Volume 119, Issue 5 p. 996-1013
Research Article
Free Access

Temporal dynamics of oxygen isotope compositions of soil and canopy CO2 fluxes in a temperate deciduous forest

E. Santos

Corresponding Author

E. Santos

Department of Agronomy, Kansas State University, Manhattan, Kansas, USA

Correspondence to: E. Santos,

[email protected]

Search for more papers by this author
C. Wagner-Riddle

C. Wagner-Riddle

School of Environmental Sciences, Guelph, Ontario, Canada

Search for more papers by this author
X. Lee

X. Lee

School of Forestry and Environmental Studies, Yale University, New Haven, Connecticut, USA

Yale-NUIST Center on Atmospheric Environment, Nanjing University of Information, Science and Technology, Nanjing, China

Search for more papers by this author
J. Warland

J. Warland

School of Environmental Sciences, Guelph, Ontario, Canada

Search for more papers by this author
S. Brown

S. Brown

School of Environmental Sciences, Guelph, Ontario, Canada

Search for more papers by this author
R. Staebler

R. Staebler

Air Quality Processes Research Section, Environment Canada, Toronto, Ontario, Canada

Search for more papers by this author
P. Bartlett

P. Bartlett

Climate Processes Section, Environment Canada, Toronto, Ontario, Canada

Search for more papers by this author
K. Kim

K. Kim

School of Forestry and Environmental Studies, Yale University, New Haven, Connecticut, USA

Search for more papers by this author
First published: 29 April 2014
Citations: 5

Contribution no 14-118-J from the Kansas Agricultural Experiment Station.

Abstract

Partitioning of CO2 exchange into canopy (FA) and soil (FR) flux components is essential to improve our understanding of ecosystem processes. The stable isotope C18OO can be used for flux partitioning, but this approach depends on the magnitude and consistency of the isotope disequilibrium (Deq), i.e., the difference between the isotope compositions of FR (δA) and FA (δR). In this study, high temporal resolution isotopic data were used (1) to test the suitability of existing steady state and nonsteady models to estimate H218O enrichment in a mixed forest canopy, (2) to investigate the temporal dynamics of δA using a big-leaf parameterization, and (3) to quantify the magnitude of the C18OO disequilibrium (Deq) in a temperate deciduous forest throughout the growing season and to determine the sensitivity of this variable to the CO2 hydration efficiency (θeq). A departure from steady state conditions was observed even at midday in this study, so the nonsteady state formulation provided better estimates of leaf water isotope composition. The dynamics of δR was mainly driven by changes in soil water isotope composition, caused by precipitation events. Large Deq values (up to 11‰) were predicted; however, the magnitude of the disequilibrium was variable throughout the season. The magnitude of Deq was also very sensitive to the hydration efficiencies in the canopy. For this temperate forest during most of the growing season, the magnitude of Deq was inversely proportional to θeq, due to the very negative δR signal, which is contrary to observations for other ecosystems investigated in previous studies.

Key Points

  • Large isotope disequilibrium was predicted at the site
  • Models captured the dynamics leaf water isotope composition
  • Precipitation was the main driver of the dynamics of flux isotope signatures

1 Introduction

Stable isotopes of CO2, such as 13CO2 and C18OO, can be valuable tools to study the CO2 exchange in a broad range of spatial scales [Bowling et al., 2008; Ogée et al., 2004; Werner et al., 2012; Yakir and Wang, 1996; Yakir and Sternberg, 2000]. The partitioning of net ecosystem CO2 exchange (FN) into canopy (FA) and soil (FR) flux components is essential to improve our understanding of biophysical controls of photosynthesis and soil respiration processes at the ecosystem level. However, the success of the isotope flux partitioning approach depends on the magnitude and consistency of the isotope disequilibrium (Deq), i.e., the difference between the isotope compositions of FR (δA) and FA (δR) [Ogée et al., 2004; Yakir and Wang, 1996]:
urn:x-wiley:21698953:media:jgrg20215:jgrg20215-math-0001(1)

Larger 18O Deq values compared to 13C Deq are typically obtained in established ecosystems as the difference in 13C composition of plant and soil sources is not very large [Ogée et al., 2004], suggesting C18OO is a better tracer for partitioning FN for these ecosystems. However, large temporal variability in 18O Deq has been observed in the few long-term studies conducted to date. Wingate et al. [2010] used automated branch chambers (for δA) and soil chambers (for δR) with continuous measurements of C18OO composition of the air using tunable diode laser spectroscopy (TDLS) and obtained typical daytime 18O Deq (denoted hereafter as Deq) of 10‰ for a maritime pine stand; however, lower disequilibrium values were observed after rain events. Griffis et al. [2011] observed values of Deq ranging from 0.3 to 17.1‰ for corn and also smaller Deq values immediately after precipitation events using TDLS technique with soil chamber and eddy covariance measurements. Hence, longer-term studies to characterize the temporal dynamics of Deq are needed.

The Deq in ecosystems arise from the isotopic equilibration between CO2 and water in soil and foliage, which usually have distinct H218O composition [Yakir and Sternberg, 2000]. The coupling between δ18O of plant and soil CO2 fluxes with their respective water pools requires prior knowledge of the isotope composition of water pools to interpret variations in δA and δR. Another important variable affecting δR is the rate of CO2 hydration in soils, which has been reported to be faster than the uncatalyzed reaction in different ecosystems [Griffis et al., 2011; Santos et al., 2012; Seibt et al., 2006; Wingate et al., 2010; Wingate et al., 2009]. Seibt et al. [2006] hypothesized that the enhancement of CO2 hydration was a result of the activity of the carbon anhydrase enzyme (CA) in the soil. The CA activity in soils and δR short-term variation [Santos et al., 2012; Wingate et al., 2010] can have important effects on Deq and should be further investigated.

The leaf-level δA can be inferred directly from branch chambers [Wingate et al., 2010], but canopy-level δA cannot be measured directly and is usually estimated using big-leaf models [Bowling et al., 2001; Lee et al., 2009] or multilayer soil-vegetation-atmosphere transfer models [Ogee et al., 2003; Ogée et al., 2004]. The canopy-scale δA variation is driven by variables such as the efficiency of CO2 hydration in the canopy, which is also dependent on CA activity in the foliage, canopy conductance, relative humidity, and the H218O composition of leaf water at the evaporation sites (δLe) [Farquhar et al., 1993]. This last variable is crucial for determining δA but cannot be measured directly at the canopy scale. However, the mechanisms governing δLe dynamics at the leaf scale are reasonably understood, and different models have been used to estimate leaf water H218O enrichment [Craig and Gordon, 1965; Cuntz et al., 2007; Dongmann et al., 1974; Farquhar and Cernusak, 2005]. The classic formulation proposed by Craig and Gordon [1965] to simulate the isotopic fractionation from a liquid water surface has been used to estimate δLe [Dongmann et al., 1974; Flanagan et al., 1991] by assuming steady state conditions, i.e., the water being transpired by the leaf has the same isotope composition of the water entering the leaf. Recent studies suggested that steady state conditions are often violated under field conditions [Lai et al., 2006; Welp et al., 2008; Wingate et al., 2010; Xiao et al., 2012]. Farquhar and Cernusak [2005] proposed a more detailed formulation, which takes into consideration the nonsteady state behavior and heterogeneity of leaf water isotope composition. There has been limited investigation of the effects of steady state versus nonsteady state models for δLe on δA for natural ecosystems over a growing season [Griffis et al., 2011; Wingate et al., 2010].

Models developed to estimate δLe use input variables measured near the leaf boundary layer [Craig and Gordon, 1965; Farquhar and Cernusak, 2005]. However, in canopy-scale studies, measurements are usually taken several meters above the canopy from which canopy-scale δLe and H218O composition of leaf bulk water (δLb) can be derived through the application of a big-leaf model [Lee et al., 2009]. Recent research has shown that the isotope exchange is significantly affected by the intense turbulent mixing above plant canopies, which is negligible in the majority of chamber-based studies [Griffis et al., 2011; Lee et al., 2009; Xiao et al., 2012]. Lee et al. [2009] suggested that the inclusion of turbulent effects on the calculation of isotope kinetic factors through addition of an aerodynamic resistance might improve the estimates of regional and global isotope budgets. Furthermore, studies over crop canopies have shown that estimates of δLb improved when turbulence effects were included in a big-leaf parameterization for isotope exchange [Griffis et al., 2011; Xiao et al., 2012]. However, additional investigation of the turbulence effects on canopy-scale H218O enrichment is required in forest ecosystems, where the magnitude of the aerodynamic resistance is much smaller than over crop canopies.

Lee et al. [2009] demonstrated that parameterized canopy-scale kinetic factors are more appropriate to study isotope CO2 and water vapor exchange at the ecosystem level, when measurements are taken at a reference point above the canopy air space. Canopy-scale studies have also indicated that some parameters of leaf-scale models, such as the CO2 hydration efficiency in the foliage, may not be scaled-up properly to the canopy level [Griffis et al., 2011; Xiao et al., 2012; Xiao et al., 2010]. Xiao et al. [2011] found that the canopy-scale CO2 hydration efficiency (θeq), derived from eddy isoforcing measurements above a soybean canopy, was 0.46, which is much lower than the leaf-scale CO2 hydration efficiency (θeq = 0.75), obtained for C-3 plants under laboratory conditions [Gillon and Yakir, 2001]. Similar results were found by Griffis et al. [2011] who reported significantly smaller canopy-scale θeq for a corn canopy than that measured in laboratory conditions. How these contrasting canopy and leaf θeq values affect the 18O Deq derived using a big-leaf model is not clear.

Advances in optical techniques in recent years have allowed the development of analyzers capable of providing robust measurements of water and CO2 isotopes under field conditions [Griffis, 2013]. Studies investigating the mechanisms controlling isotope exchange at the ecosystem scale and their temporal dynamics are needed [Griffis et al., 2011; Xiao et al., 2012; Xiao et al., 2010]. In this study, the biophysical processes governing canopy-scale isotope exchange in a forest ecosystem were investigated using detailed isotope measurements during a growing season. Atmospheric C18OO and H218O composition (δa and δv, respectively) were measured quasi-continuously and simultaneously above a temperate deciduous forest during a growing season using TDLS. In addition, detailed measurements of the H218O composition of ecosystem water pools were taken throughout the growing season. These measurements and additional supporting variables were used (1) to test the suitability of existing steady state and nonsteady state leaf enrichment models to estimate H218O enrichment of the foliage at the canopy scale in a mixed forest, (2) to investigate the temporal dynamics of δA using a big-leaf parameterization, which takes into account turbulence effects on isotope gaseous diffusion in plant canopies, and (3) to quantify the magnitude of Deq in a temperate deciduous forest throughout the growing season and to determine the sensitivity of this variable to CO2 hydration efficiency.

2 THEORY

2.1 C18OO Composition of Canopy CO2 Flux

Lee et al. [2009] proposed a big-leaf parameterization to study land-air isotope fluxes, in which the C18OO exchange between plant canopy-atmosphere is an extension of previous leaf-scale formulations [Farquhar et al., 1993; Flanagan et al., 1991] and the isotope exchange pathways are described using a series of resistances. In this approach, the C18OO composition of the canopy flux (δA) is expressed as
urn:x-wiley:21698953:media:jgrg20215:jgrg20215-math-0002(2)
where Ca is the CO2 molar density (µmol m−3) in the air and δa is the air's C18OO composition measured above the forest canopy (section 3.2), Cc is the CO2 molar concentration (µmol m−3) in the chloroplasts, estimated using a physiological model (Appendix A), δLe is the canopy-scale H218O composition of the leaf water at the evaporative site (‰, VPDB scale), θeq is the extent of the CO2 hydration in the canopy, and εk,c is the CO2 canopy-scale kinetic fractionation factor. The εk,c takes into consideration the net fractionation of CO2 during diffusion through the leaf boundary layer (5.8‰) and stomata (8.8‰) and was calculated using a resistance weighting method [Lee et al., 2009] as follows:
urn:x-wiley:21698953:media:jgrg20215:jgrg20215-math-0003(3)
where ra is the aerodynamic resistance and rb,c and rc,c are the boundary layer and canopy resistances for CO2, respectively. The resistances ra and rb,c were calculated using land surface modeling standard procedures following Lee et al. [2009] (equations A8 and A9), and rc,c was estimated by scaling up the stomatal resistance (equation A10) obtained using the physiological model described in Appendix A.

We initially adopted a leaf-scale θeq of 0.96 derived for forest/shrub vegetation by Gillon and Yakir [2001] based on an extensive survey of CA activity and CO2 exchange rate in species from the major plant groups. However, recent studies have found evidence that canopy-scale θeq values are much lower than the leaf-scale values [Griffis et al., 2011; Xiao et al., 2010]. Xiao et al. [2010] used measurements of C18OO isoforcing, obtained using the eddy covariance technique, to optimize θeq for a soybean canopy. They found that canopy-scale θeq was 0.46 for midday periods, which is much smaller than the leaf-scale default θeq of 0.75, observed under laboratory conditions [Gillon and Yakir, 2001]. Griffis et al. [2011] performed an optimization using isoflux measurements and found θeq of 0.196 for a corn canopy during the growing season. This value was significantly smaller than the average θeq (~0.70) determined for corn leaves at their site. To quantify the impact of θeq on Deq, we also calculated δA using lower values of 0.25 and 0.6 for θeq in equation 2.

2.2 H218O Composition of Leaf Water

The foliage water isotope composition was simulated using two different formulations. The steady state formulation proposed by Craig and Gordon [1965] (CG, hereafter), with the assumption that the leaf water pool is well mixed, so that the H218O composition at the evaporative site (δLe,s) and bulk leaf water (δLb,s) are the same under steady state conditions (i.e., δLe,s = δLb,s). The second formulation used in this study is the more sophisticated model developed by Farquhar and Cernusak [2005] (FC, hereafter), which takes into account nonsteady state conditions and the progressive H218O enrichment of leaf water as the water molecules move away from the xylem to foliar evaporative sites.

The CG model has been extensively used to calculate the H218O composition of leaf water at evaporative site [Dongmann et al., 1974; Flanagan et al., 1991; Welp et al., 2008; Xiao et al., 2012]. This formulation is based on the main assumption that transpiration is in isotopic steady state; i.e., the water leaving the leaf has the same isotope composition of xylem water (δx) entering the leaf. The H218O composition of leaf water under steady state conditions (δLe,s) is given by
urn:x-wiley:21698953:media:jgrg20215:jgrg20215-math-0004(4)
where εeq = (1 − 1/αeq), αeq is the water equilibrium fractionation factor between liquid water and water vapor (> 0, ‰) [Majoube, 1971], h is the relative humidity at the canopy temperature, δv is the measured H218O composition of the air above the forest canopy (section 3.2), and εk,w (‰) is the canopy-scale kinetic fractionation factor for vapor diffusion. The canopy temperature needed for derivation of h was estimated from outgoing long-wave radiation measured using a four-component net radiometer (section 3.3) and the Stefan-Boltzmann equation. The εk,w was estimated, similarly to equation 3, following [Lee et al., 2009]
urn:x-wiley:21698953:media:jgrg20215:jgrg20215-math-0005(5)
where rb,w and rc,w are the boundary layer and canopy resistances for H2O, respectively, and are given by rb,w = 1.4rb,c. and rc,w = 1.6rc,c.
Farquhar and Cernusak [2005] proposed a formulation (hereafter FC) for situations in which the isotope composition of leaf transpiration may depart from steady state conditions and the leaf water pool is not necessary well mixed. In this formulation, the isotope composition of water at the evaporative site (δLe) and of the leaf bulk water (δLb) for nonsteady state conditions are given by
urn:x-wiley:21698953:media:jgrg20215:jgrg20215-math-0006(6)
urn:x-wiley:21698953:media:jgrg20215:jgrg20215-math-0007(7)
where αk is the fractionation factor for diffusion (αk = 1 + εk,w/1000), rt = ra + rb,w + rc,w, wi is mole fraction of (light) water in the intercellular space, P is the Péclet number, W is the foliage water content, and urn:x-wiley:21698953:media:jgrg20215:jgrg20215-math-0008 is described below. In this study, W was assumed to be constant and equal to 47.8 mol m−2, which was the average W found for the foliage of a temperate mixed forest in northeastern United States [Lee et al., 2007]. Equations 6 and 7 were solved iteratively by changing the values of δLe and δLb on the right-hand side until they matched the expressions on the left hand side.
The term urn:x-wiley:21698953:media:jgrg20215:jgrg20215-math-0009 in equation 7 represents the relationship between the isotope composition of bulk leaf water and water at the evaporative site as mediated by the Péclet effect [Farquhar and Cernusak [2005]:
urn:x-wiley:21698953:media:jgrg20215:jgrg20215-math-0010(8)
The Péclet number represents the ratio of transpiration advection of unenriched xylem water to the diffusion of H218O-enriched water from evaporating sites [Farquhar and Lloyd, 1993] and is given by
urn:x-wiley:21698953:media:jgrg20215:jgrg20215-math-0011(9)
where L is the effective diffusion length derived using an optimization procedure, C (mol m−3) is the density of liquid water and D is the temperature dependent diffusivity of H218O in water [Cuntz et al., 2007], and ET is the transpiration rate of the foliage (mol m−2 s−1) given by
urn:x-wiley:21698953:media:jgrg20215:jgrg20215-math-0012(10)
where ρa is the air density, Mw is the water molar mass, q(Tc) is saturation specific humidity at the canopy surface temperature (Tc), and qa is the specific humidity at the reference height.

2.3 C18OO Composition of Soil CO2 Flux

The C18OO composition of soil CO2 flux was estimated using an analytical formulation [Tans, 1998; Wingate et al., 2009]. Santos et al. [2012] used this formulation to estimate δR at the same site and observed agreement with measured δR using the isotope flux ratio method near the forest floor. This formulation takes into consideration the different processes controlling the C18OO budget, such as C18OO production and diffusion in the soil and exchange of oxygen molecules between CO2 and liquid water in the soil. These processes are described in the soil C18OO budget equation [Amundson et al., 1998; Hesterberg and Siegenthaler, 1991; Tans, 1998; Wingate et al., 2009]. Although Santos et al. [2012] conducted measurements of δR at the same site over the same time of year as this study, these measurements were not coincident in time (i.e., they were restricted to a few days a week). Hence, an analytical solution of the soil C18OO budget equation assuming steady state conditions and isothermal and uniform soil water conditions [Wingate et al., 2010; Wingate et al., 2009] was used to estimate δR. Although this is a simplification of field conditions, estimates of δR were in agreement with direct measurements of δR obtained using the isotope flux ratio method [Santos et al., 2012] at the site, providing confidence in modeled δR (‰, VPDB) values using the Wingate et. al. [2009] analytical formulation:
urn:x-wiley:21698953:media:jgrg20215:jgrg20215-math-0013(11)
where δeq,s (‰, VPDB) is the isotopic composition of CO2 in isotopic equilibrium with the soil water, FR (µmol m−2 s−1) is the soil CO2 flux, measured using soil chambers (as described in section 3.3), εd,eff is the effective isotopic fractionation during CO2 diffusion in soil pores [Wingate et al., 2009], and vinv (m s−1) represents the rate at which oxygen atoms in CO2 present in a column of air above the soil are exchanged with oxygen atoms in soil liquid water [Tans, 1998]. This term is given by urn:x-wiley:21698953:media:jgrg20215:jgrg20215-math-0014, where B is Bunsen's solubility coefficient for CO2, and a function of soil temperature (Ts) (B = 1.739 × exp(−0.0390 × Ts + 0.000236 × Ts2) [Weiss, 1974], θw is the soil water content, and D18 is the effective diffusivity of C18OO in soil air. The ks is the effective rate of oxygen exchange between CO2 and liquid water given by
urn:x-wiley:21698953:media:jgrg20215:jgrg20215-math-0015(12)
where fCA is the relative increase in hydration resulting from activity of the CA enzyme in the soil [Riley et al., 2003; Seibt et al., 2006; Wingate et al., 2008] and kh is the rate of oxygen isotope exchange between CO2 and water, equal to 1/3 × 0.037 × exp[0.118 × (Ts-25)] [Skirrow, 1975; Wingate et al., 2008]. Here we adopted a fCA value of 20, which resulted in the best agreement between measured and modeled δR at this site during the experimental period [Santos et al., 2012].
The effective diffusivity of CO2 in the soil air is given by
urn:x-wiley:21698953:media:jgrg20215:jgrg20215-math-0016(13)
where αd = 1 + εd/1000, where εd (−8.7‰) is the full kinetic fractionation in the soil pores, D25 is the molecular diffusivity of CO2 (1.4 × 10−5 m2 s−1) at 298 K, θa is the proportion of soil pores filled with air, θa = θsat − θw, where θsat is the soil water content at saturation estimated to be 0.46 m3 m−3 at the site [Saxton and Willey, 2006], b is the soil water retention parameter, determined to be 6.2 in this study [Cosby et al., 1984], T25 = 298 K, and n is 1.5 [Bird et al., 2002; Stern et al., 2001].
The isotope composition of CO2 in isotopic equilibrium with the soil water [Brenninkmeijer et al., 1983] is given by
urn:x-wiley:21698953:media:jgrg20215:jgrg20215-math-0017(14)
where δs,zeq is the isotopic composition of soil water at the equilibration depth (zeq), urn:x-wiley:21698953:media:jgrg20215:jgrg20215-math-0018, defined as the shallowest depth where full isotopic equilibration between water and CO2 molecules occurs [Wingate et al., 2009]. The H218O composition of soil water (δs) determined from soil samples throughout the experiment (section 3.2) was converted into the VPDB scale and used to estimate half-hourly δs. An interpolation procedure to estimate δs for times between samplings and exponential functions, adjusted using δs determined for three depths (section 3.2), was used to predict δs,zeq [Santos et al., 2012; Wingate et al., 2008].

3 Experimental Methods

3.1 Site Description

The field experiment was conducted in a temperate deciduous forest at the Environment Canada research station in Borden, ON, Canada (44°19′N, 79°56′W) from mid-June to August 2009. The forest at this site is approximately 100 years old and represents natural regrowth on abandoned farm land. A survey in 2006 indicated the predominant tree species at the site were red maple (Acer rubrum L.), eastern white pine (Pinus strobus L.), large-tooth aspen (Populus grandidentata Michx.), and white ash (Fraximus americana L.), with percentages of occurrence of 52.2, 13.5, 7.7, and 7.1%, respectively [Teklemariam et al., 2009]. The stand height was approximately 22 m, and the leaf area index (±SE) ranged from 3.9 (±0.13) to 3.6 (±0.15) m2 m−2 from June to September 2009. Textural analysis shows the soil at the site to be a loamy sand [Santos et al., 2012].

3.2 Isotope Measurements

Two tunable diode laser trace gas analyzers (TGA100, Campbell Scientific, Logan, UT, USA; hereafter TGA), one dedicated to CO2 and the other to H2O analysis, were used to measure the mixing ratios of 12C16O2, C18O16O, H216O, and H218O above the forest. The δa was expressed according to the delta notation, in reference to Vienna Peedee Belemnite or VPDB scale, and δv was expressed in reference to the Vienna Standard Mean Ocean Water (VSMOW) standard [Allison et al., 1995; Griffis et al., 2005].

The air was drawn continuously from two air intakes, set up at 25.8 and 36.8 m above the ground, to the TGA sampling systems, kept in temperature-controlled enclosures set up in trailers. In this study, the measurements taken at 25.8 m were used for the estimates of δA (section 2.1). Typical differences of CO2 concentration between the two air intakes above the canopy were quite small (< 5 µmol mol−1), so the use of concentration measurements obtained at 36.8 m would have negligible effect in δA calculations. To prevent condensation of water vapor in the sampling lines of the H2O isotope TGA, the air inlet tubes were heated as described by Lee et al. [2007]. In the CO2 isotopomer TGA, critical orifices located downstream of the heated air filters kept a low pressure in the sampling lines preventing condensation inside the tubing. Each intake was measured for 15 s during 4 min. At the end of each measurement cycle, gas was sampled from calibration tanks during 1 min. A detailed description of the calibration procedure of the CO2 isotopomer TGA is provided by Santos et al. [2012]. Water and CO2 isotope measurements were corrected for the nonlinearity of the TGAs using calibration gases as described in previous studies [Bowling et al., 2003; Griffis et al., 2005; Lee et al., 2007; Welp et al., 2008]. Simultaneous measurements of water and CO2 isotopes were available from day of year (DOY) 187 to 232. Mixing ratios of water vapor and CO2 isotopomers, used in this study, were measured during week days (Monday to Friday). Weekend measurements of CO2 isotopomers were taken near the forest floor and have been reported by Santos et al. [2012], who also provided further details on experimental setup at the site.

The H218O compositions of bulk leaf (δLb) and xylem (δx) water were determined for three tree species: red maple (Acer rubrum L.), large-tooth aspen (Populus grandidentata Michx.), and white ash (Fraxinus americana L.). Five to six red maple leaves and approximately 10 large-tooth aspen and white ash leaves were collected on each sampling day. The leaf samples had the center vein removed and were placed in sealed glass vials. Nongreen twigs from these species were also sampled, had the phloem removed, and were sealed in vials to determine δx. Leaf and twig samples were taken every 2–5 days at 12:00 EST during the experimental period excluding days when leaves were wet, for a total of 26 sampling times.

Soil samples were obtained near the flux tower at three depths: 5, 10, and 50 cm to determine the H218O composition of liquid water in the soil (δs). The soil samples were placed in sealed vials and kept refrigerated. The liquid water from soil samples was extracted using the cryogenic vacuum extraction method [Ehleringer and Osmond, 1989]. In addition, to determine the H218O composition of precipitation water (δp), rain water was also collected at the site on an event basis using a 15 cm diameter plastic funnel placed over a thermally insulated plastic bottle. Ground water samples were also collected from two wells at the site in two dates during the growing season to determine H218O composition of ground water (δg). After each precipitation event, the rain water was transferred from the plastic bottle to a sealed glass vial. The δLb, δx, δs δp, and δg of water samples were determined using the CO2 equilibration method on a gas bench auto sampler attached to a mass spectrometer (Delta Plus XL, Thermo Finnigan, Bremen, Germany) with precision of 0.1‰. The H218O composition of liquid water was expressed relative to the VSMOW scale in the delta notation.

3.3 Supporting Measurements

The eddy covariance technique was used to measure FN and latent heat flux above the forest. The three wind velocity components were measured using a sonic anemometer (SATI-Sx, ATI, Longmont, CO) set up at 33.4 m above the ground. A closed-path infrared gas analyzer (Li-6262, LI-COR, Lincoln, NE, USA) was used to determine the H2O and CO2 mixing ratios in the air at 33.4 m. The half-hourly eddy covariance CO2 flux included in our analysis met the following criteria: direction ranging from 90° to 255° and friction velocity larger than 0.45 m s−1, following Teklemariam et al. [2009] who also provide further details on the long-term eddy covariance measurements at the site. The eddy covariance data were used to optimize the physiological model described in Appendix A.

The soil CO2 flux was measured using an automated dark chamber (LI-8100, LI-COR) with a diameter of 20 cm. The forest understory vegetation was sparse, and small seedlings growing inside the chamber collars were excluded. Chamber measurements were taken every 15 min and averaged into 30 min intervals. To correct single point measurements for the effects of spatial heterogeneity, the soil CO2 flux was measured 2–3 times a week using the same gas analyzer and a portable chamber, at 30 locations near the flux tower. A linear regression model was then used to scale up the automated soil chamber measurements as described by Santos et al. [2012].

In addition, air temperature and relative humidity (HMP45A, Vaisala, Vantaa, Finland), photosynthetically active radiation (LI 190SB, LI-COR), and outgoing long-wave radiation (CNR1, Campbell Scientific) were measured at a height of 33.4 m. The precipitation was measured using a tipping bucked rain gauge (Belfort, Baltimore, MD) in a nearby open pond area. Soil temperature (Ts) and water content (θw) were measured using thermocouples (105T, Campbell Sci.) and soil moisture probes (CS615-L, Campbell Sci.), respectively, at a depth of 10 cm. Leaf wetness was monitored at 18 m and 10 m height using electronic flat plate sensors (237, Campbell Sci.), and these data were used to select times with no leaf wetness when interpreting model results.

4 Results and Discussion

4.1 Environmental Conditions and Isotope Compositions of CO2 and Water Vapor in the Air

Daily average air temperature increased throughout the experiment, ranging from 8.4 (DOY 152) to 26.2°C on DOY 230 (Figure 1a). The total rainfall from June to August 2009 was 251 mm, which was above the long-term average of total precipitation (228 mm) measured in these months from 1989 to 2006 at a nearby weather station (Egbert, ON, Canada). The loamy sand texture leading to good soil drainage at the site combined with frequent precipitation events (Figure 2a) was responsible for large variations in soil water content (Figure 1a), with values ranging from 0.08 (DOY 204) to 0.18 m3m−3 (DOY 223).

Details are in the caption following the image
Time series of (a) daily average air temperature (Tair, black solid line) measured at 33 m and soil water content (SWC, gray solid line) obtained at 10 cm depth, (b) hourly values of H218O composition of water vapor in the air (δv), and (c) half-hourly values of C18OO composition of the air (δa). δa and δv are expressed in reference to the VPDB and VSMOW scales, respectively, in the delta notation, and were measured at 25.8 m above a temperate deciduous forest in Borden, ON, Canada.
Details are in the caption following the image
Time series of total precipitation during (a) rain events and H218O composition of (b) rain (δP), (c) soil water (δs) at three depths 5 cm, 10 cm, and 50cm ground water (δg), (d) xylem water (δx), and (e) observed leaf bulk water (δLb) obtained from twig samples of white ash (Fraxinus americana L.), large-tooth aspen (Populus grandidentata Michx.), and red maple (Acer rubrum L.) in a temperate deciduous forest. The isotope composition of liquid water expressed relative to the VSMOW scale in the delta notation.

Values of δa, measured above the forest canopy, showed a clear diurnal pattern for most days, with half-hour values reaching a maximum in the early evening (18:00 EST) and a minimum in the early morning (05:00 EST) (Figure 1b). The causes for the δa diurnal pattern are the expansion of the boundary layer during the day and entrainment of more enriched air from atmospheric layers above the surface as well as the retrodiffusion of CO2 from the leaves to the air after isotopically equilibrating with foliage water [Griffis et al., 2005; Yakir and Sternberg, 2000].

Values of δv obtained above the forest showed large day-to-day variability, with hourly values varying from −27 to −11.2 ‰, and average δv equal to −19.7‰. Welp et al. [2012] used the oxygen and hydrogen isotope compositions of water vapor to study the mechanisms controlling the water vapor variability in the surface layer for six sites around the world, including our experimental period at the Borden forest. They postulated that evaporation from the Great Lakes during the summer months (June to August) added significant amounts of moisture to the surface layer at our study site.

4.2 Isotope Composition of Ecosystem Water Pools

The temporal variation of isotope composition of different ecosystem water pools, expressed in the VSMOW scale, is shown in Figure 2. Average δp, weighted by precipitation magnitude, was −7.7‰. However, large event-to-event variation in δp was observed during the experimental period, with values ranging from −14.6‰ (DOY 180 and 206) to −0.2 ‰ on DOY 171 (Figure 2b). Event-to-event variations in δp are related to differences in the trajectory and in the isotope composition of water sources of different storm systems [Welker, 2000]. At the Borden forest, the Great lakes are expected to be a significant source of moisture for precipitation during the summer [Bowen and Revenaugh, 2003].

Figures 2c and 2d show δs and δx variation during the growing season. Values of δs and δx showed a downward trend from the beginning of the measurements until approximately DOY 220 when values of these variables started to increase following large rainfalls on DOY 221 and 222, in which total precipitation was 49.4 mm and δp was more enriched than in the previous precipitation events. Large δs vertical gradients were not observed at this site (Figure 2c) in comparison to previous studies [Seibt et al., 2006]. The largest difference between δs values at 5 and 50 cm (3.6‰) was observed after precipitation events on DOY 221 and 222. Average δg was −12.3‰, and small variation in δg (0.3‰) was observed between sampling dates (Figure 2c).

Average δx values were −9.1, −8.9, and −8.4‰ for large-tooth aspen, white ash, and red maple, respectively (Figure 2d). The mean δx values were more negative than average values of δs: −7.3, −7.3, and −7.9‰ at 5, 10, and 50 cm, respectively. This may be an indication that all species utilized water from deeper layers in the soil, close to the water table, which had water more depleted in H218O (Figure 2c). A large spike in δx of ash (DOY 177) was observed after a precipitation event on DOY 171, in which δp was more enriched (0.2‰) than in previous rainfalls. This spike in δx indicates that the ash tree probably has a shallower rooting depth distribution than the other two species, which allows capturing short duration precipitation events with often more enriched precipitation water, while having also the ability to extract water from deeper soil layers, which was indicated by more negative values of δx than the values of δs observed at the soil surface. Values of δLb measured at midday showed a large day-to-day variability, ranging from −4.8 to 17.6‰ for large-tooth ash, −4.7 to 17.0‰ for aspen, and −2.7 to 21.8‰ for red maple. This variation is related to other variables besides δx and will be discussed in the next section.

4.3 Calculated Isotope Composition of Leaf Water

Comparisons between measured and modeled δLb, using δx of white ash, large-tooth aspen, and red maple, showed R2 ranging from 0.75 to 0.88 (CG) and 0.75 to 0.78 (FC) (Table 1). The RMSD for this comparison ranged from 1.8 to 4.6‰ (CG) and 2.0 to 2.1‰ (FC). Estimates provided by the FC model are sensitive to the effective diffusion length (L) of the Péclet effect (equation 9), which is a poorly constrained parameter in leaf isotope enrichment models and has been assumed to be species-specific [Ferrio et al., 2009; Song et al., 2013; Xiao et al., 2012]. In this study, we optimized L values to improve the agreement between δLb measurements and the predicted values obtained by the FC model using equation 7. The optimized L values ranged from 0.1 mm (ash and aspen) to 0.2 mm (maple). Our results suggest that the Péclet effect on H218O leaf enrichment is negligible at the canopy scale, which has also been reported in recent studies [Xiao et al., 2012; Xiao et al., 2010]. A negligible P implies that the foliage water pool is well mixed; however, previous studies show L does not scale up well from the leaf to the canopy scale and further investigation on the behavior of L in canopy-scale leaf water isotope composition models is needed [Griffis, 2013].

Table 1. Coefficient of Determination (R2), Root Mean Square Deviation (RMSD), and Mean Error (ME) for the Relationships Between Midday Measured (δLb) and Modeled Isotope Compositions of Leaf Water (Using the CG [Craig and Gordon, 1965] and FC [Farquhar and Cernusak, 2005] Models (Equations 4 and 7 With Measured Isotope Composition of Xylem Water) for White Ash (Fraxinus americana L.), Large-Tooth Aspen (Populus grandidentata Michx.) and Red Maple (Acer rubrum L.)
Tree CG Model FC Model
R2 RMSD (‰) ME (‰) R2 RMSD (‰) ME (‰)
Ash 0.88 3.6 3.3 0.75 2.1 −0.10
Aspen 0.75 4.6 4.1 0.78 2.0 −0.14
Maple 0.85 1.8 −0.4 0.78 2.0 0.04
Average 0.88 2.7 2.3 0.75 2.1 −0.10
  • The row labeled “Average” shows comparison of measured isotope composition of bulk leaf water averaged for the three species and modeled values obtained using the average isotope composition of xylem water in equations 4 and 7. The effective diffusion length (L) needed for the FC model was optimized for each species and for the average conditions.

Lee et al. [2009] demonstrated that turbulent mixing enhances the isotope kinetic fractionation at the canopy scale. The inclusion of aerodynamic resistance in the kinetic fractionation calculations improved the agreement between modeled and measured δLb in agricultural ecosystems [Griffis et al., 2011; Xiao et al., 2012]. Here, we evaluated the sensitivity of δLb estimates to turbulent effects by excluding ra from εk,w calculations (equation 5). The exclusion of ra resulted in small differences in model accuracy (RMSD variation < 1.2‰) for both CG and FC models (Table 1). The effect of turbulence on canopy-scale kinetic fractionation will be discussed further in the next section.

Figure 3 shows the comparison between observed midday δLb and predicted values using the CG and FC models. The CG model overestimated δLb indicating that a departure from steady state conditions occurred at our site at midday. Although, we could not evaluate the FC and CG model performance throughout the entire day because leaf water was sampled only at midday, our results indicate that nonsteady state effects are strong at the site. A larger departure from steady state is expected to occur at night (supplemental material) and has been confirmed in intensive sampling field campaigns of δLb [Welp et al., 2008; Xiao et al., 2012] and high temporal resolution (30 min) δLb data derived from branch chamber measurements [Wingate et al., 2010]. However, our results show a departure from steady state conditions even at midday conditions. The implications of nonsteady state conditions on δA calculation will be further explored in the next section.

Details are in the caption following the image
Comparison between modeled values of H218O composition of leaf water (VSMOW) assuming steady state (δLb,s equation 4, CG model) and nonsteady state (δLb equation 7, FC model) and average values obtained at midday from leaf samples of white ash (Fraxinus americana L.), large-tooth aspen (Populus grandidentata Michx.), and red maple (Acer rubrum L.) in a temperate deciduous forest.

4.4 Temporal Dynamics of Isotope Composition of Ecosystem CO2 Fluxes

Figure 4 shows half-hourly values of isotope composition of net CO2 assimilation and some of the variables used for its estimation during a selected week. A noteworthy feature in this graph is the coupling between δa and H218O composition of leaf water at the evaporative site (Figures 4a and 4b). During this selected period, the daytime correlation coefficients for the relationship between δa and δLe,s (CG model, equation 4) were r = 0.83 (p < 0.0001) and r = 0.72 (p < 0.0001) for the nighttime. However, when nonsteady state effects were considered, higher correlation coefficients were observed for δa and δLe (FC model, equation 6), with r = 0.90 (p < 0.0001) during the daytime and r = 0.63 (p < 0.0001) for nighttime. The coupling between δa and modeled δLe is an indication of the influence of the canopy on δa dynamics above the forest. The higher correlations between δa and δLe during the daytime are related to the larger exchange of CO2 between the vegetation, when canopy resistance was lower and the turbulent mixing was more intense.

Details are in the caption following the image
Half-hourly values of (a) C18OO composition of the air (δa) at 25.8 m; (b) H218O composition of leaf water at the evaporating site (VSMOW scale) assuming steady state (δLe,s black solid line, equation 4) and nonsteady state (δLe, gray solid line, equation 6) conditions; (c) CO2 retroflux of CO2, represented by the fraction Cc/(Cc − Ca); (d) resistances values of canopy (rc, solid black line), boundary layer (rb, dashed black line) and aerodynamic (ra, dashed blue line) to H2O diffusion; (e) CO2 canopy-scale kinetic fractionation factor (εk,c), calculated considering (black solid line) and ignoring (dashed line) turbulence effects, and (f) C18OO composition of the canopy CO2 flux (δA), calculated using δLe assuming steady state (black line) and nonsteady state (gray line) conditions. The shaded areas in the graph indicate night time periods.

The canopy CO2 retroflux represented by the ratio Cc/(Cc − Ca), estimated using a physiological model (Appendix A), showed a distinct diel cycle (Figure 4c), with more negative values during the nighttime, when the difference between Cc and Ca was smaller than during the day. The daytime rc,w estimates, provided by the physiological model, were in agreement with rc,w derived from the Penman-Monteith equation and eddy covariance measurements (Figure A1). In the nighttime, however, the physiological model rc,w estimates were higher than calculated using the Penman-Monteith equation, so we assumed a constant nighttime canopy resistance equal to the average rc,w derived from the Penman-Monteith equation (further details in Appendix A). Values of rc,w were about 1 order of magnitude larger than ra and rb,w (Figure 4d). Daytime average ra, rb,w, and rc,w were 18, 12, and 250 s m−1, respectively, while during the nighttime, average values for those resistances were 70, 13, and 1000 s m−1, respectively (Figure 4d). The average εk,c during the selected period was 8.3‰; however, εk,c variation was up to 0.8‰ between sequential half-hour periods (Figure 4e). When turbulence effects were not considered in εk,c calculations, average εk,c was slightly larger (8.7‰). The influence of the aerodynamic resistance on εk,c was limited in our study because of the small magnitude of the ratio between ra and rc. Over rough surfaces, such as forest canopies, the aerodynamic resistance is small and plays a minor role in the exchange of mass and energy relative to its influence on short vegetation surface-atmosphere exchange. Differences in turbulent mixing and transpiration rates between tall and short plant canopies can have a significant impact on the variables governing the isotopic exchange. Still et al. [2009] used a land surface model to study the isotopic exchange in a broad leave deciduous forest in a C4 grassland. Their results show that the relative humidity within the grass canopy was higher than the one determined for forest ecosystem, due to higher transpiration rates in grasslands. Higher relative humidity values resulted in more depleted leaf water isotopic composition in the grassland in comparison to that of the forest canopy, even when all the other forcing variables in their model were kept constant. So, turbulence effects can have a great impact on gas exchange in plant canopies and should always be considered in ecosystems scale studies.

Half-hourly values of δA, calculated using equation 2 and values of δLe, derived from the CG (equation 4) and FC (equation 6) models, are shown in Figure 4f. Differences between nighttime values of δA, calculated assuming steady state and nonsteady state conditions, were large (up to 27‰). During the daytime, the average difference between δA values calculated assuming steady state and nonsteady state conditions was 3.5‰. Although these daytime differences in δA values are much smaller than the differences in δA values obtained at the nighttime, these differences can be significant for CO2 flux partitioning, considering that canopy CO2 fluxes are large during the daytime and the typical values of Deq found for this site (data shown below).

Average modeled daytime δA was obtained from half-hourly δA, calculated using FC model δLe estimates, and excluding half-hour periods when the leaf wetness duration sensors indicated the presence of liquid water, and periods of inadequate fetch (Figure 5a). Average daytime δR was calculated using half-hourly δR estimated using equation 11. Daytime average values of modeled δR are shown in Figure 5a. The daytime average δR reached its minimum value (−24.6‰) on DOY 210 and then increased steadily until DOY 225 when average δR was −9.6‰. The temporal dynamics of δR were driven mainly by δs (Figure 2b). Precipitation events on DOY 221 and 222 resulted in significant enrichment of δs and consequently δR. Evaporative enrichment of δs can produce large vertical gradients of δs near the soil surface [Dubbert et al., 2013]. Due to budget constraints, we only measured δs at three soil depths, but future studies should consider a sampling scheme that include more depths specially close to the surface where δs variation is expected to be large.

Details are in the caption following the image
(a) Daytime average (± 1 standard deviation) C18OO composition of canopy net CO2 flux (δA) and net soil CO2 flux (δR); (b) and absolute daytime values of isotope disequilibrium, Deq = δR − δA, in a temperate deciduous forest from July to August 2009. In Figure 5b, δA was calculated assuming different values for the extent of the CO2 hydration in the leaves (θeq).

The range of modeled δR values obtained in this study are in agreement with observations reported by Santos et al. [2012] who observed half-hourly measured δR ranging from −31.4‰ to −11.2‰ for four selected periods during the growing season in this same ecosystem. These modeled δR are similar with the δR obtained using soil chamber for a corn ecosystem in Minnesota [Griffis et al., 2011] and at a southern boreal forest ecosystem in Canada [Flanagan et al., 1997] but much lower than δR values reported by Seibt et al. [2006] for a forest ecosystem in the UK and by Wingate et al. [2008] in a Mediterranean ecosystem in southern Portugal.

Average daytime δA showed large day-to-day variation throughout the experiment (Figure 5), with values ranging from −19.1 (DOY 195) to −11.3‰ on DOY 205. The average daytime δA for the whole period was −16.2‰. This average δA was used to calculate the C18OO canopy-scale discrimination factor (Δ), assuming Δ ≈ δa − δA [Lee et al., 2009], which ranged from 10.5 (DOY 189) to 20.8‰ (DOY 195). Wingate et al. [2010] reported a C18OO discrimination factor, obtained from branch chamber measurements, ranging mostly from 10 to 25‰ in a Maritime pine stand during the growing season. The values of Δ obtained in this study are in agreement with the canopy-scale discrimination factor values calculated by Griffis et al. [2011] for a corn canopy. They observed a decrease of the canopy evaporative isotope enrichment caused by a higher relative humidity and also reported lower δLb in their ecosystem during rain events. In our study, higher relative humidity also reduced δLb and δLe values (data not shown).

We evaluated the effect of lower values of θeq. on the C18OO disequilibrium (Deq) seasonal dynamics in this ecosystem. Daytime average δR and δA, modeled using equation 2 and three values of θeq, were used to calculate Deq, excluding half-hour periods when leaf wetness sensors indicated the presence of liquid water over the canopy or when wind direction did not provide adequate fetch for the site. Values of Deq showed a large temporal dynamics throughout the season, with highest values observed near DOY 210, when the lowest δR values were estimated for the experimental site (Figure 5b). The magnitude of Deq was strongly influenced by the value of θeq. Average absolute values of Deq for the whole period were 4.8, 7.4, and 10.6‰ for θeq values of 0.96, 0.60, and 0.25, respectively. A strong Deq was observed on DOY 210, when the isotope disequilibrium at the site reached 18.7, 16.1, and 9.4‰ assuming θeq equal to 0.25, 0.6, and 0.96, respectively. Throughout the season, the magnitude of Deq was indirectly proportional to θeq, except toward the end of the growing season, on DOY 225 and 226, when highest absolute Deq values (8.5‰) were obtained for θeq = 0.96.

The inverse relationship between θeq and Deq was not expected, based on previous studies The expected relationship is that higher hydration efficiencies in the canopy, i.e., higher θeq, result in higher rates of oxygen atom exchange between CO2 and enriched water molecules in the foliage. When these CO2 molecules diffuse back to the atmosphere, the difference between δA and δR is expected to increase, resulting in higher magnitude in Deq. However, depleted 18O precipitation at our site throughout the growing season (Figure 2b) resulted in quite negative δR values, which were often lower than δA (Figure 5a). Under this condition (δA > δR), the increase of hydration efficiency would result in more negative values of δA (equation 2), so δA would approach δR therefore producing lower Deq as θeq approaches 1. The expected relationship between θeq and Deq was observed after DOY 219, when precipitation events with more enriched δp resulted in more positive δR (Figure 5b). Consequently, increased θeq lead to more negative δA and resulted in a larger absolute Deq. This inverse relationship between θeq and Deq is expected to occur in mid latitude ecosystems, where δR is expected to be much larger than the values observed in this study [Wingate et al., 2008].

Previous studies have shown a relationship between the magnitude of Deq and precipitation events. Sturm et al. [2012] observed a reduction in δA caused by precipitation in a mixed deciduous forest. They attributed this decrease in δA to lower canopy evaporative isotope enrichment due to higher relative humidity during rainy days. Griffis et al. [2011] observed a seasonal variation in C18OO disequilibrium in a corn canopy and low Deq during rainy periods. Wingate et al. [2010] also reported a reduction in Deq during rain events, because the precipitation water reset the ecosystem water pools to the same isotopic composition as δp. However, our results show that precipitation events during the growing season caused an increase of Deq at the Borden Forest. The magnitude of Deq was closely related to δp, which was the main variable affecting δs (Figures 2b and 2c). Low values of δs near the soil surface (5 cm depth) on DOY 210 were responsible for a significant reduction in δR (Figure 5b), which led to a strong Deq over the same period (Figure 5b). These results point to a need for ecosystem specific assessments of Deq and consideration of the temporal dynamics of Deq.

The strong predicted Deq indicates that C18OO has a great potential to be used as tracer to study carbon exchange in natural ecosystems. However, the large temporal variability of δp at our site shows that an event-based precipitation sampling strategy is required to interpret the temporal dynamics of soil and plant isotope signals in this ecosystem. In addition, to improve the accuracy in Deq estimates, a better understanding of the mechanisms controlling the extent of canopy CO2 hydration is needed. Our results show that Deq is highly sensitive to θeq (equation 2), emphasizing the need for canopy-scale studies to investigate the extent of CO2 hydration at the ecosystem scale and appropriate schemes to scale up variables derived from leaf observations to canopy-scale isotope models.

5 Conclusions

Bulk leaf water isotope composition showed large day-to-day variability at our research site with values ranging from −2.7 to 21.8‰. Leaf water enrichment simulations indicate that nonsteady state effects were important even at midday in your site and should be considered when estimating δA. The inclusion of turbulence effects resulted in small improvement of δLb estimates in this study, which is related to the small magnitude of the aerodynamic resistance compared to the relatively large canopy resistance in the forest.

Values of δA showed large day-to-day variation at our site, with daytime average values ranging from −19.1 to −11.3‰. The dynamics of δR were mainly driven by precipitation events with variable δp. Large Deq values (up to 11‰ for θeq = 0.96) were observed at our site; however, the magnitude of the disequilibrium was variable throughout the season, and near zero Deq values were observed near the end of the experimental season for θeq = 0.96. The temporal dynamics of Deq was largely driven by the variability of δp at the site, which was demonstrated by our high-frequency sampling scheme for precipitation water. The magnitude of Deq was also very sensitive to the hydration efficiencies in the canopy. For this temperate forest during most of the growing season, the magnitude of Deq was inversely proportional to θeq, due to the very negative δR signal, which is contrary to observations for other ecosystems investigated in previous studies. The large Deq values observed at our site indicate that C18OO can be a very powerful tool for partitioning plant and soil components of the net CO2 ecosystem exchange. However, additional work is still needed to understand better the mechanisms controlling the extent of CO2 hydration in foliage and soils.

List of Symbols

  • Am
  • photosynthetic rate at saturating light intensity, mg CO2 m−2 s−1.
  • Am,max
  • maximum photosynthetic rate at saturating light intensity, mg CO2 m−2 s−1.
  • Amin
  • residual photosynthesis rate, mg CO2 m−2 s−1.
  • An
  • the leaf net assimilation, mg CO2 m−2 s−1.
  • B
  • Bunsen solubility coefficient for CO2, m3 of air m−3 of water.
  • B
  • slope of the soil water retention function.
  • C
  • density of liquid water, mol m−3.
  • Ca
  • CO2 molar concentration in the air, µmol m−3.
  • Cc
  • CO2 molar concentration in the chloroplasts, µmol m−3.
  • Ci
  • CO2 molar concentration in the intercellular space, µmol m−3.
  • D
  • diffusivity of H218O in water, m2 s−1.
  • D18
  • effective diffusivity of C18OO in the soil air, m2 s−1.
  • D25
  • the molecular diffusivity of CO2 at 298 K, m2 s−1.
  • Deq
  • C18OO disequilibrium, ‰ VPDB.
  • Dmax
  • maximum leaf-to-air saturation deficit, g kg−1.
  • Ds
  • leaf-to-air saturation deficit, g kg−1.
  • ET
  • transpiration rate of the foliage, mol m−2 s−1.
  • F
  • intercellular and atmospheric CO2 concentration factor.
  • f0
  • maximum ratio of the intercellular to the atmospheric CO2 concentration.
  • FA
  • canopy CO2 flux, µmol m−2 s−1.
  • fCA
  • relative increase in CO2 hydration resulting from the carbon anhydrase enzyme activity in the soil.
  • FN
  • Net CO2 ecosystem exchange, µmol m−2 s−1.
  • FR
  • soil CO2 flux, µmol m−2 s−1.
  • gc,c
  • canopy-scale conductance to CO2, m s−1.
  • gl,c
  • leaf-scale conductance to CO2, m s−1.
  • gm
  • unstressed mesophyll conductance, m s−1.
  • gmin
  • cuticular conductance to water vapor, m s−1.
  • h
  • relative humidity.
  • k
  • von Karman constant.
  • kh
  • rate of oxygen isotope exchange between CO2 and water, s−1
  • ks
  • effective rate of oxygen exchange between CO2 and liquid water, s−1.
  • L
  • effective diffusion length, m.
  • LAI
  • he leaf are index, m2 m−2.
  • lw
  • leaf dimension, m.
  • Mw
  • water molar mass, kg mol−1.
  • P
  • Péclet number.
  • PAR
  • photosynthetically active radiation, J m−2 s−1.
  • q(Tc)
  • saturation specific humidity at the canopy surface, kg m−3.
  • qa
  • specific humidity at the reference height, kg m−3.
  • ra
  • aerodynamic resistance, s m−1.
  • rb,c
  • boundary layer resistance to CO2, s m−1.
  • rb,w
  • boundary layer resistance to H2O, s m−1.
  • rc,c
  • canopy resistance to CO2, s m−1.
  • rc,w
  • canopy resistance to H2O, s m−1.
  • Rd
  • dark respiration, mg CO2 m−2 s−1.
  • U
  • wind speed, m s−1.
  • u*
  • friction velocity, m s−1.
  • vinv
  • invasion velocity, m s−1.
  • W
  • foliage water content, mol m−2.
  • wi
  • mole fraction of (light) water in the intercellular space, mol mol−1.
  • zeq
  • equilibration depth, m.
  • zm
  • measurement height, m.
  • z0
  • roughness length, m.
  • αk
  • fractionation factor.
  • δA
  • C18OO composition of canopy of CO2 flux, ‰ VPDB.
  • δa
  • C18OO composition of the air, ‰ VPDB.
  • δeq,s
  • isotopic composition of CO2 in isotopic equilibrium with the soil water, ‰ VPDB.
  • δg
  • H218O composition of ground water, ‰ VSMOW.
  • δLb
  • H218O composition of leaf bulk water, ‰ VSMOW.
  • δLb,s
  • H218O composition of leaf bulk water assuming steady state conditions, ‰ VSMOW.
  • urn:x-wiley:21698953:media:jgrg20215:jgrg20215-math-0033
  • H218O composition of leaf bulk water assuming steady state conditions, calculated taking into consideration the Peclet effect, ‰ VSMOW.
  • δLe
  • H218O composition of leaf water at the evaporation sites, ‰ VSMOW.
  • δLe,s
  • H218O composition of leaf water at the evaporation sites assuming steady state conditions, ‰ VSMOW.
  • δR
  • C18OO composition of soil CO2 flux, ‰ VPDB.
  • δs
  • H218O composition of soil water, ‰ VSMOW.
  • δs,zeq
  • H218O composition of soil water at the equilibration depth, ‰ VSMOW.
  • δv
  • H218O composition of water vapor, ‰ VSMOW.
  • δx
  • H218O composition of xylem water, ‰ VSMOW.
  • ε
  • light conversion efficiency, mg J−1.
  • εd
  • full kinetic fractionation in the soil pores, ‰.
  • εd,eff
  • effective isotopic fractionation during CO2 diffusion in soil pores, ‰.
  • εeq
  • water equilibrium fractionation, ‰.
  • εk,c
  • CO2 canopy-scale kinetic fractionation factor for C18OO, ‰.
  • εk,w
  • canopy-scale kinetic fractionation factor for H218O, ‰.
  • θa
  • proportion of soil pores filled with air, m3 m−3.
  • θeq
  • CO2 hydration efficiency.
  • θsat
  • the soil water content at saturation, m3 m−3.
  • θt
  • total CO2 porosity, m3 m−3.
  • θw
  • soil water content, m3 m−3.
  • ρa
  • air density, kg m−3.
  • ΦM
  • stability function.
  • Г
  • CO2 compensation point, mg m−3.
  • Acknowledgments

    Funding for this research was provided by the Natural Science and Engineering Research Council. The first author acknowledges funding by the Brazilian National Council for Scientific and Technological Development (CNPq). Xuhui Lee was supported by the US National Science Foundation (grant ATM-0914473) and the Ministry of Education of China (grant PCSIRT). We would also like to thank the two anonymous reviewers for their insightful suggestions.

      Appendix A

      A plant physiological model was used to estimate some physiological variables, such as rc,c and Cc used to calculate δA and δL,e (section 2.1). In this approach, the canopy-scale stomatal resistance is derived from the net CO2 assimilation, which is a function of environmental variables such as water vapor deficit, photosynthetically active radiation, and canopy temperature [Calvet et al., 2004]. The leaf-scale conductance for CO2 (gl,c) is expressed as
      urn:x-wiley:21698953:media:jgrg20215:jgrg20215-math-0019(A1)
      where gmin is the cuticular conductance for water vapor—in this study we adopted the gmin value (0.15 mm s−1) found by Calvet et al. [2004] for woody broadleaved species—An is the leaf net assimilation, Am is photosynthetic rate at saturating light intensity, Amin is the residual photosynthesis rate, Ds is leaf-to-air saturation deficit (g kg−1), Dmax is the maximum Ds in well-watered conditions, Rd is the dark respiration, and Rd = 0.11Am, where Am is given by
      urn:x-wiley:21698953:media:jgrg20215:jgrg20215-math-0020(A2)
      where gm is the unstressed mesophyll conductance and Г is the CO2 compensation point. The variables gm, Г, and Am,max are functions of the canopy temperature and were estimated using Q10-type functions as described by Jacobs [1994]. The leaf internal CO2 concentration was estimated using the following closure equation:
      urn:x-wiley:21698953:media:jgrg20215:jgrg20215-math-0021(A3)
      where f is the coupling factor given by
      urn:x-wiley:21698953:media:jgrg20215:jgrg20215-math-0022(A4)
      where f0 is the maximum ratio of the intercellular to the atmospheric CO2 concentration, when Ds is zero, f = f0.
      The net assimilation is limited by photosynthetically active radiation (PAR) and is calculated as
      urn:x-wiley:21698953:media:jgrg20215:jgrg20215-math-0023(A5)
      where ε is the light conversion efficiency, ε = ε0(Ci − Г)/(Ci − 2Г), where ε0 = 0.017 mg CO2 J−1.
      The residual photosynthesis rate is expressed as
      urn:x-wiley:21698953:media:jgrg20215:jgrg20215-math-0024(A6)
      The parameters fo and Dmax were tuned [Calvet et al., 2004; Xiao et al., 2010] to improve the agreement between measured and simulated values of canopy CO2 flux. Measured canopy CO2 flux was assumed to be the difference between the net CO2 ecosystem exchange and soil CO2 flux, obtained using the eddy covariance technique and soil chamber measurements, respectively. The canopy CO2 flux was estimated as follows:
      urn:x-wiley:21698953:media:jgrg20215:jgrg20215-math-0025(A7)
      The aerodynamic resistance (ra) was estimated based on the logarithmic wind profile:
      urn:x-wiley:21698953:media:jgrg20215:jgrg20215-math-0026(A8)
      where zm is the measurement height, z0 is the roughness length, k is the von Karman constant (k = 0.4), u is the wind speed at the measurement height, and ΦM is the dimensionless stability function.
      The boundary layer resistance (rb,c) is given by
      urn:x-wiley:21698953:media:jgrg20215:jgrg20215-math-0027(A9)
      where br = 283 s0.5 m−1, lw is leaf dimension, and LAI is the leaf are index.
      The canopy resistance (rc,c) was estimated from scaling up gl,c [Ronda et al., 2001], which integrates An and Rd over the canopy, assuming an exponential decay of PAR within the canopy as follows:
      urn:x-wiley:21698953:media:jgrg20215:jgrg20215-math-0028(A10)
      where a is equal to 1/(1 − f0), D* is D0/(a − 1) and L is the leaf area. An analytical solution for equation A10 is provided by Ronda et al. [2001].
      The CO2 concentration in the chloroplasts of the leaves (Cc), used to estimate the isotope composition of the canopy CO2 flux (equation 2), was calculated according to Xiao et al. [2010] as follows:
      urn:x-wiley:21698953:media:jgrg20215:jgrg20215-math-0029(A11)
      where rm is the canopy-scale mesophyll resistance, given by rm = 1/(gmLAI). Average daytime (10:00–15:00 EST) values of Ca, Ci, and Cc during the experiment were 378.4, 202.2, and 126.3 µmol mol−1. Typical differences between Ci and Cc are within the range of values reported in the literature [Flexas et al., 2008; Keenan et al., 2010].

      Optimized values of fo and Do found for this study were 0.55 and 33 g kg−1, respectively, which are within the range of values reported for broad leaved woody species [Calvet et al., 2004]. Comparisons between Fc derived from flux measurements and estimated using the big-leaf model showed R2 = 0.85 and root mean square error of 3.4 µmol m−2 s−1. The inclusion of the soil water stress corrections proposed by Calvet et al. [2004] in the plant physiological model calculations did not improve Fc estimates, which indicates that the vegetation at the site was not under water stress.

      To evaluate the performance of this physiological approach, the canopy resistance was compared to estimates obtained from the Pennan-Monteith (PM) equation and eddy covariance measurements as follows:
      urn:x-wiley:21698953:media:jgrg20215:jgrg20215-math-0030(A12)
      where rc,w is the canopy resistance to H2O diffusion, given by rc,w = rc,c/1.6, D is the vapor density deficit, urn:x-wiley:21698953:media:jgrg20215:jgrg20215-math-0031 is the water vapor flux, urn:x-wiley:21698953:media:jgrg20215:jgrg20215-math-0032 is the sensible heat flux, s is the slope of the saturation vapor density curve, and rt is total resistance (rt = ra + rb,w), where rb,w is the boundary layer resistance to water vapor diffusion given by rb,w = rb,c/1.4. For the calculation of rc,w using equation A12, we exclude periods, in which the wind speed did not satisfy the fetch conditions for the site, periods with u* < 0.45 m/s, and when leaf wetness sensors indicate the presence of water in the forest canopy.

      Figure A1 shows the comparisons between half-hour daily ensemble average of rc,w calculated using the physiological model and the PM equation. During the daytime, reasonable agreement was obtained by the two methods, indicating that the physiological model used at this study was suitable to estimate daytime rc,w. However, at nighttime the physiological model overestimated rc,w, when compared to half-hour rc,w values estimated using PM equation. This could be related to limitations of the approach used to scale up leaf resistance to the canopy scale. To overcome this problem, we adopted a constant value of resistance at nighttime equal to the average rc,w (1001 s m−1) calculated using the PM equation during dew and rain free nights.

      Details are in the caption following the image
      Ensemble average half-hourly values of canopy resistance to water vapor (rc,w, ± 1 standard deviation) modeled using a physiological model and derived from the Penman-Monteith (PM) equation.