|Home | About | Journals | Submit | Contact Us | Français|
Understanding how tree growth strategies may influence tree susceptibility to disturbance is an important goal, especially given projected increases in diverse ecological disturbances this century. We use growth responses of tree rings to climate, relationships between tree-ring stable isotopic signatures of carbon (δ13C) and oxygen (δ18O), wood nitrogen concentration [N], and contemporary leaf [N] and δ13C values to assess potential historic drivers of tree photosynthesis in dying and apparently healthy co-occurring northern red oak (Quercus rubra L. (Fagaceae)) during a region-wide oak decline event in Arkansas, USA. Bole growth of both healthy and dying trees responded negatively to drought severity (Palmer Drought Severity Index) and temperature; healthy trees exhibited a positive, but small, response to growing season precipitation. Contrary to expectations, tree-ring δ13C did not increase with drought severity. A significantly positive relationship between tree-ring δ13C and δ18O was evident in dying trees (P < 0.05) but not in healthy trees. Healthy trees’ wood exhibited lower [N] than that of dying trees throughout most of their lives (P < 0.05), and we observed a significant, positive relationship (P < 0.05) in healthy trees between contemporary leaf δ13C and leaf N (by mass), but not in dying trees. Our work provides evidence that for plants in which strong relationships between δ13C and δ18O are not evident, δ13C may be governed by plant N status. The data further imply that historic photosynthesis in healthy trees was linked to N status and, perhaps, C sink strength to a greater extent than in dying trees, in which tree-ring stable isotopes suggest that historic photosynthesis was governed primarily by stomatal regulation. This, in turn, suggests that assessing the relative dominance of photosynthetic capacity vs stomatal regulation as drivers of trees’ C accrual may be a feasible means of predicting tree responses to some disturbance events. Our work demonstrates that a dual isotope, tree-ring approach can be integrated with tree N status to begin to unravel a fundamental question in forest ecology: why do some trees die during a disturbance, while other conspecifics with apparently similar access to resources remain healthy?
Predicting how trees will respond to disturbance events is a challenge of increasing importance, given projected increases in the frequency of diverse kinds of disturbance events in the coming decades (McKenney et al. 2007, Woodall et al. 2009). Multiple studies explore the physiological distinctions among species that can drive tree survival or mortality during disturbances (Breshears et al. 2005, 2009, McDowell et al. 2008; Eilmann and Rigling 2012, Hu et al. 2014, Mitchell et al. 2014, Renninger et al. 2014), but in spite of some studies of insect–plant interactions (e.g., Edmunds and Alstad 1978, Whitham and Slobodchikoff 1981), far fewer investigate why conspecific individuals may diverge in their response to disturbances. Given that intraspecific variation in response to environmental conditions can be as great as or even greater than variation across species (Clark et al. 2004, Clark 2010), investigating variation in individual tree responses to a disturbance event within a population, especially when trees have apparently similar access to resources, can make important contributions to our understanding of patterns of tree mortality (Billings et al. 2015).
Tree-ring studies are particularly helpful for developing an understanding of tree responses to their environment. For example, growth indices provide evidence of distinct drought responses among co-occurring species (Adams et al. 2009) and conspecifics (Haavik et al. 2008, Levanic et al. 2011), varied growth rates over time (Johnson and Abrams 2009) and apparent susceptibility to disturbances such as oak decline events (Voelker et al. 2008). Recent work describes how tree rings can reveal linkages between tree growth strategies and survival during disturbance (Billings et al. 2015), and for decades, investigators have used tree rings to help understand historical responses of trees to insect outbreaks (Muzika and Liebhold 1999, Girardin et al. 2002, Simard et al. 2008).
Stable isotopic signatures of carbon (δ13C) and oxygen (δ18O) of tree rings (typically the α-cellulose extracted from rings; McCarroll and Loader 2004), in spite of the complex and concurrent processes driving their values (Roden and Farquhar 2012, Gessler et al. 2014), can augment knowledge gained from tree-ring growth patterns by permitting inferences of drivers of past tree C dynamics (Barbour et al. 2001). Isotopic theory suggests that the ratio of leaf intercellular [CO2] to atmospheric [CO2] (ci/ca), which depends on photosynthetic capacity and the response of stomatal conductance (gs) to moisture availability, is reflected in biomass δ13C values (Farquhar et al. 1982, 1989, Francey and Farquhar 1982). Because fluctuations of δ13C in tree-ring α-cellulose are associated with both historic gs and photosynthetic capacity, δ13C of plant tissue is often used to infer a plant’s response to atmospheric demand for water (Leavitt and Long 1989, Hubick and Gibson 1993, Saurer et al. 1995) and, less frequently, N availability, which can govern a plant’s C sink strength (Gebauer and Schulze 1991). In contrast to δ13C, δ18O of vegetation appears insensitive to photosynthetic capacity (Scheidegger et al. 2000, Barbour et al. 2001, Barbour 2007). Instead, δ18O of vegetation reflects the isotopic signature of source water taken up during its formation, effects of evaporation and diffusion during transpiration, and internal exchange of O between water and carbonyl groups during organic compound formation (Sternberg et al. 1986, Yakir 1992, Farquhar and Lloyd 1993, Barbour et al. 2002, Barbour 2007). Because of the linkages between δ18O of vegetation and source water, vegetation δ18O can reflect rooting depth, given that deep water sources are often δ18O-depleted relative to shallower water sources (Roden et al. 2005, Pataki et al. 2008). Because water evaporation and diffusion during transpiration discriminate against 18O and are positively linked to gs, gs is predicted to have a negative influence on plant biomass δ18O (all else being equal); this influence likely is small, but is stronger for leaves that are relatively less coupled to their microclimate, as is the case for broad-leaved trees relative to conifers (Barbour et al. 2002). Thus, gs can influence both vegetation δ13C and δ18O, especially in broad-leaved forests, and photosynthetic capacity can influence δ13C but, apparently, not δ18O.
Exploiting this phenomenon, investigators have used plant biomass δ13C and δ18O to infer the importance of CO2 source strength (i.e., gs) relative to C sink strength (i.e., photosynthetic capacity) as limitations to growth (Barbour and Farquhar 2000, Scheidegger et al. 2000, Barbour et al. 2002, Grams et al. 2007, Cabrera-Bosquet et al. 2010). Because δ13C, but not δ18O, can vary with photosynthetic capacity, and both δ13C and δ18O are predicted to have a qualitatively similar (negative) response to gs, isotopic theory predicts a positive relationship between vegetation δ13C and δ18O if gs is more limiting to photosynthesis than photosynthetic capacity, all else being equal (Scheidegger et al. 2000, Barbour et al. 2002). Theory further predicts that the slope of any such relationship will increase with atmospheric water demand (Barbour et al. 2002). In contrast, models predict no relationship between δ13C and δ18O if photosynthetic capacity is the limiting factor, assuming all other parameters remain constant (Scheidegger et al. 2000, Barbour et al. 2002, Hilasvuori and Berninger 2010). Such analyses thus can be used to infer the relative extent to which gs vs photosynthetic capacity govern variation in δ13C of tree-ring α-cellulose and the relative degree to which these parameters have driven photosynthesis and growth. This approach is best employed among co-occurring trees with similar access to light, given the importance of light as a driver of photosynthetic capacity (Yakir and Israeli 1995).
Integrating measures of tree nitrogen (N) status using both wood and contemporary leaves can further elucidate the drivers of tree responses to variation in climate and resource availability across time (Poulson et al. 1995, Amundson et al. 2003, Guerrieri et al. 2011). Although N translocation after wood formation sometimes can confound tree-ring [N] and δ15N (Poulson et al. 1995), tree-ring δ15N has been applied successfully in studies of ecosystem N loss and retention (McLauchlan et al. 2007), tree sources of N (Poulson et al. 1995) and the impacts of N deposition (Guerrieri et al. 2011) and fertilization (Balster et al. 2009) on tree growth. Links between increased N availability and enhanced photosynthetic capacity (Evans 1989, Reich et al. 1999) and between photosynthetic capacity and tree-ring δ13C (Farquhar et al. 1982, Dawson and Ehleringer 1993) are well established, suggesting that assessing tree-ring and leaf δ15N and [N] data in conjunction with tree-ring growth and δ13C likely can help illuminate the degree to which tree responses to environmental conditions depend on tree N status. Given the link between plant N status and photosynthetic capacity (Evans 1989, Reich et al. 1999), vegetation N analyses may provide guidance for inferring drivers of tree C dynamics from tree-ring δ13C and δ18O.
To date, dual isotopic approaches to tree-ring analyses typically are invoked to explore drivers of both δ13C and δ18O and to develop conceptual and quantitative models for understanding these factors. Here, we apply these ideas and incorporate them with investigations of tree N status to begin to unravel why some trees died during a disturbance event while conspecifics of similar stature and with apparently similar access to resources remained healthy. Specifically, we use tree-ring growth patterns, δ13C and δ18O of wood-derived α-cellulose, wood [N], and leaf [N] and δ13C to infer the ecophysiological mechanisms influencing conspecific trees’ susceptibility to a decades-long insect outbreak in northern red oak-dominated (Quercus rubra L. (Fagaceae)) forests of northwest Arkansas, USA. Following historically high infestations of a native, wood-boring insect (red oak borer, Enaphalodes rufulus (Haldeman) (Coleoptera: Cerambycidae)) from the mid-1970s through the early 2000s, trees in these forests experienced high mortality rates on a landscape scale (Stephen et al. 2001, Muzika and Guyette 2004, Fierke et al. 2005a, 2007, Haavik and Stephen 2011, Haavik et al. 2015). Red oak borer larvae bore directly through bark into phloem tissue where they feed and overwinter; they then move into sapwood and heartwood, form galleries where they overwinter again, and then emerge as adults (Fierke et al. 2005a). At low red oak borer attack densities, trees appear to tolerate their presence well, but repeated years of high numbers of attacks can lead to tree death (Fierke et al. 2005b). In the current study, many forest stands experiencing oak mortality contained multiple, apparently healthy trees immediately adjacent to trees experiencing severe crown dieback and subsequent mortality.
Poor site quality and even-aged stands may have been contributing factors to this oak decline event (Oak et al. 1996), but several studies implicate cyclical drought beginning decades prior to observable decline in the region as a key factor determining oak vigor at these sites (Tainter et al. 1990, Jenkins and Pallardy 1995, Haavik et al. 2008, 2010). Indeed, tree-ring growth patterns at these sites suggest that antecedent drought was an important factor governing tree susceptibility to insect infestation and eventual mortality (Haavik et al. 2008, Haavik and Stephen 2011). However, δ13C values of α-cellulose from tree rings of more susceptible trees in nearby stands did not exhibit the increases one might predict with drought stress from isotopic theory (Haavik et al. 2008). Equally intriguing, time series forecasting models suggest that trees surviving this regional disturbance exhibited nonlinear growth dynamics (nonuniform growth responses to environmental cues) all their lives, in contrast to the uniform growth responses to environmental cues of those that died (Billings et al. 2015). Time series forecasting models cannot identify mechanisms driving a dynamic signal, but one plausible explanation for nonlinear growth dynamics in surviving trees may be an ability to take advantage of resource pulses in ways inaccessible to dying trees. Although drought likely played an important role in governing tree response to this disturbance event (Haavik et al. 2008, Haavik and Stephen 2011), the mechanisms governing the variable susceptibility of co-occurring trees remain mysterious.
We explore these issues, working at a subset of the sites investigated in the forecasting modeling work of Billings et al. (2015) to assess how tree growth strategies may have influenced susceptibility vs survival of co-occurring conspecifics. Because drought has been implicated as a driver of tree susceptibility to this insect-associated disturbance (Haavik et al. 2008, 2010), we might anticipate that declining trees’ tree rings may reveal evidence of moisture availability serving as a greater constraint to growth relative to their surviving neighbors. We assess how tree-ring growth responded to past climatic conditions, invoke δ13C and δ18O of tree-ring α-cellulose to infer the historic relative importance of gs vs photosynthetic capacity as drivers of δ13C (Barbour and Farquhar 2000, Barbour et al. 2002), and use wood and contemporary leaf [N] to infer how tree N status and its presumed link to, photosynthetic capacity, may have influenced tree survival or mortality. We address three questions relevant for understanding tree responses to this oak decline event and, more broadly, oak-dominated forest strategies for coping with disturbance: (i) Do trees that remained apparently healthy throughout this disturbance exhibit different historical growth responses to climate parameters compared with those that suffered mortality? (ii) Do surviving trees exhibit distinct relationships between δ13C and δ18O in tree-ring α-cellulose, suggesting different dominant drivers of historic C acquisition in these trees (i.e., C source vs C sinks) compared with their dying neighbors? (iii) Do tree rings or contemporary leaves reveal differences in the influence of N status on C acquisition between co-occurring, surviving and dying trees?
We sampled healthy and dying oaks at three pre-established study sites in the Boston Mountains of the Ozark National Forest in northwest Arkansas, USA (Fierke et al. 2007). The three sites, Oark (35° 43′ 21″ N, 93° 24′ 18″ W), Fly Gap (35° 44′ 10″ N, 93° 46′ 49″ W) and White Rock (35° 40′ 43″ N, 93° 58′ 04″ W), were selected on the basis of red oak borer presence. Details of site conditions can be obtained in Fierke et al. (2007). Briefly, the three sampling sites are oak-hickory dominated (USDA Forest Service 1999), support similarly aged oaks (mean of 80 years) on acidic, clay-rich soils with low organic matter content (Sander 1965), and are all within 50 km of each other. Regional climate is temperate with warm summers and mild winters; mean annual temperature is 16 °C. Mean annual precipitation totals 124 cm and falls mostly during spring, summer and fall. Bedrock composition is composed of limestone, shale and sandstone, with cliff elevations reaching up to 750 m. Depth to bedrock was relatively shallow at all sampling sites, limiting the variation in depth from which trees could take up soil water.
Climate records were obtained from the National Oceanic and Atmospheric Administration (NOAA) website (http://www1.ncdc.noaa.gov/pubdata/cirs) for the climate regions of interest. Mean annual temperature for the years of interest was calculated using average annual temperature across the four NOAA regions encompassed by the sampling area. Similarly, total annual precipitation for the years of interest was calculated using annual precipitation for each calendar year across those same NOAA regions. We determined seven time periods of specific interest consisting of alternating periods of extreme wet and dry conditions, using the NOAA Palmer Drought Severity Index (PDSI; Palmer 1965) and the regional climate data. The PDSI is an index of moisture availability derived from water balance equations (Palmer 1965). The time periods of specific interest because of their relatively high or low PDSI values were 1952–56 (dry), 1957–59 (wet), 1962–67 (dry), 1972–75 (wet), 1979–81 (dry), 1992–95 (wet) and 1998–2001 (dry). We also calculated average (temperature, PDSI) and total (precipitation) values for May through August to use as growing season climate indices.
At each site, we sampled trees experiencing moderate to heavy levels of red oak borer infestation, as well as their apparently healthy neighbors, in 2002, 2003 and 2007. Tree health was assessed based on the number of borer emergence holes and percent crown cover (Fierke et al. 2007), and we created categorical groups of apparently healthy trees (referred to throughout as healthy trees) and trees experiencing severe crown dieback (dying trees). All trees, regardless of health status, were well interspersed with no apparent differences in access to belowground resources or light; all trees exhibited similar canopy stature, height, diameter at breast height (DBH) and age. Six healthy trees from Fly Gap were sampled during 2002 and 2003. At each of the three sites in 2007, we sampled two trees categorized as healthy (six total) and three dying trees (nine total), for an additional 15 trees. Along with the six trees felled in 2002–03, these trees comprised our δ13C and δ18O data set (n = 21). We scanned cookies, cut from tree boles sampled in 2002, 2003 and 2007, into Adobe Acrobat 8 Professional, and measured and dated ring widths using event-year markers as outlined in Haavik et al. (2008). Most trees exhibited ring width series from 1930 to date of felling. We cross-dated the tree rings using COFECHA software (Holmes 1983), which ensures proper dating through event-year comparisons. We measured tree-ring widths to the nearest 0.01 cm on a total of three, evenly spaced radii around each tree’s cookie, and averaged these widths to determine ring width in each year. Though growth patterns were assessed throughout each tree’s life, we isolated tree rings only from the seven periods of climate extremes for isotopic analyses, to maximize the climate signals captured.
We generated between 1 and 2 g of whole-wood sawdust using a Dremel tool (Bosch Tool Corporation, Chicago, IL, USA) from each cookie for each of the seven time periods, being careful to avoid the earlywood visible at the start of each time period to limit the inclusion of C fixed prior to the period of interest (McCarroll and Loader 2004). Sawdust was placed in fiber filter bags (ANKOM, Fairport, NY, USA) and heat sealed for α-cellulose extraction as per a procedure modified from Leavitt and Danzer (1993). We placed the filter bags into a Soxhlet apparatus, where they boiled for 24 h in a 2 : 1 mixture of toluene : ethanol, followed by another similar treatment with replaced extractant. After drying, samples were boiled for two successive 24 h periods with 95% ethanol. Bags were dried and then boiled for 1 h in a mixture of sodium chlorite, glacial acetic acid, and de-ionized water to extract soluble sugars and low molecular mass polysaccharides. This procedure was performed three times, rinsing samples completely between cycles. Samples were then washed with sodium hydroxide and glacial acetic acid and subjected to a final rinsing. The remaining α-cellulose was then oven dried for 24 h at 70 °C.
In 2007, we collected five clusters of full-sun leaves from the top of five healthy trees and eight dying trees (the same trees sampled for tree-ring analyses). We stored all leaves in coolers until return to the laboratory, where they were frozen in flattened, air-tight bags for later analysis. We subsequently scanned leaves in Adobe Photoshop and calculated leaf area using ImageJ software (National Institutes of Health (1997–2011), Bethesda, MD, USA). Area measurements were used to generate areal indices of leaf N concentrations (Narea). Leaves were then dried at 65 °C for >48 h, weighed and ground to produce homogenized samples.
We weighed α-cellulose (0.09–0.11 mg) and whole-wood sawdust samples (15–20 mg) into tin capsules for analysis of δ13C (α-cellulose) and δ15N and %N (sawdust), respectively. Following procedures for nonresinous trees like northern red oak (Bukata and Kyser 2005), we did not wash sawdust with distilled water prior to N analyses. Tin capsules were combusted with excess O2 in an elemental analyzer (EA-1100, Carlo Erba, Milano, Italy) connected through an open split interface (Conflo II, Finnigan MAT, Bremen, Germany) to an IRMS (Delta-S, Finnigan MAT) at the University of Kansas Keck Paleoecology Stable Isotope Laboratory (KPESIL). We loaded ~1 mg of tree-ring α-cellulose into silver capsules for δ18O analysis via pyrolysis to CO on an elemental analyzer (EA-1110, Carlo Erba) coupled with a continuous flow mass spectrometer (Delta Plus, Finnigan MAT) at KPESIL. All leaf samples were analyzed for [C], [N], δ13C and δ15N using an elemental analyzer coupled to an Isotope Ratio Mass Spectrometer (Delta Plus IRMS, Thermofinnigan, San Jose, CA, USA) at Kansas State University. Isotopic values are reported as delta values (in per mil, ‰) relative to international standards (V-PDB for δ13C, V-SMOW for δ18O and atmospheric N2 for δ15N).
We calculated basal area increments (BAIs) for all years in all trees from measurements of ring width and DBH measurements using the following formula:
where Rn represents the radius in a particular year and Rn−1 the radius of the previous year, with the assumption that stem growth approximates the area of a circle (Fritts 1976). For analyses using ring widths, we produced ring width indices (RWI) by standardizing raw ring widths to zero for each individual tree. We calculated both BAI and RWI because of their respective utilities for examining growth trends across time (BAI) and the influence of multiple features of climate on growth (RWI; Johnson and Abrams 2009). Basal area increment is generally used in the field of dendroecology to reflect growth over time, while RWI is more typically used for examining physiological responses to climatic conditions (Esper et al. 2002, Johnson and Abrams 2009).
We used analysis of variance (ANOVA) to test for the influence of site and tree health class on BAI and RWI. Because site was a significant factor driving growth, we performed subsequent statistical tests of growth-related features for each site separately. At each of the three sites, we used regression analyses to assess relationships between these two indices of growth and three climate variables: PDSI, precipitation and temperature. We performed these regressions using both annual or, in the case of PDSI, multi-annual climate measures, and those calculated for May through August, the main part of the growing season. We also assessed potential lags in growth responses to climate, but present here only the relationships of growth to current year climate data, which were more robust.
We used these same climate-related variables to test for their influence on tree-ring stable isotopic and [N] data as well, using ANOVA. When site was not a significant driver of the response variable for these analyses, we grouped all sites’ data together. We employed ANOVA to test for the influence of relatively wet vs dry conditions on δ18O and δ13C of tree-ring α-cellulose and on δ15N and N content of tree-ring whole-wood sawdust. We used repeated-measures ANOVA to assess the influence of site, health status, time and their interactions on tree-ring α-cellulose δ13C and δ18O and tree-ring wood δ15N and [N]. Further, we used correlation analyses to assess relationships between tree-ring α-cellulose δ13C and δ18O in trees of different health status.
We also analyzed contemporary leaf data, focusing on healthy (n = 5) and dying (n = 8) trees. We tested for the influence of site, health status and their interaction on 2007 leaf sample [C], [N], δ13C and δ15N using ANOVA. We used correlation analysis to assess relationships between these isotopic signatures and three different measures of leaf N : Nmass , Narea and leaf C : N ratios (by mass). For all analyses, we determined statistical significance at P < 0.05. All errors presented are 1 SEM. All analyses were performed in SAS (version 9.1.3, SAS Institute, Cary, NC, USA).
In spite of statistically similar DBH at the time of sampling among tree health classes, at all three sites BAI of trees that remained healthy during the high mortality event diverged from those that died (P < 0.05, Figure Figure1).1). This divergence occurred decades prior to any obvious mortality (Figure (Figure1).1). Both healthy and dying trees exhibited positive correlations between RWI and PDSI, and negative correlations between RWI and temperature (Table (Table1).1). Only healthy trees exhibited positive correlations between RWI and precipitation; these correlations for healthy trees were present only at one site for annual average precipitation and at all three sites for average May–August precipitation.
In all trees, tree-ring α-cellulose δ13C ranged from −27.3 to −23.0‰. Values of δ13C did not vary with health class, site, or across time (Figure (Figure2),2), or with any assessed climate variable. δ18O of tree-ring α-cellulose varied from 27.5 to 31.8‰ and exhibited a significant, positive relationship with temperature (P < 0.001), and significant, negative relationships with precipitation (P < 0.05) and PDSI (P < 0.05) when examined for all trees at all sites. An interaction among year, site and health class was a significant determinant of δ18O. δ18O values of tree rings at one site (Oark) were lower in dying than in healthy trees, exhibiting an ecologically small but statistically significant (0.6‰) difference (29.2 ± 0.2 vs 28.6 ± 0.2‰, respectively, P < 0.05, Figure Figure3).3). Average δ18O of tree-ring α-cellulose among both healthy and dying trees was significantly depleted at that same site (Oark, 28.8 ± 0.1‰) compared with the other two sites (White Rock, 29.3 ± 0.1‰, P < 0.01; Fly Gap, 29.2 ± 0.1‰, P < 0.05); though these differences were statistically significant, their magnitude was small (0.5 and 0.4‰, respectively).
There was no difference in tree-ring [N] among sites. Across all sites, we observed a significant difference in tree-ring [N] between dying and healthy trees, with dying trees exhibiting greater tree-ring [N] than their healthy neighbors (Figure (Figure4).4). Tree-ring [N] exhibited relatively slight variation across time until the final sampling period (1998–2001), when [N] in healthy trees increased 34.8% from the lowest point to an average of 1.15 mg g−1, and [N] in dying trees increased 45.6% from the lowest point to an average of 1.43 mg g−1. There were no differences in tree-ring δ15N between sites or tree health status, and no change in tree-ring d15N across time.
Nitrogen concentrations in 2007 leaves ranged from 19.9 to 35.3 mg g−1. We observed no significant site differences or differences between tree health classes in any measure of leaf N or C : N, with one exception: site was a determinant of leaf Nmass. There were small but significant differences in leaf Nmass between Fly Gap and Oark (P < 0.005, 28.9 ± 1.2 and 22.8 ± 0.8 mg g−1, respectively) and between White Rock and Oark (P < 0.05, 24.4 ± 1.2 and 22.8 ± 0.8 mg g−1, respectively). Leaf δ15N was significantly, negatively related to leaf C : N ratios across all sites (P < 0.05, R2 = 0.19), with no interaction with tree health status.
Across all sites, tree-ring δ13C and δ18O from healthy trees revealed no significant correlation (Figure (Figure5a),5a), but dying trees exhibited a significant, positive relationship between δ13C and δ18O (P < 0.05, R2 = 0.57, Figure Figure5b).5b). We also observed a significant, positive relationship between leaf δ13C and leaf Nmass of healthy trees (P < 0.05, R2 = 0.77, Figure Figure5c).5c). Dying trees exhibited no relationship between leaf δ13C and leaf Nmass (Figure (Figure55d).
Though we cannot conclusively answer the pressing ecological question of why some trees die while others remain healthy during forest disturbance events—an especially confounding question when trees are conspecifics with similar access to resources—these data help to illuminate divergent responses of conspecifics to some disturbances. Our data suggest that divergent tree responses to this insect-associated disturbance event may have been linked to antecedent tree water relations and N status. Relationships between tree-ring α-cellulose δ13C and δ18O, which integrate across the trees’ lives, suggest that co-occurring healthy and dying trees likely experienced different relative dominance of the drivers of photosynthesis and δ13C values for decades prior to the disturbance. This idea is consistent with the suggestion that lifelong growth strategies can contribute to tree survival or mortality during disturbance events (Levanič et al. 2011, Billings et al. 2015). Differences between healthy and dying trees’ tree-ring N and divergent relationships between contemporary leaf δ13C and Narea in healthy and dying trees also are consistent with this hypothesis.
Basal area increment in healthy and dying trees diverged following a severe drought that began in 1979–80 (Figure (Figure1).1). These data are congruent with those reported in Haavik et al. (2008), a study taking place at the same sites but examining a smaller number of nearby trees, and indicate that drought disturbance decades prior to insect outbreak may have influenced tree susceptibility to mortality. Decreased growth has been associated with oak decline events in the USA and Europe (Pedersen 1998, Drobyshev et al. 2007, Levanič et al. 2011) and has been linked to drought (Klos et al. 2009). Though we observed decreased bole growth later in life for dying trees, DBH was similar among health classes, a counterintuitive phenomenon highlighting how tree-ring measurements can provide far greater insight into comparative tree growth dynamics than coarser, diameter-based measurements.
With the exception of growing season precipitation, healthy and dying trees exhibited no clear distinction in their responses to the quantified climate parameters. Relationships between growing season precipitation and RWI (Table (Table1)1) suggest that the growth of trees that remained healthy during the late 1990s dieback were able to respond positively to moisture inputs, while dying trees were not. These data highlight how moisture availability can be a particularly important driver of growth patterns in these forests, similar to other studies of oak-dominated forests (Pan et al. 1997, Friedrichs et al. 2009, Hilasvuori and Berninger 2010), but even these significant relationships between precipitation and RWI were fairly weak, explaining only one-quarter (approximately) of the variation in RWI. However, bole growth patterns in isolation from other data sets do little to help us understand divergent responses of healthy vs dying trees to precipitation, or contrasting susceptibilities of co-occurring, co-dominant oaks to this disturbance in these forests. Stable isotopes can help shed light on tree responses to the environment that are sometimes difficult to discern using growth patterns (Cernusak and English 2015). As such, to infer potential drivers of tree fate during the late 1990s dieback event, we turn to the stable isotopic signatures of tree-ring α-cellulose and leaf biomass.
If antecedent drought was an important factor driving tree mortality during the insect-related disturbance, as suggested by growth patterns diverging after 1979, we might expect susceptible trees to exhibit elevated δ13C values relative to their healthy neighbors prior to their obvious crown dieback and eventual mortality. Stable isotope theory predicts that tree-ring α-cellulose can become 13C-enriched with moisture stress (Farquhar and Lloyd 1993). However, sampling additional trees for this study did not alter the earlier conclusions of Haavik et al. (2008), who observed no variation in tree-ring α-cellulose δ13C at these sites with health status. Further, similar to Haavik et al. (2008), we observed little variation in δ13C with climate parameters typically associated with vegetation δ13C signatures (Figure (Figure2).2). It may be that regional precipitation records do not provide a sufficiently detailed picture of tree water availability, and that tree δ13C did, in fact, reflect each individual tree’s water status. However, a lack of relationship between biomass δ13C and water status is sometimes reported in other systems (Ramesh et al. 1986, Cullen and Grierson 2007), highlighting how drivers other than moisture availability can govern plant δ13C signatures and the complexities of interpreting them (Roden and Farquhar 2012).
Incorporating δ18O of tree-ring α-cellulose into our analyses can provide clues about the mechanisms driving the decades-long, markedly lower growth trajectory of trees destined to die during this oak decline event, and about these trees’ growth strategies. In spite of the lack of correspondence of in tree-ring α-cellulose δ13C with wet and dry conditions (Figure (Figure2),2), we observed a positive relationship between tree-ring α-cellulose δ13C and δ18O in dying trees (Figure (Figure5b).5b). No such relationship was evident in apparently healthy trees (Figure (Figure5a).5a). If δ13C in these trees reflects stomatal behavior—thus implying that the regional precipitation records were not a good measure of individual trees’ water status—it suggests a greater degree of stomatal control in dying compared with healthy trees (Voltas et al. 2013). As discussed in multiple works (e.g., Barbour et al. 2002), this in turn would suggest that C source–sink dynamics in trees that remained healthy during the insect outbreak were governed less by water status than by other factors. Given that photosynthetic capacity can be an important driver of vegetation δ13C signatures (Farquhar et al. 1989, Barbour and Farquhar 2000), it seems plausible that photosynthetic capacity may have been a dominant driver of C source-sink dynamics in these healthy trees. In contrast, the significant, positive relationship between tree-ring δ13C and δ18O in dying trees suggests that C source–sink dynamics in these trees were governed to a greater extent by stomatal responses to evaporative demand.
The lack of data describing historic atmospheric water demand at these sites and gs for these trees prohibits us from directly linking these variables to tree-ring δ18O. We also cannot know the extent to which gs influenced the slope of the positive relationship between δ13C and δ18O in dying trees. This is an important caveat, because when gs exhibits a limited range of values, the influence of atmospheric water demand on the slope of the relationship between δ13C and δ18O can be great (Barbour et al. 2002). However, these data are consistent with another study comparing co-occurring declining vs healthy trees’ tree-ring δ13C and δ18O: Herrero et al. (2013) report significant, positive relationships for these parameters in trees dying during drought, and less significant trends for trees that remained healthy. Our work expands this kind of analysis in two ways, by exploring a different disturbance type—insect outbreaks—and by using tree N status to help explain drivers of biomass δ13C in trees exhibiting no relationship between δ13C and δ18O.
Given the importance of N as a driver of photosynthetic capacity (Evans 1989) and, accordingly, of growth (Raison et al. 1990) and vegetation δ13C (Korol et al. 1999, Balster et al. 2009, Brooks and Mitchell 2011, Guerrieri et al. 2011), integrating N dynamics into these analyses may help us understand these trees’ responses to their environment and susceptibilities to this disturbance event. Tree N status also may help us understand why tree-ring δ13C in these forests did not vary with moisture availability during drought years. Tree-ring N data can be confounded by mobility of nitrogenous compounds across tree rings after wood formation (Poulson et al. 1995), but multiple studies have successfully used tree-ring N data to observe changes across time in tree N status, particularly when environmental perturbations are sufficiently great to generate a strong signal (Saurer et al. 2004, McLauchlan et al. 2007, Balster et al. 2009, Savard et al. 2009, Guerrieri et al. 2011).
The enhanced tree-ring N of dying relative to healthy trees provides a clue about these trees’ response to their environments, and their eventual susceptibility to mortality during the insect outbreak. Given the co-occurring, interspersed nature of healthy and dying individuals and their co-dominant canopy status at each site, it is unlikely that these two populations experienced differences in their environment that drove differences in wood N concentration. Further, tree-ring δ15N data (not shown) give no indication that healthy and dying trees obtained N from different sources. We infer that trees dying in the early 2000s either allocated more N to their bole wood or were able to access enhanced amounts of plant-available N from the soil. Given that allocation of limiting resources to bole wood is a relatively low priority for trees, particularly those under stress (Waring and Schlesinger 1985), these data suggest that dying trees were less limited in N than healthier ones for decades prior to the insect outbreak, for reasons that remain unclear.
It is unclear if tree-ring N does, in fact, reflect different N limitations of these trees, but tree-ring N has been linked to N availability in other studies (McLauchlan et al. 2007). If tree-ring [N] data are reflective of different N status in dying vs healthy trees, we might expect to find other evidence consistent with lower N availability in healthy trees. Our data provide two such pieces of evidence. First, the lack of a relationship between tree-ring α-cellulose δ13C and δ18O in healthy trees (Figure (Figure5a)5a) is suggestive of photosynthetic capacity being a relatively more important driver of those trees’ historic photosynthesis than stomatal responses to water demand (Barbour et al. 2002). We might expect such a feature in trees for which N is more limiting, given that N can be a key limitation on photosynthetic capacity (Ripullone et al. 2003). Second, we observed a significant, positive relationship between leaf δ13C and leaf Nmass in healthy, but not in dying, trees (Figure (Figure5c5c and d). Such a relationship is consistent with ideas derived from isotopic theory: an increase in N can prompt increases in photosynthetic capacity, which, all else being equal, is linked to increases in leaf δ13C (Barbour et al. 2002, Ripullone et al. 2003). We cannot know if a positive relationship between leaf δ13C and leaf [N] in healthy trees existed in past years. We did not observe a relationship between tree-ring α-cellulose δ13C and tree-ring wood [N], but these analyses likely were confounded by increasing [N] in tree rings of all trees over time. The significant relationship observed in contemporary leaves in healthy trees, however, hints that N is more of a factor driving C source–sink dynamics in these trees compared with their dying neighbors. Alternatively, the oak decline event may have altered the physiology of dying trees such that the influence of N on C source–sink dynamics was diminished. Although insect infestation may have masked relationships between leaf δ13C and [N] in dying trees, the observed relationship in healthy tree leaves is consistent with those trees’ lack of significant relationship between tree-ring α-cellulose δ13C and δ18O and their apparent, long-term greater N limitation. Together, these data sets imply that something other than tree water relations—perhaps N and associated photosynthetic capacity—was a more dominant driver of photosynthesis and tree-ring α-cellulose δ13C values in healthy trees.
The mechanisms linking N availability and tree health status are not clear. Nitrogen availability is an important feature of tree resistance to disturbance (Waring and Pitman 1983). However, we observed greater bole [N] in dying trees. The reasons for this observation remain unclear, but tree resource allocation priorities can change with stress (Waring and Pitman 1983, Waring 1987, Sala et al. 2012), and differences in resource allocation patterns between oak health classes have been hypothesized as an influential factor driving susceptibility of oaks to insect-associated disturbance at these sites (Haavik and Stephen 2011). Low C reserves in the presence of excess of N can increase tree susceptibility to disturbance-associated mortality (Matson and Waring 1984). If dying trees’ growth was more limited by C source strength than that of their healthy neighbors, as implied by the positive relationship between tree-ring δ13C and δ18O in dying trees, the combination of low C reserves and relatively greater N availability for these trees may have influenced their fate during this disturbance event (Matson and Waring 1984). Indeed, Billings et al. (2015) suggest that oaks that remained healthy throughout this insect outbreak may have been more responsive to pulses of N availability. Though we remain uncertain of the linkages between tree N and health status in our study, differential C allocation strategies driven by varying degrees of N limitation may have played a role in governing eventual susceptibility to this disturbance.
These data highlight two key features for using a multi-isotope approach to infer historic tree responses to environmental conditions and for understanding tree resistance vs susceptibility to disturbance events. First, tree-ring δ13C values do not always vary as predicted with historic precipitation, a feature that creates challenges for dendroecologists attempting to understand past tree responses to their environment. Regional precipitation records may be too coarse an indicator of individual trees’ past access to water. However, these data also highlight the potential importance of N as a driver of biomass δ13C. Although isotopic theory predicts that both moisture and N can drive biomass δ13C, linkages between δ13C and N availability are discussed relatively infrequently in the literature, and rarely if ever in the context of tree-ring studies. We do not know why tree-ring δ13C did not vary with past precipitation in healthy or dying trees, but these data suggest that in healthy trees, N may have been an important driver of biomass δ13C.
Second, inferences drawn when tree-ring α-cellulose δ13C and δ18O exhibit a positive relationship can be coupled with biomass δ13C and [N] data to infer relative dominances of the drivers of historic and contemporary tree photosynthesis. We can use the inferences derived from such analyses to develop hypotheses describing potential mechanisms governing tree survival during forest disturbances. These forests hint that photosynthesis of dying trees was governed to a greater degree by stomatal regulation, a phenomenon that influences both δ13C and δ18O of biomass, than by photosynthetic capacity, which is governed in large part by leaf N content. This conclusion is tempered by a frequent problem in tree-ring studies: we do not know historic atmospheric water demand, gs or N availability. However, the idea that dying trees’ past photosynthetic fluxes were governed by stomatal regulation more so than by photosynthetic capacity is bolstered by two observations: tree-ring [N] was lower in healthy trees, suggesting that healthy trees were more limited by N, and 77% of the variation in contemporary, healthy tree values of leaf δ13C was explained by leaf [N], while dying trees exhibited no such relationship. These observations suggest that individuals within this tree population possessed distinct strategies for coping with water limitations vs insect outbreaks.
As impacts of varied disturbances on forests increase in the future, predicting forest responses to decline events and developing indices to predict those responses will become increasingly important. Our study highlights how historic and contemporary stable isotope data may be integrated with tree N status to provide us with a means to predict responses among co-occurring trees to insect-related disturbances. Illuminating the degree to which C source vs sink strengths govern tree C dynamics under different environmental conditions, and different disturbances, will depend on experiments in which multiple measures of tree N status and water relations are quantified simultaneously with biomass δ13C and δ18O. Though we still cannot explain the ultimate mechanism governing how trees’ growth strategies may vary in ways influential in governing responses to eventual disturbances, the application of a dual-isotope approach can help us begin to unravel these mysteries.
This work was funded by the USDA Forest Service’s Southern Research Station and Forest Health Protection, the University of Kansas General Research Fund and National Science Foundation grant 0801522.
We thank C-CHANGE IGERT program director Dr Joane Nagel and KPESIL manager Greg Cane at the University of Kansas, along with Anna Tartarko, Nameer Baker, Ryan Rastok and Alison King for laboratory assistance. We thank Dr Melissa Fierke, Dr Laurel Haavik, Vaughn Salisbury, and Larry Galligan for their contributions at the field sites, and Dr Christoph Lehmeier for a critical review of the manuscript.