Acacia mangium was the only overstory species planted at a spacing of 3 m × 4 m in 1984, and reached maturity after ∼ 20 years of growth.The understory vegetation was very sparse and its contribution to stand CO 2 uptake was considered to be negligible for the purposes of this study.fast-growing legume species that has nitrogen fixing nodules on its shallow roots (Cole et al., 1996).It was introduced to China from Southeast Asia in the late 1970s and has been widely used for restoration of vegetation in tropical and subtropical China (Yang et al., 2009).A study plot with an area of 640.5 m 2 (36.6 m × 17.5 m) was set up within the stand.The plot comprised 47 trees (734 trees per hectare) with a basal area of 26.6 m −2 ha −1 .These trees included both trees originally planted in 1984 and naturally regenerated trees.Tree height ranged from 2.4 m to 22.8 m with an average of 15.7 m.Stem diameter at breast height (DBH) was between 3.8 cm and 37.5 cm with an average of 20.1 cm.
Air temperature (T a , • C), air relative humidity (RH, %), and photosynthetically active radiation (PAR, µmol m −2 s −1 ), were monitored at a weather station located 150 m away from the study site.T a and RH were measured using an HMP35E sensor (HMP35E; Vaisala, Finland).PAR was measured with a LI-190SA quantum sensor (Li-Cor, Lincoln, US).Since the root system of A. mangium was mainly distributed in the upper 30 cm of the soil, volumetric soil water content (SWC, m 3 m −3 ) was measured using three soil moisture probes (ML2x, Delta-T Device, UK) inserted into the upper soil (20 cm-30 cm) amongst the selected trees.The environmental data were sampled and recorded as 10 min means by a data logger synchronized to the logger for sap flow measurements.Vapor pressure deficit (D) was calculated from T a and RH as follows: where a, b, and c are fixed parameters, which are 0.611 kPa, 17.502, and 240.97 • C (Campbell and Norman, 1998).SWC data were converted to soil moisture deficit (SMD, dimensionless), defined as where SWC max and SWC min are the maximum and minimum values of SWC during the study period.SWC max (0.49) occurred on 18 July 2005, while SWC min (0.11) occurred on 26 December 2005.

Projected crown area, leaf area index and sapwood area
Sapwood area (A s ), projected crown area (A c ), and leaf area index (LAI) are key parameters for calculating stand transpiration and canopy stomatal conductance.A c of each tree in the plot was calculated as the area of an ellipse based on measurements of the widest and narrowest canopy widths.The projected crown area (A c ) of each tree in the plot was estimated to range from 0.8 m 2 to 53.0 m 2 , with an average of 13.8 m 2 .LAI was measured at 20 random locations within the stand using a plant canopy analyzer (CID-110, CID COR., USA) at dawn, dusk or under cloudy conditions every month.A s of each tree in the plot was calculated based on strong allometric relationships between DBH and A s (A s = 0.1930 (DBH) 1.944 ).This relationship, established from measurements on 23 surrounding trees, has been reported in previous studies (Zhao et al., 2005b;Wang et al., 2012).

Sap flow measurement
Sap flux density (J s , g H 2 O cm −2 s −1 ) was measured with self-made Granier-type sensors (Granier, 1987;Zhao et al., 2005b).Fourteen trees, representative of the DBH distribution in the plot, were selected for sap flow measurement, with tree height ranging from 12 m to 22.8 m, DBH between 13.4 and 37.5 cm and A c ranging from 4.6 m 2 to 47.7 m 2 (see Supplement A for details).1, 4, 5, 2, 2 trees at the < 15 cm, 15-20 cm, 20-25 cm, 25-30 cm, and > 30 cm DBH classes were selected for J s measurement, respectively.For tree nos.1-4, sensors were installed on the eastern, southern, western, and northern sides of the trunk, while for tree nos.5-14, sensors were installed only on the northern side.The sensors and adjacent portions of stem were wrapped with plastic insulation to protect the probes from mechanical damage, and the entire assembly was enclosed in an outer layer of aluminum film in order to minimize spurious temperature gradients caused by radiant heating of the stem, as well as to protect against water running down the trunk (Zhao et al. 2005b).To ensure good measurements, new sets of sensors were installed 2-5 cm from the original holes at both the beginning and end of the dry season (Ma et al., 2008a).Sensors consisted of two 20 mm-long probes inserted into the xylem at breast height, with one placed 10 cm above the other.Each probe contained a copper-constantan thermocouple, and the upper probe was continuously heated with heating wire supplied with constant power.The temperature difference between the two probes was taken every 30 s and recorded as 10 min means by a data logger (Delta-T Devices Ltd., Cambridge, UK).We selected the highest temperature difference between heated and unheated probes during times of zero flux ( T m ) as the baseline for each day using the Baseliner program developed by Yavor Parashkevov from Duke University (Granier, 1987;Phillips et al., 1997).Deviation from this baseline was used to estimate J s (Granier, 1987).It was found that the nighttime sap flow of the same Acacia mangium plantation was mainly used to refill water in the trunk, rather than nighttime transpiration (Wang et al., 2012).Hereby, using the maximum temperature difference for each night did not lead to underestimation of sap flow.

Stand transpiration and canopy stomatal conductance
Whole-tree transpiration (E t , g s −1 ) was calculated as A. mangium is a diffuse-porous species for which the radial variation in J s is considered to be small (Phillips et al., 1996;Ma et al., 2008a), and sapwood depth of sample trees measured with cores was close to 2 cm.Therefore, radial variation in J s was considered to be negligible in this study.Furthermore, the variation in J s at different aspects was random for the studied trees (Ma et al., 2007).Hence, J s at the north aspect was assumed to be representative of J s of individual trees.For tree nos.1-4, J s was calculated as the mean of the values from four aspects, while for tree nos.5-14, J s was the value measured from the northern side.
The 47 trees within the stand were classified into 5 groups by DBH categories (Zhao et al., 2006) (see Supplement B for details).The average sap flux density of class i (J si ) was calculated as: where J sj is the sap flux density of sample tree j , and A sj is the sapwood area of sample tree j .Sap flow measurements on 14 sample trees were extrapolated to stand level as in Zhao et al. (2005a): where E s is the stand transpiration of A. mangium (g s −1 ), and A si is the sum of total sapwood area in class i. E s was converted to stand transpiration per unit of leaf area (E L , mm s −1 ) as in Zhao et al. (2005a): where LAI is the leaf area index (1.95m 2 m −2 ) and A G is the total stand area (640.5 m 2 ).
For our stand, a strong coupling of canopy surface to the atmosphere can be assumed.This assumption requires three conditions to be met.Firstly, the A. mangium plantation had a low LAI, and thus an open, well-ventilated, aerodynamically rough canopy.Secondly, the wind speed (w) had an effect on canopy stomatal conductance (G s ) of the A. mangium plantation.G s significantly decreased with w in July and December from 2005 to 2007 (p < 0.05; R 2 = 0.106; G s = −7.3437w + 62.548).However, most of the time, the daily average w above the plantation was greater than or close to 2 m s −1 from 2005 to 2007.Thus, the boundary layer conductance was much higher than canopy stomatal conductance (Larcher, 1983).Finally, the decoupling coefficient for our stand was less than 0.2 in 2005 (Ma, 2008).Furthermore, since the alteration of the transpiration signal due to depletion and replenishment of stem-stored water was relatively small (Wang et al., 2012), the average sap flux in the sapwood multiplied by sapwood area: leaf area is equal to transpiration (E L ).Therefore, the mean canopy stomatal conductance (G s , mmol H 2 O m −2 s −1 ) of the study stand was calculated based on a simplification of the Penman-Monteith equation (Köstner et al., 1992) where ρ is the density of water (998 kg m −3 ), G V is the universal gas constant adjusted for water vapor (0.462 m 3 kPa K −1 kg −1 ), T a is the air temperature ( • C), and D is vapor pressure deficit (kPa).We used all data except for those that were excluded based on the following criteria: (1) data during and 2 h after rainfall were excluded to avoid the discrepancy between evaporation and tree transpiration (Granier et al., 2000); (2) when global radiation, vapor pressure deficit, or stand transpiration were too low (< 5 % of the maximum value), because of the large relative uncertainties in computing G s under these conditions (Granier et al., 2000); (3) when D < 0.6 kPa for G s estimation, to keep errors in G s estimates to less than 10 % (Ewers and Oren, 2000).
Relationships between the ratio of canopy stomatal conductance to the maximum value (G s /G smax ) and photosynthetically active radiation (PAR), vapor pressure deficit (D), and soil moisture deficit (SMD) were determined by boundary-line analysis and non-linear least squares, using the maximum of average G s of 14 trees for different PAR (step width: 50 µmol m −2 s −1 ), D (step width: 0.2 kPa), and SMD classes (step width: 0.1).The estimated boundary line was used to compare the responses among seasons or years (Webb, 1972;Rico et al., 1996).
The response of G s /G smax to PAR was represented by the Michaelis-Menten quadratic hyperbolic function (Thornley and Johnson, 1990) where a is a fitted parameter, representing the value of global radiation that reduces canopy conductance to one half of its maximum value.G s was related to D within seasons and among years in order to estimate the sensitivity of stomata to D. The response function was described as (Oren et al., 1999) where the parameter -m, the slope of the regression, quantifies stomatal sensitivity to D, and the parameter b, the intercept, is a reference G s when D = 1 kPa (ln D = 0).−m and b were estimated using least squares regression with Sigma Plot 10.0 (Systat Software Inc., San Jose, California).
The quadratic response function between G s /G smax and SMD was described as follows: where the parameters y 0 , k 1 and k 2 were estimated using least squares regression with Sigma Plot 10.0 (Systat Software Inc., San Jose, California).).The leaf sap, composed of xylem sap, phloem sap or cell contents, was forced out through the petiole using a portable pressure chamber (PMS Instruments, Corvallis, Oregon, USA).According to the diurnal pattern leaf water potential of Acacia mangium, the pressure exerted varied between 1.5 bar in the morning and 15 bar at noon.After the leaf sap appeared at the petiole, 25 µl leaf sap was pipetted into a 5 ml vial, then immediately frozen and stored in the freezer at 5 • C. Samples were heated in a quartz ampoule at 600 • C, cooled down and vacuumized for 3 hr.After that, they were heated in a muffle furnace at 860 • C for 2 hr and converted to purified CO 2 .The carbon isotope ratio of these samples was determined on a gas chromatograph-isotope ratio mass spectrometer (Finnigan MAT 252, Finnigan, Bremen, Germany).Carbon isotopic compositions of soluble sugar, water-soluble organic matter in leaves and phloem sap have all been used as indicators of recent canopy photosynthate (Gessler et al., 2004;Hu et al., 2010;Rascher et al., 2010).In our study, we used carbon isotopic compositions of leaf sap as an indicator of recent photosynthate, which was needed to match with the high-resolution sap flow data.
Five air sampling points were situated at 10 m to 15 m above the ground in the canopy of the A. mangium plantation.Air samples were collected twice per hour during the periods when leaf sap was collected.The air samples were collected with an electromagnetic pump and a plastic tube.The samples were then injected into pre-evacuated 500 ml gas sampling bags.At the same time, atmospheric CO 2 concentration was recorded by an infrared gas analyzer (IRGA) (LI-6262, Li-Cor Inc., Lincoln, NE, USA).The CO 2 isotope ratio of air samples with cryogenic preconcentration was also determined on Finnigan MAT 252.Carbon isotopic compositions are specified as δ 13 C values (Keeling, 1958) where R p is the abundance ratio of 13 C/ 12 C of the examined sample and R s refers to the internationally recognized standard abundance ratio of 13 C/ 12 C (PDB).

Canopy CO 2 uptake rate estimation
Canopy CO 2 uptake of A. mangium, estimated by combining sap flow and 13 C techniques, as proposed by Zhao et al. (2005a), was calculated as follows.The CO 2 uptake rate (F CO 2 , µmol CO 2 m −2 s −1 ) was then given (Farquhar et al., 1982) by where g CO 2 is the stomatal conductance for CO 2 .
In order to account for the influence of the 13 C/ 12 C ratio in the air, isotope discrimination ( ) was used to express the discrimination of plants against 13 C (Farquhar et al., 1982 where δ 13 C a and δ 13 C p stand for carbon isotope compositions in air and leaf sap. is frequently used to derive intercellular CO 2 concentration (C i ) according to the linear relationship (Farquhar et al., 1982 where a stands for the fractionation that occurs as CO 2 diffuses through stomates (4.4 ‰), b for fractionation during carboxylation (27.5 ‰), C i for intercellular CO 2 concentration, and C a for ambient CO 2 concentration.Combining Eqs. ( 12) and ( 14), we then had Stomatal conductance to water vapor (g H 2 O ) was converted to conductance for CO 2 by dividing the stomatal conductance by 1.6, giving Equation ( 16) depicts a relationship linking F CO 2 , g H 2 O , C a , and at the leaf level.Subsequently this relationship was scaled up to canopy level by incorporating G s calculated with Eq. ( 7) using the sap flow-based stand transpiration.
G s is expressed as a product of LAI and g H 2 O : Thus, combining Eqs. ( 7), ( 14) and ( 17) led to a mean F CO 2 of our A. mangium stand where is seasonally integrated 13 C discrimination taken over four representative days in the wet and dry seasons, separately.Notably, accounting for nocturnal sap flux caused by the recharge of water to trunks and branches, as well as nocturnal transpiration, is a vital step for accurately estimating canopy transpiration, and thereby canopy CO 2 uptake.Allocation of nighttime sap flow to the refilling of depleted water storage has been found in the same A. mangium trees (Wang et al., 2012).In view of the obvious nocturnal water recharge, we used the 24 h diurnal sap flux to accurately estimate canopy CO 2 uptake.Furthermore, the daily canopy CO 2 uptake for our mature A. mangium plantation was obtained based on the sap flow measurement-derived canopy stomatal conductance and seasonally integrated 13 C discrimination.The annual canopy CO 2 uptake for our mature A. mangium plantation was obtained by summarizing monthly canopy CO 2 uptake during the whole year.Daily and annual mean standard errors were also estimated.Subsequently, these values were compared with estimates from other methods, including the gas exchange method, carbon allocation studies, and modeling approaches.

Leaf gas exchange measurements
Leaf photosynthetic rate (P n ) and stomatal conductance (g s ) of 10 sun-exposed leaves and 10 shaded leaves from A. mangium trees at the study plot were measured with a portable gas exchange system (Li-Cor 6400, Li-Cor Inc., Lincoln, NE, USA).A frame platform was erected on the upper slope of the study plot so that the leaves within the canopy on the lower slope could be accessed.The measurements were taken on the same trees every hour from 06:00 to 19:00 (GMT + 08:00) on 6 and 17 December 2006 (winter, dry season), 9 and 10 May 2007 (spring, wet season), 26 July 2007, 1 August 2007 (summer, wet season), and 6 and 7 November 2007 (autumn, dry season).

Statistics and sensitivity analysis
Statistical analyses were performed in SPSS 16.0 (SPSS Inc., Chicago, USA) and Sigmaplot 10.0 (Systat Software Inc., San Jose, CA).Two-way ANOVA (repeated measures) (GLM procedure) was applied for parameters (C a , C i , C i /C a , δ 13 C a , δ 13 C p , ) to evaluate the main effects of season (wet season and dry season), time (from 06:00 to 19:00) and their interaction over time.Partial correlation analysis was used to determine the correlations between and PAR, D, C a , and G s .Linear regression analyses between g s and G s , and between P n and F CO 2 , were performed.The regression lines for sunlit and shade leaves and for the dry and wet seasons were compared by analysis of covariance.Non-linear regressions were performed on the correlation between G s and D among seasons and years.
In order to quantify error propagation when modeling F CO 2 based on measurements of G s , C a and , we applied a Monte Carlo method (Hollinger and Richardson, 2005;Hu et al., 2010) (Matlab R2008B, The Mathworks).This approach accounted for all known parameter uncertainties in calculating F CO 2 , including C a and (differences influenced by seasons), and G s (differences influenced by tree size and seasonality in leaf area).For each selected day (8 d between December 2006 and November 2007), we modeled F CO 2 in an iterative manner (10 000 times) and ran-domly sampled from a range of uniformly distributed G s , C a and values to calculate F CO 2 .Hence, we were able to estimate F CO 2 without the constraint of limited sampling frequency and could determine the sensitivity of the estimated F CO 2 to G s , C a and .The range of the G s (between 4.25 and 64.52 mmol m −2 s −1 ), C a (between 381.51 and 416.58 µmol mol −1 ) and (between 20.47 and 23.89 ‰) values used for constructing the pool of simulations was determined a priori.Because F CO 2 was modeled using both measurements of G s , C a and , we used random values drawn from the same pool of F CO 2 values used in the Monte Carlo analysis, and examined the sensitivity of F CO 2 to G s , C a and as independent variables.We plotted the linear regression relationship plus 95 % confidence intervals from this analysis for the data reported in Fig. 7.

Microclimate, leaf area index, and stand transpiration
Seasonal and inter-annual variability in vapor pressure deficit (D), photosynthetically active radiation (PAR), precipitation (P ), soil water content (SWC), leaf area index (LAI), and stand transpiration (E L ) from 2004 to 2007 are shown in Fig. 1.E L has clear seasonality, corresponding to D, PAR, and P seasonality, in contrast to the relatively small seasonal variation in LAI (p = 0.151).The means of annual averages of D, SWC, and LAI were 0.46 kPa, 0.26 m 3 m −3 , and 1.95 m 2 m −2 , respectively.The means of annual totals of PAR, P , and E L were 5752 mol m −2 , 1391 mm, and 220 mm, respectively.The inter-annual variability of D, PAR, SWC, and LAI was relatively small, with coefficients of variation (CV) of 8.6 %, 3.0 %, 7.1 %, and 4.2 %.However, large interannual variability of P and E L (CV 40.7 % and 26.7 %) was observed.

Diurnal and seasonal variability of physiological parameters and carbon isotope data
The diurnal ambient CO 2 concentration (C a ), intercellular CO 2 concentration (C i ), the ratio of the intercellular and ambient CO 2 concentrations (C i /C a ), and carbon isotope compositions in leaf sap (δ 13 C p ) were higher in the morning and at dusk, but lower at noon (Fig. 2a, b, c, d, f), while isotope compositions in the canopy air (δ 13 C a ) and the photosynthetic 13 C discrimination ( ) were lower in the early morning, gradually increased, and then slightly decreased from noon to dusk (Fig. 2e, g).The diurnal C a , δ 13 C p , δ 13 C a , and in the wet season fluctuated, while they were relatively flat in the dry season.The diurnal canopy stomatal conductance (G s ) increased rapidly just after sunrise, reached a maximum in the early or late morning, then decreased progressively, eventually declining to near-zero values around sunset (Fig. 2h).G s exhibited a narrow peak with a maximum around 100 mmol m −2 s −1 in the wet season, with a wider peak with a maximum around 50 mmol m −2 s −1 in the dry season.The opening of stomata, as well as the time of maximum value (around 08:00), occurred earlier in the wet season than in the dry season (around 11:00).
Diurnal C a , C i , C i /C a , δ 13 C p , δ 13 C a , and fluctuations were observed (Table 1).CVs of diurnal C a , C i , C i /C a , δ 13 C p , δ 13 C a , and were below 5.5 %, 11.6 %, 10.1 %, 6.4 %, 12.7 %, and 11.2 %, respectively.Because of the lack of strong variability in diurnal values, the average value was used in subsequent calculations of F CO 2 for the whole season.Furthermore, values of diurnal C i /C a for sunlit leaves ranged between 0.55 and 0.94, while those of shaded leaves ranged between 0.68 and 0.95.The mean daily C i /C a value of 0.76 for sunlit leaves was significantly lower than that of 0.83 for shaded leaves (p = 0.001).At the seasonal scale, CVs of seasonal C a , C i , C i /C a , δ 13 C p , δ 13 C a , and were below 5.1 %, 5.0 %, 3.3 %, 5.1 %, 14.4 %, and 9.8 % ( the wet season and dry season (p = 0.212, p = 0.644, p = 0.446, p = 0.420, p = 0.090, p = 0.988), and this did not change through time (see Supplement C for details).
Hourly was negatively correlated with C a (r = −0.549,p < 0.001) in the wet season, negatively correlated with PAR (r = −0.425,p = 0.002) and positively correlated with G s (r = 0.464, p < 0.001) in the dry season, and positively correlated with D (r = 0.250, p = 0.009) over the entire year (Table 2).Boundary-line response curves using 10-minute values of G s and PAR, D, and SMD were created to determine the G s response patterns in both the wet and dry seasons (Fig. 3).G s in relation to PAR is shown in Fig. 3a-h.G s increased with PAR, and appeared to be light-saturated at 400 and 200 µmol m −2 s −1 in the wet season and dry season, respectively.However, all the regressions between G s and PAR were non-significant.G s in relation to D was shown in Fig. 3i-p.G s decreased exponentially from D values of 0.6 and 0.8 kPa in the wet season and dry season, respectively.Nearly all the regressions between G s and D were significant (p < 0.0001), with R 2 ranging from 0.60 to 0.99.Stomatal sensitivity (m) was greater in the wet season than that in the dry season (Table 3).m increased linearly with reference canopy stomatal conductance (G sref ) (slope = 0.70, R 2 = 0.96, n = 8).G s in relation to SMD was shown in Fig. 3 q-t

Canopy CO 2 uptake rate of mature A. mangium plantation
At the leaf level scale, diurnal leaf stomatal conductance (g s ) and net photosynthesis (P n of A. mangium) obtained through gas exchange measurements reached their peak around 09:00 in the wet season, and around 11:00 in the dry season (Fig. 4).The diurnal means of g s in all leaves, sunlit leaves, and shaded leaves were 136.40, 129.42, and 144.34 mmol m −2 s −1 in the wet season, and 58.65, 56.74, and 60.42 mmol m −2 s −1 in the dry season, respectively.The diurnal means of P n in all leaves, sunlit leaves, and shaded leaves were 3.42, 3.93, and 2.84 µmol m −2 s −1 in the wet season, and 1.85, 2.17, and 1.34 µmol m −2 s −1 in the dry season, respectively.Notably, CVs were high (ranging mostly from 40 % to 60 %) for g s and P n .At the canopy level scale, diurnal canopy CO 2 uptake rate (F CO 2 ) peaked around 09:00, with a maximum of 7.3 µmol m −2 s −1 in the wet season (Fig. 5a).Diurnal F CO 2 was more symmetrical and smoother, reaching a maximum of 3.9 µmol m −2 s −1 around 11:00 in the dry season (Fig. 5a).The diurnal means of G s and F CO 2 in all leaves were 56.70 mmol m −2 s −1 and 2.78 µmol m −2 s −1 in the wet season, and 28.81 mmol m −2 s −1 and 1.32 µmol m −2 s −1 in the dry season.Furthermore, as shown in Fig. 6, values of both g s and P n were higher than those of G s and of F CO 2 , with regression coefficients of determination (R 2 ) of 0.58 and 0.19 in the wet season, 0.70 and 0.65 in the dry season, and 0.44 and 0.35 in all seasons, respectively.An analysis of covariance confirmed that neither the slopes nor the intercepts of nearly all the regression lines for sunlit and shade leaves, for wet and dry seasons, were significant (Table 4, p > 0.05).

Canopy CO 2 uptake rate sensitivity analysis
We examined the sensitivity of F CO 2 to changes in G s , C a and between December 2006 and November 2007 (Fig. 7).The positive correlation between G s and F CO 2 was significant, and the R 2 value was high (R 2 = 0.8263; p < 0.0001).Although there were significant correlations between C a and F CO 2 (p = 0.0001) and and F CO 2 (p < 0.0001), the R 2 values were very low (R 2 = 0.0016 and R 2 = 0.1348, respectively).

The theoretical basis for estimating canopy CO 2 uptake
To our knowledge, all methods for direct measurement of canopy CO 2 uptake rely on scaling.Morén et al. (2001) combined water use efficiency (WUE) on the based branch gasexchange measurements and canopy transpiration (E) as a scalar to get canopy carbon assimilation of a boreal forest, and found that canopy WUE showed a strong dependency on vapour pressure deficit (D).Catovsky et al. (2002) combined whole-tree sap flow measurements with micrometeorological monitoring and leaf-level gas exchange to determine wholetree carbon gain, and showed that estimates of canopy photosynthesis were most sensitive to measurements of D and the relationship between photosynthesis and conductance.However, Catovsky's and Morén's methods required a lot of measurements of gas exchange on leaves or branches.More recently, to solve the difficulties mentioned above, Zhao et al. (2005a) made use of the sap flow technique for scaling and isotope technique for plant water use, and then proposed a Table 3. Stomatal sensitivity to vapor pressure deficit (D) estimated as the slope of the relationship between canopy stomatal conductance and ln(D) (Oren et al., 1999), applied to those data in which G s decreased exponentially with D.  2010) used observed transpiration rates and needle sugar carbon isotope ratios to estimate whole-tree carbon assimilation rates, and then combined with species distribution and tree size to estimate gross primary productivity (GPP) using an ecosystem process model.

Temporal patterns of canopy CO 2 uptake and its dependence on environment and stomata
Canopy CO 2 uptake rate (F CO 2 ) varied diurnally (Fig. 5).
The sharp peak in photosynthesis in early morning periods during the wet season may be due to higher G s and C a , as well as lower during this period.During the wet season G s peaked around 08:00 and then rapidly dropped for the rest of the day (Fig. 2h).This diurnal pattern was similar to that of canopy conductance on well-watered 9-year-old Sultana grapevines (Lu et al., 2003), native forest composed of Lomatia hirsuta, Schinus patagonicus, Nothofagus antarctica and Diostea juncea (Fernandez et al., 2009), and pristine Nothofagus forest (Köstner et al., 1992).During the dry season, lower values of G s and less acute peaks were due to a water supply from the soil that was inadequate to meet the evaporative demand.
Strong seasonality in F CO 2 of A. mangium was observed (Fig. 5).Based on this, we related F CO 2 to the seasonally changing environmental variables.Higher F CO 2 in the wet season coincided with the season of rapid growth, ample supply of water, and strong solar radiation, the typical environment of subtropical ecosystems (Figs. 1,5).Notably, in the dry season, water supply did not match evaporative demand at the study site (P and SWC were relatively low, while PAR and T a were still high).These conditions may induce water stress in A. mangium, which caused a significant reduction in stomatal conductance.Furthermore, lower annual F CO 2 in 2005 may be due to the frequent rainfall that occurred in  June and July.Higher monthly F CO 2 in May 2005 may be attributed to the after-effect of soil water stress on stand transpiration, thereby F CO 2 .
Canopy stomatal conductance is a critical factor regulating the F CO 2 of A. mangium.It was reported that similar canopy conductance estimates for this A. mangium plantation were obtained from both the Penman-Monteith formula and the  simplified equation by Köstner et al. (1992) (Ma, 2008).The diurnal pattern of F CO 2 was similar to that of G s (Figs. 2, 5).G s was mainly influenced by D and PAR (Fig. 3).When D was lower than 0.6 kPa, PAR was the main driver of G s , because it induced stomatal opening.When D approached 0.6 kPa and beyond, D became the dominant factor.The in-crease in D led to a decrease in G s despite the increase in PAR, because the hydraulic system needed to be protected from cavitation.High D led to varying degrees of decline in G s (Table 3), and thus the corresponding reduction in F CO 2 .The slope of this linear function between stomatal sensitivity (m) and reference canopy stomatal conductance (G sref ) was (see text for further explanation).All regressions were highly significant (p < 0.01).Data were obtained by the SF/SI approach and leaf gas exchange measurements.
consistent with the shallower slope of approximately 0.60 observed for the mesic species in Oren et al. (1999).

Comparison with estimates from gas exchange measurements
With regard to the patterns, the diurnal dynamics of gas exchange results agreed well with that of the sap flux/stable isotope method (SF/SI) (Figs. 2h,4,5).Better agreement between F CO 2 and P n in the dry season was likely due to the fact that the lower leaf area index in the dry season (1.91 ± 0.38 in the dry season vs. 2.04 ± 0.22 in the wet season) would tend to minimize the effects of leaf shading, which can diminish conductance in lower-canopy leaves.However, F CO 2 from SF/SI was 27.0 % lower than P n from gas exchange measurements (Fig. 6).The diurnal C i /C a from gas exchange measurements and stable isotope values ranged from 0.63 to 0.94 and from 0.57 to 0.95, respectively.The seasonal C i /C a from gas exchange measurements and stable isotope values in the wet season (0.80, 0.78), dry  season (0.78, 0.78), and whole year (0.79, 0.78) also matched well.Thus, the differences in F CO 2 and P n resulted mainly from differences in stomatal conductance.G s from the SF/SI approach was 67.3 % lower than g s from gas exchange measurements, though the diurnal dynamics were similar.Furthermore, the differences may also be attributed to high variation in gas exchange results, which were variable by leaf position and age, and thus any one measurement was not representative of the entire canopy; variability of the average response was high.Given that F CO 2 was calculated based on and C i /C a averaged across a variety of age and light classes, this source of variation was reduced for the results from the SF/SI approach.The analysis of the differences between the gas exchange and results from the SF/SI approach highlighted the advantages of the latter method.

Comparison with estimates from other approaches for estimating F CO 2
For comparison, the gross primary production (GPP) of forests in different climate zones, using several different methods, was summarized in Table 5. GPP in tropical forests (8.35 ± 0.23 gC m −2 d −1 ) was significantly higher than that in subtropical forests (5.38 ± 0.46 gC m −2 d −1 , p < 0.001) and temperate forests (2.85 ± 0.45 gC m −2 d −1 , p < 0.001).Most of the differences in the gross primary production (GPP) of forests in various climate zones were related to climate.Our estimate of canopy CO 2 uptake for a mature A. mangium plantation (2.13 ± 0.40 gC m −2 d −1 , Fig. 5) approached the lower range of values for subtropical mixed forests (Gebremichael and Barros, 2006;Gu et al., 2006).Our lower estimates of GPP may be due to the lower mean canopy stomatal conductance (0.135 cm s −1 vs. 0.726 cm s −1 , 0.25 cm s −1 ), higher C i /C a (0.76 to 0.84), greater tree height (17.8 m vs. 10.8 m, 2 m), and lower LAI (1.95 m 2 m −2 vs. 3.20 m 2 m −2 , 3.77 m 2 m −2 ), compared to other subtropical mixed forests (Ma, 2008;Gebremichael and Barros, 2006;Gu et al., 2006).The canopy stomatal conductance of individual trees decreases with tree height (Schäfer et al., 2000), while the lower LAI might lead to higher conductance and more control by stomatal conductance, since there is less shading (Granier et al., 2000).The C i /C a ratio, which is maintained at a constant or nearconstant value in many plant species, represents a balance between the rates of inward CO 2 diffusion (controlled by stomatal conductance) and CO 2 assimilation (controlled by photosynthetic light/dark reactions) (Ehleringer and Cerling, 1995).In our study, the C i /C a ratio varied from 0.76 to 0.84, which was higher than the ranges found in cottonwoods in a riparian woodland (0.75-0.78) (Letts et al., 2008), in 13year-old loblolly pine (Pinus taeda) trees (0.45-0.80) (Maier et al., 2002), and in nine well-watered conifer species (0.57-0.68) (Brodribb, 1996).Relatively high C i /C a ratios have also been found in tropical rain forest species (Lloyd and Farqhar, 1994;Ishida et al., 1996), a Canarian laurel forest tree species (Laurus azorica) (González-Rodríguez et al., 2001).There are three possible reasons for the inference that the C i /C a ratio may result in a decrease in photosynthetic rate of our study species.Firstly, high C i /C a values may suggest a relatively small stomatal limitation to net photosynthetic rate and non-conservative water use (González-Rodríguez et al., 2001;Ishida et al., 1996).Consistent with this inference, Cienciala et al. (2000) also observed no apparent limitation to water flux in A. mangium.It is likely that high transpiration rates may cause a localized leaf water deficit, which depletes photosynthetic capacity (Sharkey, 1984).Secondly, non-stomatal limitations may also be responsible for the decline in assimilation rates (Lauer and Boyer, 1992).Thirdly, the reduction in photosynthesis associated with leaf senescence should result in high C i /C a values (Ponton et al., 2006).Notably, our estimate of canopy CO 2 uptake for the 20year-old A. mangium stand was much lower than that for 4 to 6-year-old A. mangium (8.77 gC m −2 d −1 ) in São Paulo, Brazil (Nouvellon et al., 2012).Our lower estimates of GPP may be partly explained by an age-related decline in photosynthesis (20-year-old vs. 4 to 6-year-old).It may also be the result of lower photosynthetically active radiation absorbed by the canopy, due to lower stand density (734 trees ha −1 vs. 1111 trees ha −1 ), and lower LAI (1.95 m 2 m −2 vs. 3.479 m 2 m −2 ) (Ma, 2008a;Nouvellon et al., 2012).Furthermore, we compared methods for quantifying GPP constrained by water use measurements.Our estimates were comparable with those of spruce (2.66 gC m −2 d −1 ) and mixed forest (0.94 gC m −2 d −1 ), also using the sap flow-based approach (Köstner et al., 2008;Hu et al., 2010), but lower than values determined by satellite-based modeling or carbon allocation studies (Table 5).These differences suggest the necessity of measuring species-specific fluxes in assessing CO 2 uptake.

Sensitivity analysis
Through our sensitivity analysis using a Monte Carlo approach, we were able to simulate many more combinations of G s , C a and than obtained from our limited 8 d samplings.We found that the dynamics in the estimates of F CO 2 were mainly driven by dynamics in G s , compared with C a and (Fig. 7).Hu et al. (2010) also demonstrated that the calculation of canopy photosynthesis was much more sensitive to transpiration rate than to δ 13 C values.In our study, covariance between G s and F CO 2 occurs because CO 2 is assimilated through the stomatal pores.Variation in between seasons was non-significant (p = 0.988), compared with variation in G s .This may be due to the fact that , the proportioning coefficient between G s and F CO 2 as shown in Eq. ( 15), is constrained as the ratio of two co-varying variables.It may also be due to the evidence that stomata functions in a way to optimize G s and F CO 2 in relation to environmental changes (Wong et al., 1979).C i /C a is also a very useful parameter for

Fig. 1 .
Fig. 1.(a) Monthly mean vapor pressure deficit (D) and photosynthetically active radiation (PAR) at the experimental site; (b) precipitation (P ) and monthly mean soil water content (SWC) at the experimental site; (c) monthly mean leaf area index (LAI) and (d) stand transpiration (E L ) of A. mangium.Error bars indicate standard deviation (n = 20).Boxes filled with the coarse pattern in gray color indicate the wet seasons.
. G s decreased as SMD increased in July in 2005 and 2006.However, no significant relationships between G s and SMD were observed across entire years (except for Tree No. 3 in 2005, p = 0.097, and Tree No. 4 in 2005, p = 0.064).

Fig. 2 .
Fig. 2. Diurnal variations of (a) ambient CO 2 concentration (C a ), (b) intercellular CO 2 concentration (C i ), (c) the ratio of the intercellular and ambient CO 2 concentrations (C i /C a ), (d) C i /C a in the sunlit and shaded leaves in all seasons, (e) carbon isotope compositions in canopy atmosphere (δ 13 C a ), (f) carbon isotope compositions in leaf sap (δ 13 C p ), and (g) photosynthetic 13 C discrimination ( ) and (h) canopy stomatal conductance (G s ) in both the wet season (9 and 10 May, 26 July, 1 August) and the dry season (6 and 7 November, 6 and 17 December).Error bars indicate standard deviation (n = 8).C i /C a was obtained by leaf gas exchange measurements.Other data were obtained by combining sap flow measurement and stable carbon isotope techniques (SF/SI approach).

Fig. 3 .
Fig.3.Scatter plots and response functions of the ratio of canopy stomatal conductance to the maximum (G s /G smax ) to changes in photosynthetically active radiation (PAR) (a-h), vapor pressure deficit (D) (i-p), and soil moisture deficit (SMD) (q-t) in both wet seasons and dry seasons from 2004-2007.Data were obtained through sap flow measurements.

Fig. 4 .
Fig. 4. Diurnal patterns of leaf stomatal conductance (g s ) and net photosynthesis (P n ) of A. mangium in both the wet season (9 and 10 May, 26 July, 1 August) and the dry season (6 and 7 November, 6 and 17 December).Error bars indicate standard deviation (n = 4).July and December are representative of wet and dry conditions, respectively.Data were obtained by leaf gas exchange measurements.

Fig. 5 .
Fig. 5. Diurnal, monthly, seasonal, and annual trends of canopy CO 2 uptake rate (F CO 2 ) of A. mangium plantation from 2004 to 2007.Error bars indicate standard error.Data were obtained by the SF/SI approach.

Fig. 6 .
Fig. 6.Scatter plots of leaf stomatal conductance (g s ) vs. canopy stomatal conductance (G s ), and leaf net photosynthesis (P n ) vs. canopy CO 2 uptake (F CO 2 ) in (a, b) the wet season (9 and 10 May, 26 July), (c, d) the dry season (6 and 7 November, 6 and 17 December), and (e, f) all seasons (9 and 10 May, 26 July, 1 August, 6 and 7 November, 6 and 17 December).The data in the ellipse in (e) occurred on 1 August (see text for further explanation).All regressions were highly significant (p < 0.01).Data were obtained by the SF/SI approach and leaf gas exchange measurements.

Fig. 7 Fig. 7 .
Fig. 7 Canopy CO 2 uptake rate (F CO2 ) sensitivity analysis.The bold black lines are the regression

Table 1 .
The diurnal ranges, seasonal means, and coefficients of variation (CV) in ambient CO 2 concentration (C a ), intercellular CO 2 concentration (C i ), the ratio of the intercellular and ambient CO 2 concentrations (C i /C a ), and carbon isotope compositions in leaf sap (δ 13 C p ), isotope compositions in the canopy air (δ 13 C a ) and the photosynthetic 13 C discrimination ( ).

Table 2 .
Partial correlations between hourly and photosynthetically active radiation (PAR), vapor pressure deficit (D), ambient CO 2 concentration (C a ), and canopy stomatal conductance (G s ).
a Indicates significant partial correlation at 0.01 level.
Farquhar et al. (1982)as estimated using relative G s values of each species instead of absolute values.D threshold is the value of the leaf to air vapor pressure difference from which G s begins to decrease in an exponential form.Reference canopy stomatal conductance (G sref ) is G s at D = 1 kPa.newapproachcombining the sap flow measurements and stable 13 C techniques to estimate forest C assimilation.Herein, the relationship between and C i /C a laid out byFarquhar et al. (1982)is integrated with the relationship between F CO 2 and stomatal conductance for CO 2 , so that a new relationship linking , F CO 2 and G s was obtained.Consistent with our study,Hu et al. (

Table 4 .
The regression lines in Fig.6and covariance analysis.