Response of Quercus velutina growth and water use efficiency to climate variability and nitrogen fertilization in a temperate deciduous forest in the northeastern USA

Abstract

Nitrogen (N) deposition and changing climate patterns in the northeastern USA can influence forest productivity through effects on plant nutrient relations and water use. This study evaluates the combined effects of N fertilization, climate and rising atmospheric CO₂ on tree growth and ecophysiology in a temperate deciduous forest. Tree ring widths and stable carbon (δ¹³C) and oxygen (δ¹⁸O) isotopes were used to assess tree growth (basal area increment, BAI) and intrinsic water use efficiency (iWUE) of Quercus velutina Lamb., the dominant tree species in a 20+ year N fertilization experiment at Harvard Forest (MA, USA). We found that fertilized trees exhibited a pronounced and sustained growth enhancement relative to control trees, with the low- and high-N treatments responding similarly. All treatments exhibited improved iWUE over the study period (1984–2011). Intrinsic water use efficiency trends in the control trees were primarily driven by changes in stomatal conductance, while a stimulation in photosynthesis, supported by an increase in foliar %N, contributed to enhancing iWUE in fertilized trees. All treatments were predominantly influenced by growing season vapor pressure deficit (VPD), with BAI responding most strongly to early season VPD and iWUE responding most strongly to late season VPD. Nitrogen fertilization increased Q. velutina sensitivity to July temperature and precipitation. Combined, these results suggest that ambient N deposition in N-limited northeastern US forests has enhanced tree growth over the past 30 years, while rising ambient CO₂ has improved iWUE, with N fertilization and CO₂ having synergistic effects on iWUE.

Introduction

The future structure, composition and productivity of the world’s forests are tightly linked to multiple global change drivers (IPCC 2013). Anthropogenic activities including intensive agriculture and the combustion of fossil fuels have increased reactive nitrogen (N) inputs to ecosystems to unprecedented levels, with some regions of the world projected to double their N deposition rates by 2050 (Galloway et al. 2008). Nitrogen deposition and rising atmospheric carbon dioxide (CO₂) are two important changes affecting forest health and productivity (Hyvönen et al. 2007), along with rising temperatures, longer growing seasons and changing precipitation regimes (Schimel et al. 2001, Allen et al. 2010). Evaluating tree- and ecosystem-level responses to these changes is crucial for projecting forests’ capacity to sequester carbon (C) in the coming decades and, in turn, to gauge how vegetation–atmosphere feedbacks will influence climate change (Bonan 2008).

Nitrogen deposition often enhances forest productivity, especially in N-limited temperate regions (Magnani et al. 2007, Thomas et al. 2010, Vadeboncoeur 2010); however, there is still a considerable variation between ecosystems that show enhanced productivity vs N saturation. Generally, N deposition improves C sequestration (Pinder et al. 2012), but elevated N can have negative biogeochemical consequences such as acidification and base cation depletion where N loads exceed biological demand (Aber et al. 1989, Wallace et al. 2007, Lovett and Goodale 2011). Estimates range widely on the amount of additional C stored per unit N, partially due to differences in species response (Thomas et al. 2010) as well as different C pools included in the estimate (i.e., aboveground, soils, ecosystems) (Nadelhoffer et al. 1999, Magnani et al. 2007, De Vries et al. 2009, 2014, Gentilesca et al. 2013, Ferretti et al. 2014, Fowler et al. 2015). The fate of sequestered C within ecosystems can also be difficult to trace, but recent work has shown that soils may be an equal or greater C sink than aboveground productivity in forest ecosystems (Lovett et al. 2013, Yanai et al. 2013, Frey et al. 2014). Further, while the tight link between N availability and photosynthesis is well established at the leaf and plant levels (Field and Mooney 1986), isolating this mechanism at different temporal and geographic scales from concomitant climate change drivers (e.g., temperature) remains difficult (Fleischer et al. 2013, Law 2013).

At the tree level, N availability strongly influences productivity and water use through increases in leaf area (Ewers et al. 2000) or increases in photosynthetic capacity through investment in photosynthetic machinery (Lambers et al. 2008), hence directly modulating intrinsic water use efficiency (iWUE), the ratio between photosynthesis (A) and stomatal conductance (g_s). Many studies of iWUE response to N deposition have involved short-term, soil application N fertilization experiments and have yielded a wide range of responses, ranging from no effect (Elhani et al. 2005, Balster et al. 2009) to an increase in iWUE due to either a stimulation in A (Ripullone et al. 2004, Walia et al. 2010, Brooks and Mitchell 2011) or a reduction in g_s (Brooks and Coulombe 2009). Other N deposition studies have observed transient reductions in g_s from long-term N fertilization (Betson et al. 2007), A stimulated from aerial mistings (Guerrieri et al. 2011), or either reductions in g_s or enhanced A from N emissions (Guerrieri et al. 2009, 2010). In contrast, few studies have assessed the long-term response of iWUE to elevated N deposition and how it interacts with other changing climate drivers, especially rising CO₂ and climatic variability.

The relatively direct relationship between N deposition and leaf gas exchange can be complicated by the interactive effects of other components of global changes, such as the increase in atmospheric CO₂ (Hyvönen et al. 2007). Plants may respond to rising CO₂ with adjustments in gas exchange at the leaf and canopy levels by changing A, g_s or both, which in turn directly affects tree productivity and water use (Ainsworth and Rogers 2007, Franks et al. 2013). Predictions that rising CO₂ would enhance tree productivity through a fertilization effect have limited support from field studies, as most have failed to document consistent growth responses to long-term increases in ambient CO₂ (Cole et al. 2010, McMahon et al. 2010, Silva and Anand 2013). Limitation by other resources such as water and nutrients (particularly N) may prevent plants from realizing the full potential fertilization effect of elevated CO₂ across ecosystems (Canadell et al. 2007, Norby et al. 2010, Reich and Hobbie 2013, Sigurdsson et al. 2013, Wieder et al. 2015). A more globally prevalent response of trees to rising atmospheric CO₂ has been an increase in iWUE (as derived from stable C isotopes) ranging from 1 to 60% over the last century depending on species and biome (Saurer et al. 2004, 2014, Andreu-Hayles et al. 2011, Peñuelas et al. 2011, Silva and Anand 2013, Xu et al. 2013, Lévesque et al. 2014, Frank et al. 2015, Van Der Sleen et al. 2015). However, disentangling the response of tree growth and iWUE to atmospheric CO₂ and other co-occurring global change drivers remains a central challenge in global change science (Franks et al. 2013).

Tree rings are valuable repositories of historical information on physiological processes due to their extensive spatial range (across large geographical areas) and fine temporal resolution (intra-annual to annual) that can be extended over millennia (McCarroll and Loader 2004, Babst et al. 2014). Dendrochronological techniques coupled with analysis of stable C and oxygen (O) isotope ratios (δ¹³C and δ¹⁸O) of tree rings are increasingly used to relate tree productivity (stem growth) with tree water use (iWUE), thereby explicitly linking tree C and water cycles. Carbon isotopes provide a time-integrated proxy for the ratio of CO₂ between the leaf intercellular spaces (c_i) and the atmosphere (c_a) (i.e., c_i/c_a), which can then be used to derive iWUE (Farquhar et al. 1982). When isotopic signatures of soil water and water vapor for the investigated trees are similar, δ¹⁸O in plant tissue can be used to evaluate changes in g_s (Scheidegger et al. 2000, Roden and Farquhar 2012), thereby making it possible to assess the relative contributions of changes in A or g_s to observed variations in c_i and iWUE. Global observations from dendro-isotopic studies have shown increases in iWUE in response to CO₂ primarily attributed to decreased g_s (Gagen et al. 2011, Nock et al. 2011, Van Der Sleen et al. 2015), with further reductions exhibited in ecosystems with trends of increasing moisture stress from changing temperature and precipitation regimes (Peñuelas et al. 2008, Andreu-Hayles et al. 2011, Saurer et al. 2014). Fewer dendro-isotopic studies have addressed the combined effects of multiple climate change drivers, especially with the inclusion of changes in N availability. A notable exception is a recent global analysis that assessed tree iWUE in relation to ambient N deposition, atmospheric CO₂ and climate metrics, which found evidence that N deposition contributed to the observed increases in iWUE from 1950 to 2000, but the effect was difficult to isolate from the other climatic and atmospheric parameters (Leonardi et al. 2012).

Here we present a study that addresses the response of tree growth and iWUE to N fertilization, rising atmospheric CO₂ and climate over three decades, including 23 years of continual N fertilization. Our objectives were to (i) assess how N fertilization influenced tree growth and iWUE, (ii) determine the primary mechanism by which iWUE is influenced by N (i.e., by altering either A, g_s or both) and (iii) evaluate the interactive effects of N fertilization, CO₂ and climate on tree growth and iWUE. To do this, we applied dendro-isotopic techniques in Quercus velutina Lamb. trees in the Chronic Nitrogen Amendment Study at Harvard Forest in Massachusetts, USA. We hypothesized that N would enhance tree growth and iWUE by stimulating A. Further, we expected the large amounts of N added to be the dominant driver influencing growth and iWUE, with the modest increases in atmospheric CO₂ concentration during the experiment exerting ancillary positive influences on iWUE. Finally, we predicted that climate would contribute most strongly to interannual variation in growth and iWUE, and that the N treatment would increase tree sensitivity to climate.

Materials and methods

Site description and sampling

This study was conducted at the Harvard Forest, a cool moist temperate transitional hardwood New England forest (42°30′N, 72°10′W). The mean annual temperature is 13 °C with 1120 mm of precipitation (Aber et al. 1993, Magill et al. 1997). Mean total inorganic wet N deposition (NH₄ + NO₃) decreased significantly from 5.7 kg N ha⁻¹ year⁻¹ before fertilization began to 3.8 kg N ha⁻¹ year⁻¹ at the end of observation period (1982–2012; P = 0.003; http://nadp.sws.uiuc.edu).

Within this forest, the Chronic Nitrogen Amendment Study has been carried out since 1988 to examine the impact of N deposition on the health and productivity of forest ecosystems (Aber et al. 1989, 1993). The study includes three 30 × 30 m plots in a single stand (one plot per treatment), which had been fertilized annually at 0, 50 or 150 kg N ha⁻¹ year⁻¹ (as NH₄NO₃) for the control, low-N and high-N treatments, respectively, since 1988 (for more details, see Aber et al. 1993, Magill et al. 1997). The soils are well-drained typic Dystrochrepts of the Canton or Montauk series with stony to sandy loam texture formed from glacial till. This stand is at least 60 years old and part of a section of forest that naturally regenerated from a salvage clear cut conducted after a hurricane in 1938 badly damaged 40–50% of the trees (Magill et al. 2004). Black oak is the dominant species within this stand (Q. velutina; 72% basal area, plots range from 67 to 85%), with smaller contributions from other hardwood species including black birch (Betula lenta L.; 10%), red maple (Acer rubrum L.; 9%) and American beech (Fagus grandifolia Ehrh.; 3%). The mean basal area is 30.3 m² ha⁻¹ (plots range from 28.0 to 33.7 m² ha⁻¹), with leaf area index estimates ranging from 4.4 to 5.3 for Harvard Forest (Zhao et al. 2011).

For this study, trees were visually inspected for health based on crown size and integrity, and presence of pests or disease. Dead and dying trees were avoided to ensure high core quality and dating accuracy. By selecting only healthy trees, we acknowledge that our sampling design is somewhat biased (Nehrbass-Ahles et al. 2014). Thus our results represent an optimal response to N fertilization, which is also appropriate given that our primary interest was to explore the optimal response and that tree decline or mortality were not main foci of the study. In October 2011, 10 healthy dominant Q. velutina trees were selected each from the control, low- and high-N plots. Three cores per tree were taken at breast height perpendicular to the slope with a 5.15 mm diameter increment borer. The mean and range diameter at breast height (DBH) of the selected trees were 25.2 (17.1–36.4), 25.6 (15.7–32.4) and 25.7 (14.6–32.8) cm in the control, low- and high-N plots, respectively.

Cross-dating, ring width measurements and basal area increment calculations

Cores were sanded, skeleton plotted and cross-dated following standard dendrochronological procedures (Fritts and Swetnam 1989). The rings were measured with a sliding scale micrometer (Velmex Measuring System, Velmex, Inc., Bloomfield, NY) using MeasureJ2X software (VoorTech Consulting, Holderness, NH, USA) and validated in COFECHA, a statistical cross-dating quality control program (Holmes 1983). Basal area increment (BAI) series were calculated from the total ring widths. For cores that passed but did not intersect the pith, its location was inferred from additional measurements, including graphical estimation based on ray convergence (Rozas 2003) and geometric estimation based on ring curvature (Duncan 1989). Basal area increment has been shown to be an appropriate index for relative growth that corrects for the negative trend in ring width occurring with tree age (LeBlanc 1990, Peñuelas et al. 2008). The usefulness of this approach increases if the juvenile growth phase or release events are avoided (LeBlanc 1990).

Stable isotope analysis

Annual rings from the three cores were cut into earlywood and latewood segments, and the latewood for each year was pooled by tree (one sample per tree per year). Latewood was selected for isotopic analysis as it is primarily derived from current year photosynthate (particularly for deciduous species), which allows for better isolation of the annual physiological signal within the tree that can then be related to annual treatment and climate conditions (McCarroll and Loader 2004). To minimize the total number of samples, while still encompassing the entire fertilization period at annual resolution, groups of years were selected for analysis: 1984–93 (10 years bracketing the onset of fertilization in 1988), 1998–2002 and 2008–11. Wood samples were shredded and then extracted for holocellulose (for δ¹³C) and α-cellulose (for δ¹⁸O), according to the procedure described by Leavitt and Danzer (1993) and Sternberg (1989). For δ¹³C analysis, 1–3 mg of holocellulose of each sample was weighed in tin capsules and combusted on an elemental analyzer (Costech 4010 Elemental Analyzer, Costech Analytical Technologies, Valencia, CA, USA) coupled to an isotope ratio mass spectrometer (Finnigan Delta Plus XP Mass Spectrometer, Thermo Fisher Scientific, Germany). For δ¹⁸O, 0.15–0.25 mg of α-cellulose of each sample was weighed in silver capsules and analyzed on an elementar PyroCube (Elementar Analysensysteme GmbH, Hanau, Germany) interfaced to a PDZ Europa 20-20 isotope ratio mass spectrometer (Sercon Ltd, Cheshire, UK). Two internal blind references run with samples had standard deviations of ≤0.1 and ≤0.4‰ between batches for δ¹³C and δ¹⁸O, respectively. Relative stable isotope abundances are expressed as ratios of ¹³C/¹²C and ¹⁸O/¹⁶O using δ-notation (in per-mil; ‰) relative to an established standard (VPDB for δ¹³C and VSMOW for δ¹⁸O; Coplen 2011).

Stable isotope theory

The δ¹³C composition of plant material (δ¹³C_p) is generally depleted relative to the δ¹³C of atmospheric CO₂, due to the biophysical and biochemical processes that discriminate against ¹³C during CO₂ uptake and fixation occurring in the leaf. According to the simplified version of the full equation proposed by Farquhar et al. (1982), δ¹³C_p is directly linked to the c_i/c_a ratio and is affected by isotopic fractionations that occur during the diffusion of CO₂ through stomata (a; 4.4‰) and carboxylation (b; ∼27‰):

$${\delta }^{13}{\mathrm{C}}_{\mathrm{p}}={\delta }^{13}{\mathrm{C}}_{\mathrm{a}}-a-(b-a)\frac{c_{\mathrm{i}}}{c_{\mathrm{a}}}$$

(1)

Leaf c_i (and δ¹³C_p) therefore reflects the balance between the diffusion of atmospheric CO₂ into the leaf (source) and CO₂ fixation through photosynthesis (sink). While this model does not include important effects of c_c (CO₂ concentration at the chloroplast) and leaf internal conductances such as mesophyll conductance and dark respiration (Farquhar et al. 1982, 1989, Seibt et al. 2008), the simplified model is appropriate for relative estimates of c_i and iWUE (Cernusak et al. 2013).

Intrinsic water use efficiency is the direct trade-off between net photosynthesis (A) and water loss due to stomata conductance to water (g_s):

$$\mathrm{iWUE}=\frac{A}{g_{\mathrm{s}}}=\frac{c_a-c_i}{1.6}$$

(2)

where $A=g_{\mathrm{C}{\mathrm{O}}_2}$ (c_a − c_i) and g_s is linked to stomatal conductance to CO₂, i.e., $g_{\mathrm{C}{\mathrm{O}}_2}$ by a constant, i.e., $g_{\mathrm{s}}=1.6\hspace{0.17em}(g_{\mathrm{C}{\mathrm{O}}_2})$ (Farquhar et al. 1989). Atmospheric CO₂ and δ¹³C were obtained from McCarroll and Loader (2004) and more recent observations (2004–11) from Mauna Loa (http://www.esrl.noaa.gov/gmd/ccgg/trends/). Thus, the δ¹³C_p from tree rings can trace the variations in iWUE through time with respect to rising atmospheric CO₂ and other climatic or physiological changes.

Whereas the variation in δ¹³C reflects a change in A, g_s or both, the δ¹⁸O of plant material is related to changes in leaf water δ¹⁸O influenced by relative humidity, isotopic composition of water vapor, evaporative and diffusional effects during transpiration, as well as isotopic exchanges between tree water and organic molecules (Barbour 2007). During transpiration within the leaf, the lighter H₂¹⁶O molecules evaporate faster than the heavier H₂¹⁸O molecules, leading to the enrichment of ¹⁸O/¹⁶O water within the leaf. The isotopic composition at the site of evaporation (δ¹⁸O_e) is expressed as the difference between the intercellular (e_i) and atmospheric (e_a) vapor pressure, as described by Craig and Gordon (1965):

$${\delta }^{18}{\mathrm{O}}_{\mathrm{e}}={\delta }^{18}{\mathrm{O}}_{\mathrm{s}}+ϵ\ast +{ϵ}_{\mathrm{k}}+({\delta }^{18}{\mathrm{O}}_{\mathrm{v}}-{\delta }^{18}{\mathrm{O}}_{\mathrm{s}}+{ϵ}_{\mathrm{k}})\frac{e_{\mathrm{a}}}{e_{\mathrm{i}}}$$

(3)

where δ¹⁸O_v and δ¹⁸O_s represent the isotope composition of water vapor and source water, respectively, ϵ* is the equilibrium fractionation factor and ϵ_k is the kinetic fractionation as water diffuses through the stomata and boundary layer. The δ¹⁸O of leaf water is further influenced by the Péclet effect at high rates of transpiration (Farquhar and Lloyd 1993, Song et al. 2013), the result of which is a reduction of δ¹⁸O in the leaf water as a whole. The signal of δ¹⁸O in leaf water is then imprinted in the leaf organic molecules, and due to the isotopic exchange between carbonyl oxygen and water in which they are formed, is further modified until it is imprinted as cellulose (Sternberg et al. 1986). When the isotopic signature of soil water and water vapor for the investigated trees are similar, the δ¹⁸O of plant tissue frequently shows a negative correlation to g_s (Barbour and Farquhar 2000, Siegwolf et al. 2001, Barbour 2007, Grams et al. 2007, Ripullone et al. 2009). Therefore, δ¹⁸O can be used to constrain the iWUE ratio and to interpret variations as due to changes in A, g_s or both.

Scheidegger et al. (2000) developed a conceptual model that analyzes the directional relationship between δ¹³C and δ¹⁸O in leaf organic matter. The application of the dual-isotope approach in tree rings violates some assumptions of this model, particularly that source water and water vapor δ¹⁸O, as well as ambient humidity, should be constant through space and time (Scheidegger et al. 2000, Roden and Siegwolf 2012). However, the synchronous analysis of δ¹³C and δ¹⁸O can still be insightful in retrospective tree rings studies (Brooks and Mitchell 2011, Barnard et al. 2012, Roden and Farquhar 2012, Battipaglia et al. 2013, Lévesque et al. 2014), especially if efforts are taken to constrain model assumptions. In this study, we minimize spatial heterogeneity by analyzing a single tree species in adjacent plots within a relatively homogenous closed-canopy stand (similar age, species composition and basal area), thereby minimizing differences in source water, water vapor and interspecies fractionation processes, with particular reference to the Péclet effect (see Song et al. 2014). Data of δ¹⁸O in soil water at the adjacent AmeriFlux site at Harvard Forest (T.E. Dawson and J.W. Munger, personal communication) showed similar values for the 2005 (−7.08‰) and 2006 (−6.02‰) growing seasons (April–October; t = −0.66, P = 0.53). Similarly, leaf water δ¹⁸O did not differ between the two years (t = −1.95, P = 0.37). To analyze the variation through the investigated time, we modeled δ¹⁸O in soil water, based on the measured δ¹⁸O in soil water during the 2005 and 2006 growing seasons. Modeled growing season (May–Septmeber) soil water δ¹⁸O had no trend over the course of the study period (slope = 0.0061, F = 0.073, P = 0.789; Figure S1 available as Supplementary Data at Tree Physiology Online). While interannual variation was present, the effects would be reduced as we assess δ¹⁸O in groups of years (i.e., time periods). Moreover, the use of the dual-isotope conceptual model is limited to evaluating directional and relative changes in A and g_s due to N fertilization and climate, and is not intended to indicate absolute changes.

Climate data

In order to be incorporated into the linear mixed model approach, we used to analyze climate effects, climate records had to cover the length of the entire BAI and iWUE analysis period (1984–2011) and be highly complete. Mean monthly precipitation and temperature data were obtained from the Harvard Forest Shaler and Fisher meteorological stations, but the closest station with hourly temperature and dew point records allowing the calculation of vapor pressure deficit (VPD) over the full study period was Worcester Regional Airport (Worcester, MA, USA; ∼40 km from Harvard Forest; National Climate Data Center). An afternoon VPD (the mean over the period from 12:00 to 17:00 h) series was calculated to characterize the part of the daily VPD cycle with the largest influence on tree physiology, particularly stomatal conductance. Exploratory multiple regression using mean VPD (24-h) vs afternoon VPD showed that afternoon VPD explained a larger proportion of the variance within BAI and iWUE data (data not shown), therefore afternoon VPD was selected for subsequent climate analyses. Regional wet inorganic N deposition data were obtained from the Quabbin Reservoir National Atmospheric Deposition Program site (site MA08; ∼20 km from Harvard Forest; http://nadp.sws.uiuc.edu) (National Atmospheric Deposition Program (NRSP-3) 2007). Annual ambient N deposition was added to the amount of N fertilization so that the total annual N (N_ann) deposition rate and total cumulative N (N_cum; 1984–2011) deposition were calculated for each treatment.

Data analysis

Cellulose δ¹³C correction

Raw δ¹³C values from tree ring cellulose were corrected for the changing isotope ratio of atmospheric CO₂ due to the burning of ¹³C-depleted fossil fuels (the Suess effect; Keeling 1979). The correction factor for each year (McCarroll and Loader 2004) mathematically removes the isotopic depletion relative to the pre-industrial atmosphere (ca 1850). Throughout the rest of the paper, δ¹³C values represent the tree-ring δ¹³C after correction.

Temporal trends in climate data

Temporal trends in temperature, precipitation, VPD and ambient N deposition were evaluated with Kendall non-parametric robust line fit with Sen’s method and the degree of significance was determined with the Mann–Kendall trend test in R version 3.0.2 using packages ZYP (Bronaugh and Werner 2013) and Kendall (McLeod 2011).

Temporal and indirect treatment effects

Because isotope analysis was conducted on groups of years throughout the treatment, δ¹³C, δ¹⁸O and iWUE analysis were compared between time periods; pre-treatment years 1984–87 (T0), and three time periods during treatment: 1989–93 (T1), 1998–2002 (T2) and 2008–11 (T3). Temporal and treatment differences for δ¹³C, δ¹⁸O, iWUE and BAI were tested for homoscedasticity using Levene’s test and were then evaluated with ANOVA and Tukey–Kramer HSD when variances were equal or evaluated with the non-parametric Welch test when variances were unequal among treatments or time periods. Analyses were conducted in JMP statistical software (Version 11, SAS Institute Inc., Cary, NC, USA).

Direct treatment effects—response ratios

Response ratios (RRs) were used to measure the treatment response of the trees in the N fertilized plot in comparison with the control plot. Response ratios are commonly used in ecology to quantify the proportional change from experimental manipulations (Hedges et al. 1999, Battipaglia et al. 2013). We calculated RRs as:

$$\ln (RR)=\ln \frac{X_{\mathrm{E}}}{X_{\mathrm{C}}}$$

(4)

where X_E represents the BAI or iWUE in the high or low fertilized plot and X_C is the response in the control plot for a given year (Hedges et al. 1999). Student’s t-test was used to test whether the mean RRs (using trees as replicates) were significant on an annual basis. Analyses were conducted in JMP statistical software (JMP Version 11, SAS Institute Inc.).

Linear mixed models—BAI and iWUE temporal trends and N fertilization–climate relations

Linear mixed-effect models were constructed to test for temporal trends, pair-wise relationships, as well as climatic and N fertilization effects on growth and water use efficiency for the Q. velutina study trees. Mixed models were selected for this analysis because they are flexible functions that can be customized for data with complex nesting, heterogeneity, autocorrelation and random effects. Simple mixed models were constructed to test for temporal trends over the observation period for BAI and iWUE within treatments, with year as the predictor variable and tree as the random effect. Similarly, the relationship between BAI and iWUE was investigated with mixed models, with iWUE as the predictor and tree as a random effect.

More complex mixed models were constructed to investigate the relationship between the N fertilization treatment and climate and their combined influence on growth and water use efficiency. Nitrogen deposition (N_ann or N_cum; treatment = ambient deposition + fertilization, and control = ambient deposition), CO₂ and monthly growing season climate predictors (May–September precipitation, temperature and afternoon VPD) were fitted to BAI and iWUE time series from 1984 to 2011 (28 and 19 years, respectively). Additionally, BAI and iWUE were further analyzed using three different models to better isolate the treatment vs climate effects: (i) a full model including all observations, (ii) treatment trees (low- and high-N) only and (iii) control trees only, for a total of six models.

Following the guidelines of Zurr et al. (2009) we began by fitting a generalized least squares function to a saturated model, including all predictor variables, to act as a baseline model. From the baseline model, successive modifications were applied to test for heterogeneity, autocorrelation and random effects. The optimal saturated models were selected with the lowest AIC (Akaike Information Criterion) and BIC (Bayesian Information Criterion), and the best set of validation graphs that checked for overall fit, heterogeneity, autocorrelation and undesirable patterns within the residuals (Zurr et al. 2009). The optimal saturated models were then simplified in subsequent iterations by removing insignificant terms as evaluated with the likelihood ratio test. Generally, fixed terms were included when P ≤ 0.05; however, for the smaller subsets that contained few observations (i.e., the control models), a higher cut off of P ≤ 0.10 was used. Fixed terms were centered to reduce collinearity between terms and the intercept. Nevertheless, a high correlation remained between N_cum and CO₂ within the model subsets (i.e., treatment and control models; VIF > 5), as both factors increased approximately linearly with time, a trend unrelated to the fertilization treatment. In these cases, the CO₂ term was removed since the focus of the analysis was on the influence of the N fertilization treatment and its interaction with the increase of atmospheric CO₂, rather than isolating the effect of the CO₂ alone. The mixed models were fit and analyzed in R, version 3.0.2 (R Development Core Team 2014), using the package NLME (Pinheiro et al. 2013).

Results

Trends in δ¹³C, δ¹⁸O, iWUE and BAI

Generally, trees that received N fertilization were less depleted in ¹³C and had higher iWUE and BAI compared with control trees, although the response varied by time period (T1–T3) and N fertilization rate (low- vs high-N; Figure 1a, c and d). Growth over the fertilization period (1988–2011) resulted in a mean cumulative basal area of 137 cm² (±15; 1 SE) in the control trees vs 251(±31) and 266(±31) cm² in the low- and high-N trees, respectively. All trees, including those in the control plot, exhibited long-term increases in iWUE over the nearly 30 years of observation. Specifically, iWUE increased by 9.6% (±2.5; 1 SE), 15.5% (±2.9) and 16% (±8.4), whereas BAI increased 62% (±8.4), 65% (±10.6) and 95% (±10.8) from pre-treatment (T0) to the end of the study period (T3) in the control, low- and high-N plots, respectively. Intrinsic water use efficiency was a significant predictor of BAI in all treatments, with the low- and high-N treatments showing similar strength relationships (mixed model P < 0.0001), and the control trees showing a slightly weaker relationship (P = 0.0025). Generally, trees with higher iWUE had higher growth rates, but this relationship was variable with some trees exhibiting a relatively strong BAI–iWUE relationship, while in other trees the relationship was absent (see Figure S2 available as Supplementary Data at Tree Physiology Online). Prior to fertilization, physiological differences existed among the trees in the three plots (T0, Figure 1). Most notably, cellulose of trees growing in the low- and high-N plots was significantly more enriched in ¹⁸O than trees growing in the control plot (P < 0.001), a difference that was largely maintained throughout the treatment period (Figure 1b).

Figure 1.

δ¹³C (corrected for Suess effect) (a), δ¹⁸O (b), iWUE (c) and BAI (d) trends through time. Each point represents the treatment mean of that time period with trees n = 8–10. T0 = pre-treatment (1984–87), T1 (1989–93), T2 (1998–2002), T3 (2008–11). Error bars represent ±1 SE.

To better evaluate the treatment effects, RRs were used to directly compare the proportional change in growth and water use efficiency between the N-fertilized (low- and high-N) and control trees (Figure 2). Both the low- and high-N fertilization treatments showed sustained significant enhancements in BAI over the control for nearly 20 years (1988–2005, low-N and 1988–2008, high-N; Figure 2a). No significant differences were detected between the mean responses of the two N-fertilization rates over the treatment period. The treatment effect on growth dropped below detection at the end of the observation period due to a combination of increased variability among individual tree responses and a large growth increase in the control trees (T3, Figure 1d). Prior to the onset of fertilization, differences were present among the trees, with the low-N trees exhibiting significantly higher growth rates than the control trees beginning in the early 1980s. Even accounting for these pre-treatment differences, the low-N trees showed significant growth enhancements compared with their pre-treatment growth rates (Figure 1d) and the growth rates of the control trees (Figure 2a) once fertilization began.

Figure 2.

Natural log RRs comparing mean annual BAI (a) and iWUE (b) in the low- and high-N treatment trees with the control trees. Vertical black lines represent the start of fertilization in 1988. Stars represent significance at P ≤ 0.05. In (a), both the low- and high-N treatments are significant during 1985, 1988–2005 and 2008 (stars shown), while the low-N is also significant during 1982–87, and high-N during 2006–07 and 2011 (not starred). Error bars represent ±1 SE.

The trends for iWUE were more variable (Figure 2b), with significant increases observed for the fertilized trees over the control trees primarily during the early and middle periods of the fertilization treatment (1989–93 and 1998–2002). When the entire treatment period was assessed as a whole, trees growing in both fertilized treatments had significantly higher iWUE than control trees (P < 0.001), with the high-N trees showing the largest stimulation (high-N > low-N; P < 0.001).

Changes in iWUE and its determinants by combining δ¹³C vs δ¹⁸O

In the treatment trees, δ¹⁸O and δ¹³C were analyzed as the difference from the control trees in order to remove climatic influences and difference between treatments for δ¹⁸O (see later in the text), and grouped into pre-fertilized and fertilized time periods to summarize the overall treatment effect (Figure 3). Both the low- and high-N plots had significant increases in δ¹³C with no detectable change in δ¹⁸O. In the control trees, the only significant change between time periods was a significant reduction in δ¹⁸O from T2 to T3 (Figure 4). All other changes in isotope composition were non-significant, but showed a consistent trend, whereby δ¹⁸O and δ¹³C increased or decreased together between time periods.

Figure 3.

Treatment response (a) of δ¹⁸O and δ¹³C as differences from the control (treatment–control) comparing pre-treatment (T0, 1984–87) with treatment mean (T1–T3, 1989–93, 1998–2002, 2008–11). Error bars represent 95% confidence intervals. Model interpretation (b) using the Scheidegger et al. (2000) conceptual model where changes in iWUE can by analyzed using the directional relationship between δ¹⁸O and δ¹³C. Variations in δ¹³C infer changes in c_i (and A) while δ¹⁸O is used to infer changes in g_s.

Figure 4.

Mean δ¹⁸O and δ¹³C (a) for the control trees during the four time periods (T0 = pre-treatment (1984–87), T1–T3 = fertilization period (1989–93, 1998–2002, 2008–11)). Error bars represent 95% confidence intervals. The short arrow represents transition from T0 to T1 and the long arrow is the transition from T2 to T3. Model interpretation (b) using the Scheidegger et al. (2000) conceptual model where changes in iWUE can by analyzed using the directional relationship between δ¹⁸O and δ¹³C. Variations in δ¹³C infer changes in c_i (and A) while δ¹⁸O is used to infer changes in g_s.

Assessing the effect of N fertilization, CO₂ and climate on changes in iWUE and BAI

The interacting effects of N fertilization, CO₂ and climate on tree growth and water use were examined with linear mixed models. All models (six in total; three for BAI and three for iWUE) included a random intercept for each tree and random slope for N_ann or N_cum, accounting for differences in baseline growth and water use, as well as unique responses by individual trees to N fertilization. To limit the number of fixed terms added to the models, monthly climate variables from the growing season (May–September) were selected for this analysis. Before simplification, the saturated models included N_ann or N_cum, annual CO₂, N_ann/N_cum × CO₂ interaction and growing season (5 months) temperature, precipitation and maximum VPD, for a total of 18 fixed predictor variables (see Materials and methods for more details). Early in the model selection process N_cum was consistently a stronger predictor than N_ann, and was therefore chosen as the N fertilization term.

The final models fit the data well, with the predicted vs the observed values clustering around the 1 : 1 line (Figure 5). All models showed significant influences of N_cum, the N_cum × CO₂ interaction (with the exception of one control subset) and two to five monthly climate predictors (Table 1). Cumulative N (ambient deposition + fertilization treatment) and atmospheric CO₂ (full models only) had positive linear effects on both BAI and iWUE. Yet, a significant negative interactive effect of N_cum and CO₂ was detected in most models.

Table 1.

Mixed effect models of BAI and iWUE as functions of N-fertilization (N_cum = ambient + treatment cumulative N dosage to date), atmospheric CO₂, N_cum × CO₂ interaction, and growing season climate (5–9 = May–September). BAI and iWUE were divided into three subset models: (1) full model = all observations, (2) control = control trees and (3) treatment = treatment trees. Table lists coefficients of predictor variables and stars represent level of significance (*P < 0.05; **P < 0.01; ***P < 0.001). CO₂ is not added to the control and treatment model subsets due to high multicollinearity. N_cum (kg N ha^–1 year^–1); CO₂ (ppm); Prec (mm day^–1) Temp (°C); VPD (kPa).

	BAI			iWUE
	Full (n = 743)	Control (n = 224)	Treatment (n = 519)	Full (n = 492)	Control (n = 139)	Treatment (n = 353)
Ncum	0.001*	0.023***	0.002***	0.004***	0.100***	0.009***
CO2	0.061***			0.282***
Ncum × CO2	−5.4 × 10−5***	−8.4 × 10−4***	−6.1 × 10−5***	−9.0 × 10−5***		−7.0 × 10−5**
Prec5				0.305**
Prec7	0.125**		0.162*	−0.389***		−0.618**
Prec9		−0.175**
Temp7	0.297***		0.360***	−1.043***		−0.826**
VPD5		0.186***		−1.176***	1.303***	1.065***
VPD6				0.851***	0.919***	0.808***
VPD7	−0.158**		−0.162*
VPD8	0.237***	0.214**	0.265***
VPD9		−0.258**

Figure 5.

Observed versus modeled values from mixed effects models for BAI and iWUE. Diagonal line represents the 1 : 1 line. (Model goodness of fit: BAI RMSE 2.057, iWUE RMSE 4.615.)

The relationships of BAI and iWUE with growing season climate metrics highlighted some interesting patterns for growth and water use trade-offs, as well as differences between control and fertilized trees. Congruity across models showed physiological sensitivities to July growing conditions and early or late season VPD depending on the modeled parameter. Within the full and treatment models significant July precipitation and temperature predictors were detected in both BAI and iWUE; however, they had opposing trends. Indeed, July precipitation and temperature were positively correlated with BAI (improving growth) while they were negatively correlated with iWUE. Over the modeled time period (1984–2011), July temperature had a non-significant increasing trend while July precipitation showed no trend. Additionally, all models had a significant effect of VPD on changes in BAI and iWUE. Consistent sensitivities to early growing season VPD (May and June) were detected in iWUE model sets, while BAI showed responsiveness to mid to late season VPD (July–September).

Discussion

Effect of N fertilization on tree growth and iWUE

Twenty-three years of N fertilization elicited strong physiological responses in these Q. velutina trees at Harvard Forest by significantly improving tree growth and iWUE, which were largely maintained over the observation period. These results likely represent the maximum stimulus after 23 years of N fertilization since the selected sample trees were the healthiest dominant Q. velutina trees on the plots. Despite the large difference in N fertilization rates (three times greater for high-N plot than low-N plot), the magnitude of the tree response (expressed as RRs) was similar in the two treatments, with the exception of a slightly larger enhancement in iWUE in the high-N plot. These findings corroborate previous work conducted at the Chronic Nitrogen Amendment plots, which also documented that N fertilization increased stem wood growth (Magill et al. 2004; Q. velutina represents 72% of basal area), and that the growth response was not substantially different between the two fertilization rates (Frey et al. 2014). In the northeastern USA (Elvir et al. 2003, Wallace et al. 2007, Thomas et al. 2010) and elsewhere (Brooks and Coulombe 2009, Gentilesca et al. 2013), N additions often stimulate tree growth, especially in regions categorized as primarily N-limited (Lebauer and Treseder 2008, Vadeboncoeur 2010), although many examples of neutral (Nadelhoffer et al. 1999, Elhani et al. 2005, Seftigen et al. 2013, Du and Fang 2014) or negative (Magill et al. 2004, Högberg et al. 2006, Wallace et al. 2007) responses to N fertilization have been reported. Ultimately, the response of trees to N additions is related to a multitude of factors such as stand age, species composition, ambient N deposition, duration and dose of applied N, as well as sensitivities of different tree species and soils (Pardo et al. 2011).

Nitrogen fertilization frequently improves iWUE by stimulating A, as plant N supply and photosynthesis are tightly coupled (Siegwolf et al. 2001, Ripullone et al. 2004, Brooks and Coulombe 2009, Guerrieri et al. 2010, 2011, Brooks and Mitchell 2011). In some cases however, detecting this relationship can be difficult. At the leaf level, some studies of instantaneous gas exchange measurements conducted on trees subjected to N additions have detected increased rates of photosynthesis (Mitchell and Hinckley 1993, Ripullone et al. 2004, Elvir et al. 2006), while others did not (Bauer et al. 2004, Talhelm et al. 2011). The responses were variable between species, and tree health was related to the amount of N added. Overall, our results are consistent with the well-established N–A relationship, as foliar % N (Frey and Ollinger 2012; http://harvardforest.fas.harvard.edu:8080/exist/xquery/data.xq?id=hf008) was a strong predictor for iWUE (⁠$R_{\mathrm{adj}}^2=0.582,$P < 0.0001) and to a lesser extent BAI (⁠$R_{\mathrm{adj}}^2=0.362,$P < 0.001; data not shown). Further, the greater foliar N content in the treatment trees (see Figure S3 available as Supplementary Data at Tree Physiology Online) is consistent with the positive relationship observed between BAI and iWUE as well as the higher productivity as indicated by their greater BAI during the study period. These findings are also consistent with the isotopic trends, whereby δ¹³C increased and δ¹⁸O remained stationary, indicating that changes in A largely explain changes in iWUE.

Recently, the limitations of the dual-isotope approach have been debated, and several authors have highlighted the challenges involved when interpreting variations in tree-ring δ¹³C and δ¹⁸O in terms of changes in A and g_s, with particular reference to δ¹⁸O (Roden and Farquhar 2012, Roden and Siegwolf 2012, Barbour and Song 2014, Gessler et al. 2014). While caution is warranted in interpreting these isotopic proxies, application of the dual-isotope approach can be insightful in simple systems. Here we assume that trees are exposed to similar atmospheric and soil water conditions, as the treatment and control plots are adjacent within the same stand (see Materials and methods for more details). However, pre-treatment differences in δ¹⁸O between the control and the treatment trees did suggest differences in source waters (e.g., different depths of soil water uptake or differing hydrologic flow paths in the soil, most likely due to variations in microsite topography). This should not be a confounding factor in our analysis, as we did not directly compare stable isotopes between treatments, but rather changes over time within plots. Furthermore, to analyze the effect of the N fertilization, we standardized for any baseline differences between treatment and control trees by considering the RR (e.g., difference between control vs treatment trees) rather than absolute values.

After several years of fertilization, significant trends in growth and iWUE were more difficult to detect because response variability increased among the N-fertilized trees (see error bars in Figure 2). This is particularly pronounced in the BAI for both treatments (2006–11), as well as iWUE in the high-N trees (2008–11). While some trees continued to show an increasing trend in growth or iWUE, other individuals stabilized or even declined in the final years of observation (see Figures S4 and S5 available as Supplementary Data at Tree Physiology Online). This diverging response among trees after prolonged N fertilization, particularly at high rates as seen in the high-N plot, could be a physiological threshold to N fertilization where benefits wane and neutral or even negative responses are produced over time (Aber et al. 1998). Another contributing factor could be increased competition between trees due to the N fertilization, where dominant trees have a greater capacity to utilize the additional N and benefit more than non-dominant trees.

Atmospheric CO₂ and climate influence on Q. velutina at Harvard Forest

By analyzing the trends in the control trees, we examined the effects of climate, rising atmospheric CO₂ and ambient N deposition on BAI and iWUE. In addition, we were interested to assess whether our results are in line with previous regional studies that explored changes in WUE at the tree and ecosystem level in response to global change drivers. Following a very similar approach to our study, Belmecheri et al. (2014) analyzed tree rings in Quercus rubra within the nearby Harvard Forest AmeriFlux tower footprint and found C isotope discrimination Δ¹³C, as derived from δ¹³C in tree ring cellulose, to be influenced by April and May precipitation, as well as May Palmer Drought Severity Index. Despite the different lengths of the study period (18 vs 28 years), we observed a similar early summer moisture signal driving variation in iWUE within the control trees, with significant influences from May and June VPD.

Harvard Forest has the longest record of eddy covariance measurements worldwide (Urbanski et al. 2007), and previous work analyzing these data has documented increases in net ecosystem C uptake, reduced transpiration as well as increased photosynthesis and water use efficiency from 1992 to 2010 (Keenan et al. 2013), attributed mostly to a CO₂ fertilization effect. A significant increase in aboveground woody biomass within the tower footprint was also reported (doubling over the 18 year time period (Keenan et al. 2012), primarily attributed to Quercus species (Urbanski et al. 2007; 52% basal area, Belmecheri et al. 2014), which have exhibited a trend of increasing BAI calculated from tree rings (Belmecheri et al. 2014). The control trees in our study site, which is located ∼1 km from the tower and outside the core footprint, follow similar trends, with significant increases in BAI (mixed model P < 0.0001; 1964–2011) and iWUE (mixed model P < 0.0001; 1984–2011). They also showed a stable c_i/c_a trend over the observation period (see Figure S6 available as Supplementary Data at Tree Physiology Online), indicating that trees actively adjusted A and g_s in response to increasing atmospheric CO₂, thus improving iWUE (Saurer et al. 2004). Furthermore, evidence from the mixed models suggests that ambient N deposition positively contributed to these increasing trends in growth and iWUE.

A fundamental predicted response to rising CO₂ is that plants will experience increased rates of A and growth accompanied by a decrease in g_s through a CO₂ ‘fertilization effect’ (Korner 2000, Norby and Zak 2011, Franks et al. 2013). However, many recent studies have documented a decoupling between iWUE and growth, whereby long-term increases in iWUE were not accompanied by increased growth (Silva et al. 2010, Andreu-Hayles et al. 2011, Peñuelas et al. 2011, Silva and Anand 2013, Lévesque et al. 2014, Van Der Sleen et al. 2015), suggesting that the CO₂ fertilization effect is not being realized in many ecosystems. In contrast to these previous studies, we observed sustained increases in growth in the control trees over the 30-year measurement period. However, the cause of this growth trend may be at least partially attributed to stand age rather than CO₂ fertilization. The stand reportedly established after the 1938 hurricane (Magill et al. 2004), and the mean chronological age of these Q. velutina trees in 2011 was at least 60 years at breast height. Typical successional patterns of Quercus in New England mixed hardwood forest show oaks maintaining constant growth rates with increasing BAI (Oliver and Stephens 1977, Oliver 1978) for up to 100 years after stand initiation (Hibbs 1983), indicating that these trees have not fully matured and may continue to grow steadily for several more decades.

N fertilization, climate and [CO₂] interactions

The N-amendment experiment at Harvard Forest offered a unique opportunity to explore the interactive effects of high rates of N fertilization, climate and ambient CO₂ on tree growth and water use over nearly three decades of dynamic environmental change. When N fertilization is added to the effects of climate and atmospheric CO₂, we gained insight as to how all three factors interact to influence Quercus tree growth and physiology. Analysis of the mixed models revealed that BAI and iWUE in all trees were positively influenced by N_cum (i.e., ambient deposition for control trees and ambient + fertilization for treatment trees) and CO₂. While some caution is warranted in interpreting the individual influence of N and CO₂ within the mixed models due to the collinearity between the predictors, we can still identify separate influences by comparing the trends between the treatment and control trees. Still, the significant negative interaction between N_cum and CO₂, albeit with very small coefficients, suggests that when combined, these two stimuli have counteracting effects on growth and water use. A possible explanation for the observed trends for iWUE may be that N and CO₂ influence different mechanisms, whereby N stimulates A (Chapin et al. 1987, Evans 1989, Ripullone et al. 2004) and CO₂ more strongly influences g_s (Ainsworth and Rogers 2007, Lammertsma et al. 2011). This interpretation is consistent with our observation of strong increases in A in the treatment trees (Figure 3), while control trees were more influenced by g_s (Figure 4).

In the Duke Forest FACE experiment, C sequestration in loblolly pine was maximized when elevated CO₂ was coupled with increased availability of N and water (Oren et al. 2001). Contrastingly, a study conducted in a mature mixed-deciduous forest in Switzerland did not find a lasting growth stimulation effect in response to elevated CO₂, even though this forest was not considered N-limited due to the high deposition load for the region (Körner et al. 2005). Other FACE studies have also documented transient or neutral responses to CO₂ depending on tree species and ecosystem type (Körner 2006, Norby and Zak 2011).

The mixed model analysis also provided evidence that N fertilization altered tree responsiveness to climate by increasing tree sensitivity to July temperature and precipitation. Significant model terms for July were detected in the full and treatment models but not within the control model subsets, suggesting that once N limitation is relieved in this stand, climate becomes a more dominant control on tree growth and iWUE. Further, this climate signal detected in July for fertilized trees showed a growth–water use trade-off whereby greater precipitation and warmer temperatures stimulated growth at the cost of decreasing iWUE. This is most likely caused by increasing g_s to stimulate A and therefore growth, a physiological response that is aided by high water availability.

A common trend observed for both the control and treatment trees was a fairly consistent sensitivity to growing season VPD. All models had significant predictor terms for early season VPD (May and June) for iWUE and late season VPD (July–September) for BAI, indicating that VPD is a strong climate control influencing interannual variations in growth and water use for Q. velutina, likely affecting g_s (Mencuccini and Binks 2015, Roman et al. 2015). Vapor pressure deficit is a well-documented driver of plant gas exchange by inducing partial stomatal closure (and reducing g_s) when VPDs between the leaf and the atmosphere increase (Lange et al. 1971, Oren et al. 1999). This is similar to findings by Voelker et al. (2014) who found VPD to be a better predictor of growth and Δ¹³C than precipitation across a broad range of riparian bur oak in the northern Mississippi basin. Voelker et al. (2014) hypothesized that VPD would be a strong regulator of leaf level gas exchange driving the variation in Δ¹³C and growth in the absence of soil moisture deficit. Similarly, Harvard Forest is not typically considered water-limited and therefore water availability is not usually implicated as constraining tree growth or iWUE. However, recent work on tree physiology has detected precipitation signals in many forest ecosystems worldwide including humid temperate forests (Allen et al. 2010, Choat et al. 2012, Pederson et al. 2012). When soil moisture does not limit tree growth and iWUE, atmospheric evaporative demand could become the dominate climate driver of tree growth and iWUE, as seen here with the influence of VPD. While Quercus species are generally considered well adapted to drought through avoidance or tolerance, Q. rubra (and closely related Q. velutina) are the least resistant species of this genus in North America (Abrams 1990). Future climate projections for the Northeast predict increases in total annual rainfall, yet more frequent growing season moisture stress due to higher temperature reduced spring soil recharge, and prolonged dry periods (Rustad et al. 2012, Swain and Hayhoe 2015) that could negatively impact Q. velutina growth.

Seasonal VPD signals influencing iWUE and BAI can also partially be explained by the timing of xylogenesis. Quercus species form the majority of their earlywood vessels before budburst to recover from winter embolism and restore hydraulic conductivity (Cochard and Tyree 1990); therefore, latewood formation can begin as early as late May or early June (Zasada and Zahner 1969, Michelot et al. 2012a, 2012b, Voelker et al. 2012, Pérez-de-Lis et al. 2015). Significant predictor terms for May and June VPD within the iWUE model suggest sensitivities to early summer conditions that influence when latewood formation begins, which is the portion of the tree ring used for assessing the δ¹³C-derived iWUE. Moreover, significant predictor terms in July, and in particular, September VPD for BAI are related to the whole dynamic of the annual ring formation, which includes both xylem size increase and cell wall thickening (lignification), with the latter continuing even after the former is completed, i.e., late summer or early fall (Cuny et al. 2015).

Conclusion

In summary, our data demonstrated that 20+ years of N fertilization stimulated tree growth and iWUE in Q. velutina trees from the Chronic Nitrogen Amendment Study. The length of the stimulation is unique given that the annual fertilization rate in the high-N plot was exceptionally high (ca 30 times greater than ambient), and the only indication that the stimulation was diminishing was the plateauing or declining response in a few individual trees at the end of the observation period. Trends in foliar N, a positive relationship between BAI and iWUE, as well as trends in latewood δ¹³C and δ¹⁸O give supporting evidence, suggesting that N fertilization stimulated A in the treatment trees. In the control trees, g_s was most likely the primary mechanism controlling changes in iWUE, and a constant c_i/c_a suggests a proportional regulation of A and g_s influenced by rising CO₂. Growth and iWUE were sensitive to growing season VPD, and N fertilization increased tree sensitivity to July temperature and precipitation. Lastly, with the aid of tree rings, stable isotopes and linear mixed-effects models, we were able to identify physiological responses to N fertilization and ambient CO₂, as well as detect climate signals influencing growth and iWUE in Q. velutina within this stand. These methodologies together allowed us to assess three global change components in concert, which is necessary for predicting tree responses in the future.

Conflict of interest

None declared.

Funding

This work was supported by funding from UNH-COLSA. R.G. acknowledges the support from NASA (grant NNX12AK56G S01).

Acknowledgments

We would like to acknowledge the contributions of John Aber, Michelle Day, Richard Maclean, Mary Martin, Scott Ollinger and Serita Frey for the establishment and maintenance of the Chronic Nitrogen Amendment Experiment at Harvard Forest, an NSF-funded Long-term Ecological Research site. Thanks to T. Dawson and W. Munger for providing data on δ¹⁸O in soil water from Harvard Forest. K.A.J. would also like to thank Kathleen Eggemyer for her mentoring during the early stages of the project, and the extensive lab assistance from Lauren Buzinski, Will Lynch and Conor Madison. R.G. acknowledges Stefano Leonardi (Universita’ of Parma, Italy) for useful discussion on the application of the mixed-effects model.