Water availability drives gas exchange and growth of trees in northeastern US, not elevated CO₂ and reduced acid deposition

Abstract

Dynamic global vegetation models (DGVM) exhibit high uncertainty about how climate change, elevated atmospheric CO₂ (atm. CO₂) concentration, and atmospheric pollutants will impact carbon sequestration in forested ecosystems. Although the individual roles of these environmental factors on tree growth are understood, analyses examining their simultaneous effects are lacking. We used tree-ring isotopic data and structural equation modeling to examine the concurrent and interacting effects of water availability, atm. CO₂ concentration, and SO₄ and nitrogen deposition on two broadleaf tree species in a temperate mesic forest in the northeastern US. Water availability was the strongest driver of gas exchange and tree growth. Wetter conditions since the 1980s have enhanced stomatal conductance, photosynthetic assimilation rates and, to a lesser extent, tree radial growth. Increased water availability seemingly overrides responses to reduced acid deposition, CO₂ fertilization, and nitrogen deposition. Our results indicate that water availability as a driver of ecosystem productivity in mesic temperate forests is not adequately represented in DGVMs, while CO₂ fertilization is likely overrepresented. This study emphasizes the importance to simultaneously consider interacting climatic and biogeochemical drivers when assessing forest responses to global environmental changes.

Introduction

Long-term changes in tree growth and forest productivity have been attributed to multiple climatic and biogeochemical drivers including regional changes in temperature and precipitation regimes, elevated atm. CO₂ concentration, nitrogen deposition, and atmospheric pollution^1,2. Still, a critical question remains “What are the simultaneous impacts of divergent and interacting environmental drivers of forest productivity?” Although moderate warming, higher atm. CO₂ concentration, and nitrogen deposition can enhance forest productivity and carbon sequestration^3,4, increased heat⁵, drought⁶, and atmospheric pollution² could counteract these positive effects. Disentangling the impact of these drivers on forest productivity is crucial for better anticipating future changes in biogeochemical cycles and ecosystem services.

One region where the simultaneous influence of multiple environmental drivers on forest productivity can be tested is the northeastern United States (US). Over the last decades, this region has experienced simultaneous and significant shifts in moisture availability⁷, increases in atm. CO₂ concentration, and reductions in acid and nitrogen deposition^4,8 (Fig. 1). While there has been a substantial reduction in acid deposition in northeastern US, there is still not a consensus that reduced pollutant loads have enhanced tree growth in temperate mesic forests^9,10,11. The simultaneous increase in water availability and decrease in acid deposition complicate our understanding of the potential benefits of reduced acid deposition. At the same time, the divergent influences between moisture stress and potential CO₂ fertilization have led to significant disagreement between remotely sensed (satellite) and modeled (DGVM) productivity of temperate mesic forests¹². Investigating the concurrent effects of varying environmental drivers on growth and gas exchange of trees is critical to improve DGVM and better understand the rates, magnitude, and trajectory of terrestrial carbon budgets.

Figure 1: Location of the study site and time series of the SO₄ and total inorganic N deposition, climatic water balance, and atmospheric CO₂ concentration.

(a) Location of Black Rock Forest and distribution range of Liriodendron tulipifera L. and Quercus rubra L. in North America. Data source: https://gec.cr.usgs.gov/data/little/. (b) Total SO₄ and inorganic N wet deposition per hectare for the water year (previous October to current September) at Black Rock Forest, NY. Deposition data were obtained from the National Acid Deposition Program (http://nadp.sws.uiuc.edu); 1981 to 1983 data are from Station NY51, located 6 km from Black Rock Forest, while 1984–2014 data are from Station NY99, located at Black Rock Forest. (c) Standardized climatic water balance during the summer (June–August) calculated as the difference between precipitation and potential evapotranspiration. Positive values in blue indicate wet conditions and negative values in red show dry conditions. Estimated atmospheric CO₂ concentrations from ice core data for the period 1895–1958 (dashed black line) and measured data from 1958 to 2014 (black line), CO₂ data source: http://www.columbia.edu/~mhs119/GHGs/CO2.1850-2015.txt. Figure created with R version 3.2.2⁶¹.

Trees acclimate to environmental changes at the leaf level by adjusting their stomatal conductance (g_s) and photosynthetic assimilation rates (A). These adjustments translate into changes in allocation and growth^13,14. Concurrent adjustments at the tree level interact and influence transpiration and carbon assimilation rates from stand to landscape scales¹⁵. Long-term information on physiological and environmental processes at annual and seasonal time-scales can be gained through stable isotopic analysis of tree rings^16,17. Stable isotopic analysis can be used to assess how stomatal conductance and photosynthesis respond to shifts in moisture availability¹⁸, increasing CO₂ concentration¹⁹, and reductions in acid deposition²⁰.

Here, we assess the simultaneous effects of changes in key environmental factors on gas exchange and tree growth in a temperate mesic forest of northeastern US using isotopic records from tree rings of two dominant and widely distributed tree species in eastern North America, Liriodendron tulipifera L. and Quercus rubra L. (see Methods, Fig. 1, Supplementary Table S1). Under dry conditions, L. tulipifera has an isohydric behavior and constrains its stomatal conductance so that mid-day water potential minima is kept below a critical threshold²¹. In contrast, Q. rubra shows an anisohydric behavior and maintains constant levels of stomatal conductance during drought at the risk of incurring xylem cavitation²¹. Contrasting physiological behavior and habitats of our study trees make them ideal for isolating growth and physiological responses to concurrent but divergent changes in key environmental factors. We first assess the simultaneous influences of changes in atmospheric CO₂ concentration, climatic water balance, and SO₄ and N deposition on tree growth and physiological mechanisms with structural equation models (SEM, Fig. 2). Second, we analyze the growth, carbon isotope discrimination (Δ¹³C), intrinsic water-use efficiency (_iWUE), and oxygen isotopic ratio (δ¹⁸O) responses of trees to shifting moisture conditions, from the extreme 1960s drought to repeated pluvial periods since the 1980s (Fig. 1c).

Figure 2: Hypothetical structural equation model used for assessing the influence of the summer climatic water balance, atmospheric CO₂ concentration, SO₄ and N wet deposition on tree growth (basal area increment, BAI), and tree gas exchange inferred from Δ¹³C and δ¹⁸O measured in tree rings.

Single-headed arrows indicate causal relationships and double-headed arrows denote covariation between response variables. Grey paths indicate covariation between explanatory variables.

Results and Discussion

Hypothetically, broadleaf trees in temperate mesic forests are sensitive to moisture availability because of poor stomatal regulation, low hydraulic conductance, high leaf area, and the high radiation and evaporative demands experienced by their large crowns²². We found support for this hypothesis through the high sensitivity of mature (>100 yrs old) L. tulipifera and Q. rubra to moisture availability irrespective of their physiological behavior (isohydric vs anisohydric) and site conditions (moist lowland vs shallow-soiled ridge site). Particularly, we observed a strong coupling between moisture availability and gas exchange of trees as indicated by the strong correlations between the isotopic tree-ring data and summer climatic water balance (Δ¹³C, r = 0.73 and 0.65; δ¹⁸O, r = −0.71 and −0.55) and maximum vapor pressure deficit (Δ¹³C, r = −0.70 and −0.70; δ¹⁸O, r = 0.71 and 0.56) (Table 1). Such high correlations indicate an even greater sensitivity of temperate broadleaf trees to drought than a recent analysis over the eastern US²³. Given that our period of study covers one of the wettest periods of the last 500 years⁷, if not the last 3000–5000 years²⁴, the strong sensitivity of tree gas exchange and to a lower degree of growth to moisture availability in this mesic region is particularly striking.

Table 1 Pearson’s correlation coefficients between tree-ring time series and June–August climate variables for the period 1950–2014.

Elevated atm. CO₂ concentration has been found to stimulate tree growth by indirectly enhancing photosynthetic rates and _iWUE^25,26. When simultaneously analyzing tree sensitivity to summer climatic water balance, atm. CO₂ concentration, and SO₄ and N deposition, however, we found that water availability was the most important factor. Climatic water balance during the summer (June, July, August) was the strongest driver of BAI, Δ¹³C, _iWUE, and δ¹⁸O in both species (Fig. 3). In contrast, atm. CO₂ concentration, SO₄, and N deposition, which showed significant covariation, exhibited negligible effects. Even if elevated atm. CO₂ concentration directly improved _iWUE (although this is partly due to the inclusion of CO₂ in _iWUE calculation, Supplementary Methods S1, eqn. 3), the inexistent or negative associations found between _iWUE and BAI, as well as between atm. CO₂ concentration and BAI indicate little to no stimulation of growth to CO₂. These findings match experiments in mature forests where elevated atm. CO₂ concentration and increases in _iWUE do not necessarily translate into enhanced radial tree growth²⁷. In those settings, heat and drought stress^16,28,29 or limited nutrient availability³⁰ override CO₂ effect. The lack of evidence of CO₂ fertilization effect on tree growth in our study cannot be attributed to moisture deficit or warming-induced drought stress because the period of the SEM analysis (1981–2014) is the wettest period of the instrumental period that began in 1895. Therefore, our results indicate that CO₂-induced growth enhancement²⁵ is unlikely for mature trees under natural conditions when the effects of the concomitant and significant covariation in climatic water balance, SO₄ deposition, and N deposition are considered.

Figure 3

Fitted piecewise structural equation models showing the relative influence of the summer climatic water balance, atmospheric CO₂ concentration, and SO₄ and N wet deposition on tree growth inferred from basal area increment (BAI), Δ¹³C (a,b), _iWUE (c,d), and δ¹⁸O. Period of analysis 1981–2014. The random tree identity effect (5 trees per species) was considered by fitting each response variable to a linear mixed effects model within the structural equation models. Single-headed arrows indicate causal relationships and double-headed arrows denote covariation between variables. The width of arrows is proportional to the strength of path coefficients. Numbers next to the paths indicate standardized path coefficients. Coefficients with ^ns are not significant, but improved the model fit. Solid and dashed paths indicate positive and negative effects, respectively. Grey paths indicate covariation between explanatory variables. Amount of variance explained by the model (R²) is listed for each response variable. The chi-square (Χ²) p-value, degree of freedom (df), and number of observations (n) are shown in the lower left.

Acid deposition can alter leaf physiology and stomatal conductance, indirectly modify isotope ratios in tree rings²⁰, and influence tree growth⁹. When we simultaneously analyzed the effects of the climatic water balance, atm. CO₂ concentration, and atmospheric deposition on trees, we did not detect any direct effect of SO₄ and N deposition on BAI and only found some direct but small effects of SO₄ and N deposition on Δ¹³C, _iWUE, and δ¹⁸O (Fig. 3). However, these direct effects on tree-ring isotopic ratios were not translated into changes in growth. Inexistent or negative correlations were found between Δ¹³C or _iWUE and BAI. L. tulipifera, a species with arbuscular mycorrhizal association, may show higher growth to increased availability of inorganic N from atmospheric deposition^4,31. However, the SEM indicated no direct positive association between N deposition and BAI of L. tulipifera. Similarly, the BAI of Q. rubra did not show any direct association with SO₄ and N deposition despite the presence of some minor correlations between tree-ring Δ¹³C and δ¹⁸O with N and SO₄, respectively. These results do not support previous findings that showed a high sensitivity of this species to S and N deposition³² and N-induced growth enhancement under natural conditions^4,33. The absence of growth response to N deposition found for both species agrees with the results of an N-addition experiment done in the Catskill Mountains of southeastern New York State where N addition had no significant effects on aboveground biomass production³⁴.

Our SEM analysis indicates that moisture is the primary driver of gas exchange and growth for both species, even during an anomalously wet period. From these analyses, the increase in forest growth recently observed in the northeastern US^4,9,35 is likely less related to rising atm. CO₂ concentration and changes in acid/nitrogen deposition and more likely driven by regional wetting^7,35. Supporting this inference, change-point detection analysis of the climatic water balance time-series identified a tipping point in moisture availability with drier conditions prior to 1983 and wetter after (Fig. 4a,b). Concurrent to this water availability increase, L. tulipifera and Q. rubra exhibited a simultaneous shift in BAI, Δ¹³C, _iWUE, and δ¹⁸O (Fig. 4) indicating a strong coupling between growth, gas exchange, and moisture conditions. A similar increase in tree-ring Δ¹³C and enhancement in growth was found in the Q. rubra Harvard Forest eddy-flux tower forest following the regional increased in water availability³⁵.

Figure 4: Trends and probability density functions.

Climatic water balance (a,b), basal area increment (c,d), Δ¹³C (e,f), _iWUE (g,h), and δ¹⁸O (i,j) of Liriodendron tulipifera and Quercus rubra for the dry 1950–1983 (light orange fill) and wet 1984–2014 (light blue fill) period. The dry and wet period were identified in the climatic water balance time series applying change-point detection test. Significance levels of the slopes from Mann-Kendall trend tests and Theil-Sen trend estimates: *P < 0.05; **P < 0.01; ***P < 0.001. P-values from the Kolmogorov–Smirnov tests between the dry and wet period are shown. Variables were standardized (z-score) before.

While BAI, Δ¹³C, and _iWUE mainly increased as the climate became wetter, tree-ring δ¹⁸O started to decrease (Fig. 4) because of changes in isotopic composition of water sources and stomatal response to lower evaporative demand of the atmosphere. Variation in tree-ring δ¹⁸O is primarily due to evaporative enrichment at the leaf level, biochemical fractionation during oxygen incorporation, and isotopic signature of tree source water, which is mainly influenced by the δ¹⁸O of precipitation and soil evaporative enrichment^36,37. The δ¹⁸O value of precipitation is essentially influenced by air temperature, precipitation amount, moisture sources, air mass trajectory, and seasonality³⁸.

In our study region, the δ¹⁸O of precipitation has changed through time (Supplementary Methods S2, Fig. S1). Between 1968 and 2010, a significant reduction in δ¹⁸O of precipitation (−0.089‰ yr⁻¹) was recorded in northeastern US due in large part to the increase in the proportion of Arctic precipitation sources which are more depleted in δ¹⁸O ratios³⁸. This decrease in δ¹⁸O of precipitation may have potentially influenced the isotopic signature of the source water of our trees and caused a gradual reduction in tree-ring δ¹⁸O values through time (Fig. 4i). However, the strong coupling between tree-ring δ¹⁸O and moisture availability (Table 1) indicates that the decrease in tree-ring δ¹⁸O was also due to a reduction in transpiration at the leaf level in response to the lower evaporative demand of the atmosphere as the climate became wetter.

Synchronically to the decrease in δ¹⁸O tree-ring values, rising Δ¹³C and BAI (Fig. 4e,c) suggest that C-assimilation and stomatal conductance also increased. Taken together, the concurrent depletion of δ¹⁸O and increase in Δ¹³C indicate that, despite the reduction in δ¹⁸O of precipitation over the study region, changes in transpiration, stomatal conductance, and photosynthetic assimilation rates occurred simultaneously and tracked the abrupt shift in climatic water balance that began in the early 1980s. The long-term trends recorded in our tree-ring δ¹⁸O time series, however, should be interpret with caution as changes in the isotopic signature of source water in time and the reduction in evaporative enrichment at the leaf level driven by the increase in water availability have likely occurred at the same time.

Overall, we found that moisture is the main driver of increased gas exchange and, to a lesser degree, increased radial growth of two broadleaf trees in a mesic temperate forest in the northeastern US during the period of opposing trajectories of acid and N deposition, atm. CO₂ concentration, and water availability. Simultaneous analysis of these drivers on tree-ring isotopic composition and growth indicates that the reported growth recovery from reduced acid deposition⁹ and N-induced growth enhancement in northeastern US forests³³ might be the result of a possible omission of the concurrent shift in water availability and atm. CO₂ concentration. Additionally, results here do not find support for atm. CO₂ fertilization on broadleaf tree growth in mature temperate mesic forests of northeastern US²⁵, a region where greening trends have been mainly attributed to increased atm. CO₂ concentration and land cover change¹. Our work emphasizes the need to simultaneously consider changes in water availability, atm. CO₂ concentration, and acid/nitrogen deposition at large spatial scales to gain a more complete understanding of future changes in forest productivity.

As the climate is getting warmer and wetter in northeastern US (Supplementary Fig. S2), our observed sensitivity of broadleaf trees to moisture availability is important. The expected increases in severity, frequency and duration of drought periods would likely have a significant impact on tree growth, mortality rates, and forest composition³⁹. Although the temperate mesic forests of the northeastern US have not experienced severe and long-lasting drought since the 1960s⁷, the strong and persistent response of trees to water availability reported here as well as in recent studies^35,39 reveals a vulnerability of mesic forests to drought.

Methods

Study sites and dendrochronological analysis

We conducted this study at Black Rock Forest (41°24′N, 74°01′W), a 1550 ha forest preserve in southeastern New York State⁴⁰. We sampled L. tulipifera in a lowland site located on a south-facing slope at 170 m a.s.l. on loamy soils and Q. rubra at the ridge of an upper slope site at 400 m a.s.l., situated 2 km away from the lowland site, and characterized by shallow soils with abundant rock outcrops. We extracted two 5 mm diameter increment cores from 15 dominant and healthy L. tulipifera and Q. rubra trees for tree-ring width measurements (Supplementary Table S1). The increment cores were air dried, glued on wood mounts, and successively sanded with finer grades of sandpaper until the xylem structure and ring boundaries were clearly visible. We measured ring widths to the nearest 0.001 mm. Individual tree-ring width series were crossdated and statistically checked with the program COFECHA⁴¹. To ensure that the number of trees sampled was sufficient and representative of the sampled population, we calculated the express population signal (EPS). All tree-ring chronologies showed EPS values ≥ 0.85, which is considered the threshold value for adequately reflecting a common signal among trees (Supplementary Table S1)⁴². To detect long-term changes in growth, we converted the individual raw tree-ring width series to basal area increments (BAI) and removed the potential age related trends that can bias long-term growth changes with a Regional Curve Standardization (RCS) approach (Supplementary Fig. S3)⁴³. We calculated an average ontogenetic growth curve for each species (i.e., the regional curve) by aligning the raw BAI measurements of each tree to the biological age of the rings. We then divided each raw individual BAI series by this average curve to produce RCS residual BAI series^44,45. RCS residual BAI series were used in further analyses.

Isotopic analysis

Tree-ring δ¹³C was used to calculate carbon isotope discrimination (Δ¹³C, Supplementary Methods S1). Δ¹³C in tree rings provides an integrated record between intercellular and atm. CO₂ concentration during the period when the carbon was fixed by the enzyme Ribulose-1,5-bisphosphate carboxylase/oxygenase (RuBisCO) in chloroplasts⁴⁶. With further calculations, _iWUE, i.e., the ratio between A and g_s, can be determined from Δ¹³C (Supplementary Methods S1)⁴⁷. By contrast, the oxygen isotopic ratios in tree rings integrate the isotopic composition of source water and the stomatal response to changes in vapor pressure deficit^36,37. Thus, the δ¹⁸O ratio contains an indirect record of g_s and can help understanding the influence of A and g_s on Δ¹³C and _iWUE¹⁸. Therefore, by examining concurrent variations in both Δ¹³C and δ¹⁸O, insights can be gained on how stomatal conductance and photosynthesis respond to shifts in climatic water balance, increasing atm. CO₂ concentration, and reductions in acid deposition.

For the isotopic analysis, we selected the five trees per species with the highest correlations with the tree-ring width master chronology and took an extra 12 mm diameter core per tree. We analyzed the isotopic ratios in each tree and each annual ring individually for the period 1950–2014. From each core, we split off the latewood of each annual ring with a scalpel under a stereomicroscope, chopped the material, and stored each latewood sample individually in centrifugal tubes before cellulose extraction. We extracted the α-cellulose following standard procedures^48,49 and homogenized the cellulose using an ultrasound treatment⁵⁰. For each sample, 200 μg of cellulose were weighted and put in silver capsules. δ¹³C and δ¹⁸O were measured simultaneously using high-temperature pyrolysis in a Costech elemental analyzer interfaced with an Elementar Isoprime mass spectrometer at the Department of Geology at the University of Maryland, USA⁵¹. The analytical precision for the in-house α-cellulose standards was ±0.17‰ for δ¹³C and ±0.34‰ for δ¹⁸O.

Climate and atmospheric deposition data

Modeled mean monthly minimum and maximum temperature, total monthly precipitation and maximum vapor pressure deficit for the period 1895–2014 were obtained from the PRISM Climate Group, Oregon State University (http://prism.oregonstate.edu). We used the mean monthly minimum and maximum temperature and total precipitation to compute a monthly climatic water balance, i.e., precipitation minus potential evapotranspiration. Evapotranspiration was calculated according to Hargreaves⁵².

To assess the potential effect of atm. CO₂ concentration and pollutants on tree growth and physiology, we used historical SO₄ and total inorganic N wet deposition data (kg/ha) for the water year (previous October to current September) from two measuring stations from the National Acid Deposition Program (Fig. 1, http://nadp.sws.uiuc.edu), and annual global average CO₂ mixing ratio values available online (http://www.columbia.edu/~mhs119/GHGs/CO2.1850-2015.txt).

Data analysis

We first explored the relationships of tree-ring chronologies (BAI, Δ¹³C, δ¹⁸O) with monthly climate variables with bootstrapped correlation functions using the R package treeclim⁵³. We identified the months (i.e., June, July, and August) that have the strongest and significant influence on tree-ring chronologies (Supplementary Figs S4, S5) and further averaged the climate variables over these months and calculated correlation coefficients between summer climate and tree-ring variables (Table 1). Tree-ring and climate time series were prewhitened (removal of the first order autocorrelations) before correlation analysis. The summer climatic water balance was used in further analyses since it showed the strongest correlations with the tree-ring variables.

We performed piecewise structural equation modeling (SEM) using the R package piecewise SEM⁵⁴ to account for concomitant changes in atm. CO₂ concentration, SO₄ and N deposition, and climatic water balance, which may be masking the influence of a single environmental variable on the growth and gas exchange of trees (Fig. 2). Piecewise SEMs use advanced multivariate statistical techniques better suited for small sample sizes and allow the simultaneous implementation of non-normal distributions, random effects, and different correlation structures within a traditional SEM framework^54,55,56. We developed piecewise SEMs for each species to address the joint effects of changing climatic water balance, atm. CO₂ concentration, and SO₄ and N deposition on BAI, Δ¹³C and δ¹⁸O of trees. We considered the random tree identity effect (individual tree series) by fitting each response variable to a linear mixed effects model⁵⁷ using the function lme from the NLME package⁵⁸. The SEMs were fit for the period 1981–2014, i.e. the time window with available deposition data, and raw tree-ring and environmental data to account for effects of the trends in environmental variables. We assessed the models fits using chi-square p-value and Akaike’s information criterion corrected for small sample size⁵⁴.

We used the change-point detection test of Pettitt⁵⁹ to test the shift in the central tendency of the climatic water balance time series. Based on this test, we identified two significant periods with contrasting climatic water balance, i.e., the dry period 1950–1983 and the wet period 1984–2014. We assessed the significance of the changes in BAI, Δ¹³C, _iWUE and δ¹⁸O in the dry vs. wet period with probability density functions and Kolmogorov–Smirnov tests. We estimated the temporal trends of the climatic water balance, BAI, Δ¹³C, _iWUE and δ¹⁸O time series, and their significance with Mann-Kendall trend tests and Theil-Sen trend estimates⁶⁰. All data analyses were conducted in R version 3.2.2⁶¹.