Tree-ring δ¹³C tracks flux tower ecosystem productivity estimates in a NE temperate forest

The forest within the flux-tower footprint (1 km²; Harvard Forest-EMS tower) is mainly deciduous and is dominated by Quercus. rubra (L.) (northern red oak; 52% of basal area), Acer rubrum L. (red maple; 22% of basal area) and Tsuga canadensis (L.) (eastern hemlock; 17% of basal area). Increment core sampling focused on canopy dominant trees. To identify interannual variability in tree growth and tree-ring δ¹³C, we cored Q. rubra and T. canadensis with a 5 mm increment borer at 1.37 m stem height (table 1). Two cores were taken from each tree to build ring-width chronologies. Additionally, we cored five and four trees respectively of each species at the same height with a 12 mm increment borer. The large cores were used to analyze the δ¹³C in the LW of the tree rings. A sample of four trees is sufficient to achieve good precision of the δ¹³C sample mean (McCarroll and Loader 2004, Leavitt 2008).

Ring-width chronologies were developed for each species using standard dendrochronological techniques (Speer 2010). For the site, Q. rubra mean age was 97 years (range 71–115 years) and T. canadensis mean age was 145 years (range 89–221 years). For isotope cores, Q. rubra mean age was 103 years (range 87–113 years) and T. canadensis mean age was 142 years (range 93–188 years). All ages were estimated from inner ring dates.

Basal area increment (BAI) was calculated for each tree and species using the dplR package in R (Bunn 2008). We used the 'outside in' function to convert raw ring-width measurements to BAI based on the diameter of the tree and the width of each ring moving towards the pith of the tree. The method assumes a circular growth pattern. BAI was used instead of ring width as a surrogate of radial growth and carbon gain because it represents more accurately tree annual biomass increment without the need for standardization (Biondi and Qeadan 2008). A mean BAI value was computed for each species at an annual resolution, averaged over all the trees cored for both carbon isotope analyses and ring-width chronologies (i.e., 16 trees/42 cores for Q. rubra and 22 trees/55 cores for T. canadensis).

We analyzed the LW portion of each tree ring from individual trees over the 1992–2010 period (table 1). The samples were milled using an ultra-centrifugation mill (Qiagen TissueLyserII) and α-cellulose was extracted from each wood sample following the Soxhlet method elaborated by Green (1963) and modified by Leavitt and Danzer (1993). δ¹³C ratios were measured on the CO₂ produced by α-cellulose combustion in a Costech elemental analyzer coupled with a Thermo Delta V-IRMS. The sample accuracy was determined to be ±0.08‰ (1 σ standard deviation calculated from the average difference between measured and true internal standard, n = 13). The isotopic value is expressed in the delta (δ) notation relative to the VPDB (‰ VPDB).

We used δ¹³C to determine carbon isotope discrimination (Δ) by the plant against atmospheric δ¹³C, and variation in plant iWUE. The Δ describes the isotopic difference between the δ¹³C of air (δ¹³C_air) and the plant (δ¹³C_plant) and results from the preferential use of ¹²C over ¹³C during photosynthesis. Δ is calculated using:

$$\begin{eqnarray}&&\Delta =\left( {{d}^{13}}{{C}_{{\rm air}}}-{{\delta }^{13}}{{C}_{{\rm plant}}} \right)/\left( 1+{{\delta }^{13}}{{C}_{{\rm plant}}}/1000 \right).\end{eqnarray} \tag{ 1 }$$

Records of δ¹³C_air were obtained from Mauna Loa from 1992 to 2002, and from aircraft measurements collected in Worcester, Massachusetts at 500 m above ground and available from 2003 to 2010 (White and Vaughn 2011). The model of Farquhar et al (1982) describes isotopic Δ for C₃ plants via:

$$\begin{eqnarray}&&\Delta =a+\left( b-a \right){{c}_{i}}/{{c}_{a}},\end{eqnarray} \tag{ 2 }$$

where a is the fractionation during CO₂ diffusion through the stomata (4.4‰: O'Leary 1981); b is the fractionation by RuBP carboxylase (27‰: Farquhar and Richards 1984); and c_i and c_a , are the leaf intercellular space and ambient CO₂ concentrations (μmol Mol⁻¹), respectively. We used the data of atmospheric CO₂ concentration (c_a) measured at 29 m height on the eddy-covariance tower to calculate c_i from equation (2). The fractionations due to diffusion and carboxylation are constant but additive; therefore, the δ¹³C records variations in c_i as regulated by two main processes: stomatal conductance (c_a/c_i) and photosynthetic assimilation rate (A). The iWUE is the ratio of the net photosynthetic assimilation rate (A) and water vapor conductance (g_H2O) and is described by Ehleringer and Cerling (1995) as:

$$\begin{eqnarray}&&iWUE=A/{{g}_{H2O}}={{c}_{a}}-{{c}_{i}}\left( 1/1.6 \right).\end{eqnarray} \tag{ 3 }$$

The iWUE derived from plant isotope data is used to compare photosynthetic properties independent from evaporative demand (Osmond et al 1980), and is therefore, often applied as an indicator of long-term trends in the internal regulation of carbon uptake and water loss of plants (Seibt et al 2008). We used summer (June, July and August) values of δ¹³C_air and c_a to calculate Δ, and to reconstruct c_i and iWUE for the period corresponding to LW formation.

We used hourly gap-filled meteorological data from the eddy-covariance tower measured by sensors above the canopy at 29 m height (Urbanski et al 2007, Harvard Forest Data Archive HF004) to compute monthly mean temperature, vapor pressure deficit (VPD), incident photosynthetically active radiation, relative humidity as well as monthly total precipitation from 1992 to 2010. Hourly gap-filled GPP was used to provide monthly cumulative GPP. We used the Palmer Drought Severity Index (PDSI) for the central Massachusetts climate region from the National Climatic Data Center.

The association between climate variables and tree-ring Δ chronologies were assessed using correlation analyses. The GPP time series during the growing season (from May to October) were compared to tree-ring Δ chronologies for each species. Correlations were calculated between monthly GPP and LW-Δ using the Pearson (pairwise) product-moment correlation. Trends over the 1992–2010 period were estimated using linear regression.

The carbon contribution to stem growth during LW formation should primarily originate from the immediate product of carboxylation thus reflecting the δ¹³C signature of recent assimilates (Helle and Schleser 2004). The interannual variability of Δ was compared with monthly GPP to identify the seasonality of carbon uptake and allocation to aboveground woody biomass. The correlations shown in figure 2 indicate that recent photosynthates provided the substrates for stem radial growth occurring in the second part of the growing season (Gessler et al 2009, Offermann et al 2011). The Δ-GPP correlations are logical based on the timing of LW formation for both T. canadensis and Q. rubra. LW tracheid formation and cell wall thickening in T. canadensis begins later in the growing season (late July–October, Skene 1972) coincident with the significant correlations. In contrast, LW formation in Quercus species begins approximately five weeks after the unfolding of first leaves and continues until radial growth cessation in late summer or early fall (Zasada and Zahner 1969, Voelker et al 2012).

Zoom In Zoom Out Reset image size

The Δ-GPP correlations are consistent with the seasonality of carbon uptake from eddy flux measurements and C storage estimates at the deciduous forest-EMS tower site and the Hemlock Forest tower site located 0.5 km from the EMS tower (Hadley and Schedlbauer 2002, Hadley et al 2009). Net C uptake at the deciduous forest site occurs between the end of May and mid-October. At the hemlock site, the peak in net C uptake and storage occurs in April and May. This period is not influencing the δ¹³C in the LW, although it shows the effect of conifers on the annual pattern of the C exchange. Conifers currently represent 17% of the total basal area at the deciduous forest site. After C uptake declines in June, it increases in July, and falls in August before a second uptake peak in October, consistent with the Δ-GPP correlations for T. canadensis.

The species seasonality of carbon allocation is further reflected in the relationship between LW-Δ and growth. The correlation for Q. rubra LW-Δ and mean BAI was significant (r = 0.72, P < 0.001), and was slightly higher with the mean Q. rubra BAI of all sampled trees (r = 0.75, P < 0.001). To illustrate the relationship between aboveground carbon gain and photosynthetic discrimination we regressed BAI against Δ, and a strong positive trend was obtained for Q. rubra (R² = 0.39 P < 0.01). The growth increase of Q. rubra could be interpreted as the result of a greater photosynthetic rate induced by increased c_i concentrations (von Caemmerer and Farquhar 1981, Schubert and Jahren 2012), and a substantial allocation of carbohydrates to LW formation and radial growth (Keel et al 2006, Palacio et al 2011). However stage of stand development may also be contributing to this increase in BAI (e.g. Foster et al 2014, figure S1). A relationship between Δ and BAI was not found for T. canadensis, probably because carbon sequestered after June contributes little to the current year radial growth or that carbon allocation differs by species (Richardson et al 2013). The trend in the BAI growth enhancement for Q. rubra is consistent with changes in the measured annual increment of aboveground biomass in the flux-tower footprint from 1993–2010 (∼20% increase, Data Archive: HF069), and provides a quantification of carbon stored in aboveground woody biomass. Interestingly, the aboveground biomass over the last decade accounted for ∼50% of the total carbon sequestered (Keenan et al 2012a) confirming the relationship we found between Q. rubra BAI and Δ, and growing season GPP (figure 2). The remaining carbon uptake could be attributed to litter or soil pools; however, these carbon proportions and stocks are not accounted for in the tree-rings δ¹³C.

The inter-annual variability and fraction of carbon uptake and sequestration explained by the tree-ring derived Δ (figures 2(c), (d)) provides a quantitative proxy that can be used to constrain models that estimate forest productivity. Process-based models fail to accurately reproduce the observed inter-annual variability of C-fluxes (Keenan et al 2012b) mainly because of inaccurate model allocation structure and lagged effects of climate variability on tree growth and physiology (Gough et al 2009). In contrast, the Δ integrates biotic factors which modulate the δ¹³C fractionation and the isotopic composition of the total pool of carbon fixed during photosynthesis at seasonal and annual cycles (Leavitt 1993).

We examined the climatic drivers of Δ using meteorological data from the eddy-covariance tower and PDSI. The climate signal within both species Δ series is dominated by monthly precipitation and PDSI during the growing season (table S1), although they have a stronger influence on T. canadensis due to its particularly drought-sensitive nature caused by a relatively shallow rooting-depth (Cook and Cole 1991).

The carbon isotope discrimination properties can be used to assess how trees are responding to increasing atmospheric concentration c_a and changes in climate. Three physiological response scenarios described by Saurer et al (2004) identify possible changes in iWUE and c_i following c_a increase: (1) c_i remains constant such that c_i/c_a decreases and iWUE increases; (2) c_i increases proportional to c_a such that c_i/c_a remains constant and iWUE increases; (3) c_i increases at the same rate as c_a, c_i/c_a increases and iWUE remains constant. In this study, Δ and c_i increased and there was no increase in iWUE for both species (figure 2, table 2). The c_i/c_a calculated for both species increases indicating c_i follows the increase in c_a at a rate of 2 ± 0.21 ppm yr^–1 for T. canadensis, and 1.7 ± 0.17 ppm yr^–1 for Q. rubra. The c_i/c_a increase and no change in iWUE suggest a weak stomatal response to c_a increase and is described as a passive response to changes in c_a (McCarroll et al 2009) where neither stomatal conductance nor photosynthetic rate change. The most common response documented in trees is an active response as stomatal conductance is reduced (Bert et al 1997, Saurer et al 2004, Peñuelas et al 2011). The iWUE trends derived from the tree-rings δ¹³C in our study show a divergence from the instantaneous WUE (the ratio of carbon assimilation to transpiration, Farquhar and Richards 1984) trend documented at the ecosystem level using continuous measurements of CO₂ and water vapor fluxes at HF (Keenan et al 2013). The canopy-integrated water use efficiency calculated as the ratio between gross ecosystem photosynthesis (GEP) and Ecosystem evapotranspiration (E) shows a significant increase over the measurements period (3.5% change yr⁻¹, P = 0.01). The latter was calculated taking into account the atmospheric evaporative demand and therefore is referred to as 'inherent' water use efficiency (W_ei). The 'intrinsic' water use efficiency as derived from tree-rings δ¹³C is often used as an indicator of long-term trends in the internal regulation of carbon uptake and water loss independently from evaporative demand (Osmond et al 1980). Therefore iWUE inferred from δ¹³C may not be representative of instantaneous or inherent WUE. Although both are affected by similar processes (i.e. stomatal conductance), they can vary independently because they are influenced by additional factors like mesophyll conductance, leaf N-content and C-respiratory losses (Griffiths et al 1999, Seibt et al 2008). We note that when iWUE is calculated from the flux tower measurements (as the ratio between GEP and canopy water conductance), the magnitude of the trend is lower and the statistical strength is no longer significant compared to those calculated for W_ei (1% change yr⁻¹, P = 0.2, Keenan et al 2013).

The Δ trend found in this study during the last 18 years is unusual but not unique. Results from literature often reported a decrease in Δ inferred from tree-ring records, and thus a strong improvement of iWUE during the last 100 years under increased atmospheric CO₂ (Arneth et al 2002). However, Marshall and Monserud (1996) showed that iWUE has remained static in sites from western US and a switch in iWUE from an active to a passive response around AD 1970 has been noted in several European sites (Gagen et al 2011). The spatial variability of stomatal regulation and iWUE response to increasing c_a is explained by the strong dependence of water and carbon usage on moisture stress experienced at a specific site and species sensitivity to moisture availability (Warren et al 2001). The increased discrimination is a reflection of the availability of CO₂. When the level of CO₂ (c_i ) is high, large discrimination is observed reflecting the RuBisCO enzymatic preference for ¹²CO₂ (Stewart et al 1995). Clearly, the Δ chronologies at HF indicate that despite higher c_a concentrations, factors like physiology and site condition contributed to stomatal openness and conductance.

The strong correlation between Δ and growing season PDSI and precipitation for both species suggest a link between water availability and stomatal conductance (Dupouey et al 1993). The climatic data for central Massachusetts indicates that precipitation and moisture (PDSI) are high at HF and are rarely limiting tree growth (Voelker 2011). In fact, the region became wetter over the period of analysis. The mean annual precipitation from 1898 to 2010 was 107.45 cm and increased 0.28 ± 0.06 cm yr^–1 (p < 0.0001). The trend in precipitation is also reflected in a trend of increasing PDSI of 0.03 ± 0.005 units yr^–1 (P < 0.0001) over the same period. Since 1992, precipitation and PDSI have increased at rates of 1.30 ± 0.67 cm yr^–1 (P < 0.07) and 0.15 ± 0.05 units yr^–1 (P < 0.006), respectively. The increase in regional scale water availability (Wang et al 2013) is also evident at the HF site. Annual precipitation and growing season precipitation have increased by 10% and 15% over the last 18 years, respectively (Keenan et al 2013).

We conclude on the basis of the photosynthetic discrimination sensitivity to water availability that the long-term increase in moisture which controls Δ drove the increase in stomatal conductance and carbon photosynthetic discrimination. A positive correlation between Δ and mean annual and summer precipitation has been confirmed for modern leaf tissues from multiple biomes in North America (Diefendorf et al 2010). This correlation is explained by the effect of the mean annual and summer precipitation levels on both soil water status and VPD (Hartman and Danin 2010). It is noteworthy that despite the mesic conditions of the HF site, precipitation and soil moisture levels remain very strong predictors of CO₂ discrimination. Similar to our findings, drought sensitivity has been demonstrated for radial growth of T. canadensis and Q. rubra growing in mesic sites located in the New York City watershed (Pederson et al 2013).

The sensitivity to water availability is further illustrated by iWUE data for the year 1999. Both species exhibited a higher iWUE reflecting lower Δ values (figure 1). The summer of 1999 was very dry with decreased soil moisture levels resulting in the restriction of the CO₂ supply to the leaf (Brugnoli et al 1988) because of stomatal closure, and in a weaker carbon uptake observed in the summertime NEE record (Urbanski et al 2007).