Hemlock woolly adelgid (Adelges tsugae) infestation affects water and carbon relations of eastern hemlock (Tsuga canadensis) and Carolina hemlock (Tsuga caroliniana)


Author for correspondence:

Jean-Christophe Domec

Tel: +1 919-687-7639

E-mail: jdomec@ncsu.edu


  • Hemlock woolly adelgid (HWA) is an exotic insect pest causing severe decimation of native hemlock trees. Extensive research has been conducted on the ecological impacts of HWA, but the exact physiological mechanisms that cause mortality are not known.
  • Water relations, anatomy and gas exchange measurements were assessed on healthy and infested eastern (Tsuga canadensis) and Carolina (Tsuga caroliniana) hemlock trees. These data were then used in a mechanistic model to test whether the physiological responses to HWA infestation were sufficiently significant to induce changes in whole-plant water use and carbon uptake.
  • The results indicated coordinated responses of functional traits governing water relations in infested relative to healthy trees. In response to HWA, leaf water potential, carbon isotope ratios, plant hydraulic properties and stomatal conductance were affected, inducing a reduction in tree water use by > 40% and gross primary productivity by 25%. Anatomical changes also appeared, including the activation of traumatic cells.
  • HWA infestation had a direct effect on plant water relations. Despite some leaf compensatory mechanisms, such as an increase in leaf hydraulic conductance and nitrogen content, tree water use and carbon assimilation were diminished significantly in infested trees, which could contribute to tree mortality.


Hemlock trees dominate almost 1 million hectares of forest throughout eastern North America from the southern Appalachians to southern Canada and west to the central Lake States. These forests are on the verge of disappearing because of the hemlock woolly adelgid (HWA), Adelges tsugae Annand, a significant ecological pest that was introduced from Japan (c. 1955) to the eastern USA (Kizlinski et al., 2002; Havill et al., 2006). In Asia, HWA does not cause severe destruction of its hosts because of native host resistance (McClure, 1996; Havill et al., 2006). In the eastern USA, HWA harms native hemlocks as it extracts through its stylet the carbohydrates from the xylem ray parenchyma cells at the base of needles from the current year's growth (Young et al., 1995). The tree responds by stopping growth and by dropping needles (McClure, 1991). After the current year's needles have been shed from the tree, HWA moves to the next most recent growth. Since its introduction, HWA has caused the severe decline of eastern hemlock, Tsuga canadensis (L.) Carr., and Carolina hemlock, Tsuga caroliniana Engelm., populations (Evans & Gregoire, 2007; Krapfl et al., 2011). The average rate of HWA spread over the past decade has been c. 25 km per year and, at this time, cannot be managed in forests. HWA-induced mortality can be explosive, exceeding 80% within 4 yr of infestation (Orwig et al., 2002; Small et al., 2005). The exact mechanism of mortality in hemlock trees is not known, but one hypothesis postulates that an excessive amount of photosynthate from the tree is extracted by infestation (Broeckling & Salom, 2003), leading to a more general mechanism applicable to all tree species that involves the cessation of metabolism or failure to defend against further biotic attack because of a negative carbon budget (McDowell, 2011).

An alternative hypothesis, put forth for the first time here, is that the trees may be responding to the feeding of HWA by a reaction similar to that of Fraser fir (Abies fraseri, Pursh.) to infestation by balsam woolly adelgid (BWA; Adelges picea Ratz; Balch et al., 1964). BWA feeds predominantly on the trunk, and infested trees respond rapidly by releasing phenolic compounds that induce the formation of abnormal xylem, which consists of shorter tracheids, thicker cell walls and smaller lumen (Puritch, 1977). Such abnormal wood formation induces a reduction in xylem permeability (Puritch, 1971; Spicer & Gartner, 1998), affecting whole-plant water acquisition and growth (Arthur & Hain, 1986; Hollingsworth et al., 1991). In Fraser fir, trees also succumb to death within 4 yr (Hollingsworth & Hain, 1991). Given the similarity of time infested, mortality and the proportion of the tree infested (throughout the entire needle-bearing fraction), it is hypothesized that eastern and Carolina hemlocks infested with HWA are exhibiting a reaction similar to that experienced by Fraser fir in response to BWA. We therefore also expect that such a response to infestation is likely to induce changes in stem anatomical properties and, in turn, the ability to move water.

At the whole-tree level, the developmental shift caused by a decrease in leaf area to sapwood area ratio following defoliation from HWA is expected to increase the overall tree hydraulic conductance per unit leaf area (Ktree; Pataki et al., 1998). However, although many studies have examined the relationship between Ktree and water loss, Ktree remains a complicated factor because it is dependent on the sapwood to leaf area ratio as well as on the root, stem and leaf conductivities. As HWA only feeds on needles, leaf water transport capacity, quantified in terms of leaf hydraulic conductance (Kleaf), should be primarily impaired and, together with the decline in leaf area, should be expected to act as one of the main components influencing Ktree. Consistent with earlier investigations documenting the coordination of stomatal conductance (gs) with Ktree (Cochard et al., 2002; Meinzer et al., 2004), Kleaf and its relationship to stomatal control have been investigated recently, revealing that, in conifers, gs is very sensitive to Kleaf within and among species (Woodruff et al., 2007; Domec et al., 2009a; Johnson et al., 2009). However, little is known about the suites of functional traits involved in limiting the dynamic coordination between Ktree, Kleaf, gs and leaf water potential (ΨL) in infested trees. A broader understanding of the patterns of stomatal regulation of ΨL and transpiration in HWA-infested trees requires a knowledge of the hydraulic architecture of the branches to which the leaves are attached and of the leaves themselves. In addition to direct hydraulic measurement, the use of stable carbon isotope ratios (δ13C) as an index of gas exchange can be useful to investigate the effect of HWA on growth. Tree ring δ13C is particularly relevant because it provides a whole-crown-weighted record of gas exchange that can be discretized to individual years, allowing analyses of the response of individual trees to known years of insect attack (McDowell et al., 2010).

The objective of this study was to ascertain whether the water relations of eastern and Carolina hemlock infested with HWA are being affected in a manner similar to that of BWA-infested Fraser fir. Our first hypothesis was that, in eastern and Carolina hemlocks, HWA infestation impairs both leaf water transport capacity and xylem hydraulic characteristics through the production of abnormal xylem cells. Our second hypothesis was that these effects reduce the water transport capacity at the whole-tree level and influence ΨL and gs. Our third hypothesis was that, when the physiological measurements from single trees are scaled up, HWA infestation influences whole-stand water use and contributes to a significant reduction in carbon uptake. In order to test these hypotheses, anatomical and physiological measurements related to leaf and stem water transport were measured in healthy and infested hemlocks. These measurements were then used in a process-based model to test whether the physiological responses to HWA infestation were sufficiently significant to induce a change in plant water use and thus plant carbon assimilation at the tree and stand level.

Materials and Methods

Study site and experimental design

The study site was located on the campus of Montreat College in the Southern Appalachian Mountains (35°38′44.62″N, 82°18′14.35″W; elevation, 797 m). The mean annual temperature is 13.3°C and the average annual precipitation is 1218 mm (Black Mountain Digital Media, Black Mountain, NC, USA; http://www.blackmtndigitalmedia.com/weather.html). This site was chosen because of the close proximity of mature healthy and infested trees. The field sample size consisted of 15 trees of roughly the same size (c. 40 yr old; diameter at breast height, 0.51 ± 0.05 m) and randomly distributed within the plot. Nine trees (four healthy and five infested) were eastern hemlock and six trees (three healthy and three infested) were Carolina hemlock. HWA had been present at the site for 4 yr at the start of the study, but some of the trees (healthy) were being treated with foliar- and soil-injected imidacloprid (Bayer Corp., Kansas City, MO, USA) to limit infestation. The healthy trees had minimal to no needle loss as a result of adelgid infestation. All the nontreated trees were, however, heavily infested, and had an estimated 20–60% needle loss in Carolina hemlock and > 60% needle loss in eastern hemlock. Most untreated trees continued to decline over the course of this study, and two of the infested eastern hemlock died before the end of this study. For both species, we visually classified the branch health from 1 to 5, with 1 being completely healthy (no needle loss) and 5 being almost entirely defoliated.

Plant water potential and stomatal conductance

Diurnal changes in leaf water potential (measured every 2 h from 05:30 to 19:30 h) were determined on three separate branches sampled from the upper crown of each tree using a pressure chamber (PMS Instrument, Albany, OR, USA). To encompass a large range of soil moisture, as seen from the decline in predawn water potentials (ΨPD) which fell from −0.3 MPa in spring to −1.2 MPa in summer, measurements were made from early June to late August of 2007 and 2008. In 2008, on the same days on which the leaf water potentials were measured, the stomatal conductance (gs) was measured on five branches per treatment between 11:00 and 13:00 h using a steady-state porometer mounted with a narrow aperture chamber (Model Li-1600-01; Li-Cor Inc., Lincoln, NE, USA). Water potentials and gs were measured on branches classified as 1 for the healthy trees (all branches were classified as 1 on these trees) and 3–4 on infested trees (> 50% needle loss).

Carbon isotope composition

Branch wood tissue samples were collected in August 2007 and January 2008 for carbon isotope analysis of the branch and leaf tissues. On the infested trees, three branch and three leaf tissues were selected from branches classified as 3–4. In August 2007, six branches were sampled from each treatment and, in January 2008, four branches from the infested trees (classified as 3–4) and two branches from the healthy trees were collected. Each branch sample from both collection times was immediately separated into year of growth (current, second year, third year and fourth year). The second year's growth was selected for analysis because it was the most recently completed growth season. The oven-dried samples (80°C) were separated into branch and leaf tissue, and then ground to a fine powder with a mortar and pestle, after being frozen with liquid nitrogen. Wood powder was washed with a mixture of toluene–ethanol to remove extractives, which also removed nonstructural carbohydrates, ensuring that the carbon isotope composition reflected entirely the carbon assimilated in the corresponding year. Plant material was analyzed for C and N concentration and 13C natural abundance (δ13C; Cornell University Stable Isotope Laboratory, Ithaca, NY, USA). The reported δ13C values were based on the PeeDee Belemnite standard (Craig, 1957). A reduction (more negative values) in δ13C is assumed to reflect improved water use efficiency (WUE) over the period of time over which carbon was assimilated (Ehleringer, 1993).

Hydraulic conductivity

For both species, hydraulic measurements were performed on samples cut in January 2008. Samples included three branches from infested trees (50–70% needle loss, scale of 3–4) and three branches from healthy trees (0% needle loss, scale of 1). One of the infested eastern hemlocks used for the January sample died over the winter, so that, during the July sampling, it was replaced with another infested tree.

Samples (whole branches) were immediately frozen and transported to the laboratory at North Carolina State University. Samples were stored in a chest freezer for 2 months before analysis with a hydraulic conductance flow meter (HCFM-XP; Dynamax Inc., Houston, TX, USA) applied in a quasi-steady-state mode, that is, under conditions in which flow and applied pressure were approximately constant with time (Tyree et al., 1993). Preliminary results showed that freezing the samples did not affect the hydraulic conductivity measurements, as also shown previously in two other conifer species (Reid et al., 2003; Rhea, 2010). The HCFM allows the perfusion of water into the base of a branch whilst measuring the flow rate of water into that branch. The whole-branch hydraulic conductance can then be calculated from the applied pressure and flow rate. Analysis began by attaching the whole-branch sample (stem and needles), freshly cut, to the HCFM at approximately the level of 4 yr of growth. This branch length was chosen because these trees did not retain their needles after the third year. This measurement was recorded as the hydraulic conductance of the whole branch including the stem and leaves (kstem-leaf, mmol s−1 MPa−1). A final conductance measurement after removing all the needles (kstem) was used to calculate the leaf hydraulic conductance Kleaf (mmol m−2 s−1 MPa−1) as:

display math(Eqn 1)

(AL, leaf area).

After achieving stable readings with the HCFM, the distal part of the stem was cut off. The remaining 10–12 cm of basal segment of the stem, left connected to the pressure coupling, was perfused again to determine the specific hydraulic conductivity of stem xylem (ks, kg m−1 s−1 MPa−1):

display math(Eqn 2)

(ksegment (kg s−1 MPa−1), hydraulic conductance of the stem (branch xylem); L, length of the segment; AS, cross-sectional sapwood area). Further, the leaf specific conductivity (LSC, kg m−1 s−1 MPa−1), which represents the efficacy of a branch in conducting water on a leaf area basis, was determined as:

display math(Eqn 3)

Leaf area was determined by first scanning a subsample of fresh needles (Epson Perfection V700 PHOTO; Epson America, Inc., Long Beach, CA, USA) and by measuring their leaf area with Image-J software (National Institutes of Health, Bethesda, MD, USA). Then, the scanned needles and the remaining needles of each sample were weighed, and the total leaf area was estimated on the basis of the specific leaf area (SLA, in m2 kg−1) of the scanned needles.

Wood anatomy

Wood anatomy was observed on samples used for hydraulics. Samples were microtomed (Model 57951; Spencer Lens Co., Buffalo, NY, USA) to 30-μm cross-sections and stained in a solution of 0.1% safranin-O. Analysis of the cross-sections was conducted under light microscopy mounted with a digital camera. Samples were photographed and analyzed with Image-Pro software (Media Cybernetics, Inc., Bethesda, MD, USA). Early- and latewood proportions, together with ray counts, were assessed at ×100 magnification on all of the growth rings present in a sample. When false rings within the abnormal wood were present, they were treated as separate rings on which early- and latewood proportions were determined. Tracheid cell wall thickness and lumen diameter were measured using ×200 magnification. For each sample separately, two radial files of tracheids from pith to bark were randomly selected and all the tracheids in each file were measured. One of the two radii measured was located in the area opposite the greatest amount of compression wood, which corresponds to the trunk section that was subjected to mechanical stress and that occurred in both healthy and infested trees (not to be confused with abnormal wood). The other radius measured was perpendicular to the first radius, thus minimizing the amount of compression wood in the measurements. The tracheid diameter distributions were also estimated for the calculation of the mean hydraulic diameter (Dh = Σd5d4, where d is the individual conduit diameter; Kolb & Sperry, 1999). Inside tracheid diameters were calculated by measuring the lengths of lines drawn across the lumen. Hacke et al. (2001) demonstrated that water-filled conduits experience large tensile hoop stresses that could lead to cell collapse under drought conditions. The basis for this relationship is that the double cell wall shared by adjacent cells behaves in a manner similar to a long plate of width b (cell diameter) and thickness t (taken as the distance across the double cell wall between two adjoining tracheid lumen), and this plate will buckle under a force proportional to (t/b). Therefore, after the determination of the mean hydraulic diameter for a section, we made measurements of (t/b) and expressed it as (t/b)2, so that the results were more directly comparable with previously published data, and could be used as an anatomical indicator of xylem water stress resistance (Hacke et al., 2001; Domec et al., 2009b). About 100–140 tracheids were used to measure t and b.


To predict the effect of HWA infestation on plant water and carbon exchange, we used the soil–plant–atmosphere (SPA) model which is designed to represent processes common to vascular plants (Williams et al., 1996, 2001; Hill et al., 2008). The scale of parameterization (leaf level) and prediction (canopy level) of the SPA model have been designed to allow the model to provide a tool for the scaling up of leaf-level processes to canopy and landscape scales. SPA is a process-based model that simulates ecosystem photosynthesis and tree water use at fine temporal and spatial scales (30-min time step, multiple canopy and soil layers). In the model, the maximum flux rate of water through vegetation is determined by the difference between the soil and leaf water potential, and is controlled by the tree hydraulic conductance (Ktree). Leaf to air energy, water and CO2 exchange consists of a coupled photosynthesis model (Farquhar model) with a stomatal conductance model. Allometric (leaf area index, tree size), hydraulic (LSC, Ktree, minimum leaf potential) and leaf nitrogen content used to drive the model were measured directly and are presented throughout the results (Supporting information Table S1). We estimated Ktree from the relationship between the rates of single-leaf transpiration (E, mmol m−2 s−1) and soil–leaf water potential difference (Loustau et al., 1998). The soil water potential (Ψsoil) was considered to be very close to ΨPD in the leaves (Breda et al., 2006), and thus Ktree was taken as:

display math(Eqn 4)

Using the ohm (electrical resistance) analogy applied to a hydraulic circuit, and because Ktree is equivalent to Kroot-leaf, the root to stem hydraulic conductance component of the model (Kroot-stem) was calculated as: Kroot-stem = (Ktree × Kstem-leaf)/(Kstem-leaf − Ktree), where Kstem-leaf equals kstem-leaf/AL. The model also requires the root total biomass, which can be calculated from the ratio of total root biomass to the projected leaf area (Table S1). Modeled E saturates when root biomass is high, as Ktree becomes more limiting than the soil to root conductance. For both species, a ratio of total root biomass to projected leaf area of 5.1 was chosen because it was sufficient to saturate E, and to come within 7–10% of measured E under high ΨPD (late spring transpiration data).

Leaf photosynthetic parameters, such as the carboxylation rate of RuBisCO (Vcmax, mmol m−2 s−1) and photosynthetic electron transport (Jmax, mmol m−2 s−1), are known to be tightly related to leaf nitrogen content (Field & Mooney, 1986; Dang et al., 1997; Reich et al., 1997a,b). For both species, Vcmax values were determined using the equation introduced by Niinemets & Tenhunen (1997) describing the theoretical anatomical and biochemical interactions that may be sources of variation in Vcmax:

display math(Eqn 5)

(6.25, ratio of the weight of RuBisCO to the weight of nitrogen in RuBisCO; Vcr, specific activity of RuBisCO, which is assumed to be only a function of temperature and is considered to be constant for C3 plants (math formula at 25°C; Harley et al., 1992; Niinemets & Tenhunen, 1998); Nm (g g−1), mass of nitrogen in the leaf per total dry mass of leaf; FR, proportion of leaf nitrogen that is invested in RuBisCO, which has been shown not to vary between species and was taken as 0.15 (Niinemets & Tenhunen, 1998)). Similarly, Jmax may be expressed as:

display math(Eqn 6)

(8.06, converts an investment of 1 g of nitrogen in bioenergetics into mmoles of cytochrome f; Jmc, capacity of electron transport per unit of cytochrome f, again considered to be constant among C3 species (156 μmol electrons μmol Cyt_f−1 s−1 at 25°C; Nolan & Smillie, 1976); FB (taken as 0.12; Niinemets & Tenhunen, 1997), fraction of leaf nitrogen in bioenergetics, reflecting the changes in the contents of the primary proteins determining the overall rate of electron transport (Evans & Seemann, 1989)).

We forced the SPA model with soil, meteorological and radiation data collected in 2005 by Black Mountain Digital Media (station Black Mountain 0.1 WSW, NC-BC 31) and available through the CRONOS (Climate Retrieval and Observations Network of the Southeast) database at http://nc-climate.ncsu.edu/cronos. This year was chosen as it represented, for this region, an average year in terms of mean annual temperature (13.5°C) and precipitation (1150 mm), and also allowed us to compare SPA outputs with direct measurements of tree water use (Daley et al., 2007; Ford & Vose, 2007) and gross primary productivity (GPP) (ecoregion 66 in Coops et al., 2009). WUE was calculated for each species for winter and summer sampling intervals by dividing GPP by tree water use (T). These WUE values provide an independent assessment of model validation, as δ13C data were not used to calibrate the model.

Statistical analysis

The mixed model procedure (PROC MIXED) in SAS version 9.1 was used for all statistical analyses (SAS, Cary, NC, USA). Species, time and their interaction were employed as sources of variation for data from healthy and infested samples, with species as a fixed effect. A separate ANOVA was executed for vigor class (healthy vs infested), time and their interaction for data from each species. Conductivity and stomatal conductance data were analyzed with the fixed effects of species, vigor class and their interaction as sources of variation. Both sets of carbon isotope data were also analyzed with the fixed effects of vigor class and their interaction as sources of variation.


Water potential

In Carolina hemlock, June predawn water potential (ΨPD) was −0.30 MPa lower in infested trees than in healthy trees. In eastern hemlock, June ΨPD never dropped below −0.35 MPa and was no different (= 0.33) between healthy and infested trees (Fig. 1a). In Carolina hemlock, diurnal patterns of midday leaf water potential (ΨMD) in June indicated that the healthy trees reached a more negative water potential earlier during the course of the day than did the infested trees (Fig. 1a), and were able to recover to predawn levels earlier than the infested trees. In June, the ΨMD value of infested trees of eastern hemlock was −0.45 MPa less negative than that of healthy trees (= 0.03). The diurnal change in leaf water status in August followed the same pattern as in June, except that ΨPD and early-morning water potentials (until 09:00 h) were more negative. In August, ΨPD of infested trees dropped below −1.0 MPa, and values of infested trees were 0.61 and 0.70 MPa lower than the values of healthy trees for eastern and Carolina hemlock, respectively (< 0.02; Fig. 1b). In both June and August, ΨPD values of infested Carolina hemlock were  0.30 MPa lower (< 0.001) than that of infested eastern hemlock. In August, there was no difference in ΨMD between healthy and infested trees in either species (P = 0.17 and = 0.33 for eastern and Carolina hemlock, respectively). Across the growing season, the water potential gradients (ΔΨ) from soil to shoot were 43% (= 0.01) and 49% (= 0.01) lower in infested eastern hemlock and Carolina hemlock trees, respectively.

Figure 1.

Diurnal variation in leaf water potentials measured in June during the peak of the growing season (a), and predawn and midday leaf water potentials and leaf carbon isotope discrimination (δ13C) measured in August during the most drought-prone period (b) in healthy and infested eastern and Carolina hemlock trees by the hemlock woolly adelgid. Error bars represent standard errors (SE).

Carbon isotopic composition, nitrogen content and SLA

For both species, the foliar carbon isotope ratios (δ13C) from summer sampling (August) were significantly less negative for infested than for healthy trees (Fig. 1b; Table 1), indicative of an increase in WUE. For winter sampling (January), there were significant differences in δ13C between healthy and infested trees for Carolina hemlock only (Table 1). In addition, isotopic fractionation resulted in some quantitative differences in δ13C between organs, with the wood signal being less negative than leaves. However, between healthy and infested trees and between species, the δ13C data of leaves were fully consistent with the δ13C data of wood (Table 1).

Table 1. Effect of the hemlock woolly adelgid on stem (wood) and leaf carbon isotope ratios (δ13C) in eastern hemlock (Tsuga canadensis) and Carolina hemlock (Tsuga caroliniana) trees
  1. Means with different letters are significantly different at the  0.05 level.

Eastern hemlock
Healthy −28.86 ± 0.13a−28.64 ± 0.83a−29.97 ± 0.16A−29.64 ± 0.45A
Infested−26.90 ± 0.34b−26.71 ± 0.30a−28.22 ± 0.18B−27.34 ± 0.24B
Carolina hemlock
Healthy −28.19 ± 0.13a−26.19 ± 0.08a−28.76 ± 0.08A−27.37 ± 0.10A
Infested−27.13 ± 0.29b−25.74 ± 0.03b−27.95 ± 0.24B−27.75 ± 0.36A

SLA was different between healthy and infested eastern hemlock trees (= 0.01, Table 2). Leaf nitrogen per gram of dry matter and per unit leaf area increased in infested eastern hemlock trees (Table 2), indicating that, in this species, the photosynthetic capacity may have increased with HWA attacks. Both infested and healthy branches appeared to share a common relationship between leaf nitrogen and δ13C (Fig. 2). Foliar δ13C values, an integrated measure of intrinsic WUE, became less negative with increasing N content (Fig. 2).

Table 2. Impacts of hemlock woolly adelgid (HWA) on branch morphological and hydraulic characteristics of eastern hemlock (Tsuga canadensis) and Carolina hemlock (Tsuga caroliniana) trees
 Eastern hemlockCarolina hemlock
  1. N, leaf nitrogen content as a percentage of leaf dry mass; SLA, specific leaf area; AL : AS, leaf area to sapwood area ratio; ks, xylem hydraulic specific conductivity; Kleaf and Ktree, leaf and whole-tree hydraulic conductance, respectively; LSC, leaf specific conductivity; gs, stomatal conductance. One-sided (projected) leaf area values were used to calculate SLA, AL : AS, Kleaf, LSC, Ktree and gs. Also given are the hydraulic resistance located in the leaf and in the stem (branch xylem) as a percentage of whole-tree hydraulic resistance (1/Ktree). Means in the same row with different letters are significantly different at the P ≤ 0.05 level.

N (%)1.16 ± 0.14a1.42 ± 0.07b0.90 ± 0.11a0.81 ± 0.07a
SLA (m2 kg−1)3.28 ± 0.13a3.40 ± 0.07b2.54 ± 0.14a2.50 ± 0.09a
AL : AS (m2 cm−2)0.26 ± 0.10a0.11 ± 0.02b0.31 ± 0.08a0.15 ± 0.04b
ks (kg m−1 s−1 MPa−1)0.72 ± 0.07a0.36 ± 0.16bc0.58 ± 0.03b0.45 ± 0.01ac
Kleaf (mmol m−2 s−1 MPa−1)10.8 ± 4.7a19.9 ± 3.9a5.2 ± 1.4a13.0 ± 2.8b
LSC (10−3 kg−1 m−1 s−1 MPa−1)0.21 ± 0.08ab0.31 ± 0.09ab0.11 ± 0.01b0.16 ± 0.04a
Ktree (mmol m−2 s−1 MPa−1)3.3 ± 0.5a3.5 ± 0.2a1.7 ± 0.2b2.1 ± 0.2c
Leaf hydraulic resistance (%)31 ± 2a18 ± 3b33 ± 2b16 ± 2c
Stem hydraulic resistance (%)14 ± 2a17 ± 1b19 ± 2b15 ± 3ab
gs (mmol m−2 s−1)297 ± 30a179 ± 22b126 ± 9c72 ± 11d
Figure 2.

Linear relationships between leaf nitrogen content (N) and leaf 13C natural abundance (δ13C) in healthy and infested eastern hemlock (EH) and Carolina hemlock (CH) trees (r2 = 0.66 and 0.57 in EH and CH, respectively). Slopes of both linear regression trendlines (< 0.05) varied between 2.45 and 2.54, and were not significantly different from each other (= 0.29). Nonlinear models were also tested, but did not lead to significant regression coefficients. See Table 2 for treatment-assigned differences within each species.

Hydraulic parameters and stomatal conductance

For a given diameter and sapwood area (AS), branches from infested trees had half as much leaf area (AL) as branches from healthy trees (Table 2). Infestation of branch segments with HWA reduced significantly the xylem hydraulic conductivity (ks) by 50% (= 0.02) and 28% (= 0.04) in eastern and Carolina hemlock, respectively. The sharp decrease in AL : AS in infested trees was offset by the decrease in ks of branch xylem, and thus LSC of infested trees increased by 45–48% (< 0.03). In Carolina hemlock, the increase in tree hydraulic conductance (Ktree) was the consequence of the large increase (= 0.05) in leaf hydraulic conductance (Kleaf; Table 2). Although Kleaf was only significantly different between healthy and infested Carolina hemlock trees, in both species, it accounted for almost one-third of the whole-tree hydraulic resistance (1/Ktree). The proportion of tree hydraulic resistance located in the leaves decreased significantly (< 0.04) in both species from > 30% in healthy trees to < 20% in infested trees (Table 2). Because the hydraulic resistance located in the stem represented 14–19% of 1/Ktree (Table 2), the root to stem component, which corresponded to the tree hydraulic resistance minus the resistances in leaf and stem, comprised 48–55% of 1/Ktree in the healthy trees and 66–70% in the infested trees (Table S1).

Differences in stomatal conductance (gs) were significant for species and vigor classes (Table 2). Eastern hemlock trees had higher gs (= 0.01) than Carolina hemlock trees. In both species, HWA infestation decreased gs by 40% (< 0.01 for both species).

Wood anatomy

Anatomical analyses of samples taken from healthy branches of both species showed normal hemlock characteristics, such as distinct rings with abrupt changes in earlywood to latewood structure, with no constitutive xylem ducts and no axial parenchyma cells (Fig. 3a,c). Anatomical results of infested branches were consistent with the hypothesis of abnormal wood formation. The infested branches had a large number of false rings and a smaller percentage of earlywood than did the healthy branches (Table 3; Fig. 3b,d), making it appear that the trees repeatedly slowed down growth sufficiently during the growing season to form some latewood and marginal parenchyma cells (Fig. 3d,e; Table 3). Although the overall mean tracheid diameter was affected by HWA in Carolina hemlock, the lumen tracheid hydraulic diameter decreased significantly in both species (< 0.02). Longitudinal analysis of the infested branches showed that, in both species, there was a decreasing trend in the number of longitudinal ray parenchyma (Table 3). However, this was accompanied by a distortion of the grain with the appearance of large abnormal, unidentified parenchyma cells, which, in eastern hemlock, surrounded abnormal (traumatic) resin canals (Fig. 3e). These cells contained reddish-brown starch deposits that completely occluded the lumina of axial parenchyma. Although wood anatomy revealed a decrease in mean hydraulic lumen diameter in infested branches (Table 3), there was no difference in (t/b)2 (> 0.42), an indicator of drought stress resistance.

Table 3. Effect of hemlock woolly adelgid (HWA) infestation on branch xylem anatomical characteristics of eastern hemlock and Carolina hemlock trees
 Ring width (cm)Total number of ringsRay numberFalse rings (%)Earlywood (%)b (μm)Dh (μm)(t/b)2
  1. b, mean tracheid diameter; Dh, mean lumen hydraulic diameter. Total number of rings includes false rings, and ray number corresponds to the total number of radial ray parenchyma per cross-section. The cell wall to tracheid diameter ratio (t/b)2 is an indicator of xylem water stress resistance. Means in the same column with different letters are significantly different at the P ≤ 0.05 level.

Eastern hemlock
Healthy0.22 ± 0.03a3.5 ± 0.6a68.2 ± 3.5a0a75 ± 2a20 ± 2a14 ± 1a0.27 ± 0.01a
Infested0.15 ± 0.02b7.2 ± 1.0b50.5 ± 3.3b63 ± 9b66 ± 5b21 ± 1a12 ± 1b0.25 ± 0.02a
Carolina hemlock
Healthy0.23 ± 0.02a3.7 ± 1.1a51.2 ± 3.3b12 ± 8a78 ± 1a23 ± 2a19 ± 1c0.23 ± 0.01a
Infested0.14 ± 0.03b6.4 ± 0.9b47.7 ± 3.3b75 ± 11b60 ± 5b16 ± 1b13 ± 1ab0.26 ± 0.03a
Figure 3.

Light microscope images showing the development of abnormal wood production in branch xylem of eastern hemlock infested with hemlock woolly adelgid (HWA). (a, b) Healthy eastern hemlock showing normal hemlock characteristics, such as distinct rings with abrupt changes in earlywood to latewood structure and very rare longitudinal ray parenchyma. (c, d) Infested eastern hemlock branches showing abnormal wood production characterized by the development of multiple false rings (F) and an increase in parenchyma cells (P). Microphotograph (e) represents an infested eastern hemlock branch section showing the presence of traumatic resin canals and P cells that are specialized for the synthesis and storage of phenolic compounds.

Modeling of tree water use

When scaled to the stand level, measured changes in tree physiology, combined with climate drivers, had strong effects on tree water use (T), GPP and WUE. Our model predicted that a pure stand of eastern hemlock and Carolina hemlock would transpire 608 and 401 mm yr−1, respectively, with 45% being transpired in the winter and spring (Fig. 4a,b). HWA infestation is predicted to have a stronger effect on Carolina hemlock than on eastern hemlock, with reductions in T of 41% and 34%, respectively. In eastern hemlock stands, the SPA model predicted that HWA infestation would reduce annual stand-level transpiration by 38% in the winter (November–April) and by 45% in the summer (May–October). In Carolina hemlock stands, HWA infestation would reduce annual stand-level T by 60% in winter and by 52% in summer. The yearly GPP rates were 1800 and 1380 gC m−2 yr−1 for healthy and infested eastern hemlock trees, respectively, and 1600 and 835 gC m−2 yr−1 (gC = grams of carbon) for healthy and infested Carolina hemlock trees, respectively. HWA infestation was also predicted to have a stronger effect on GPP in Carolina hemlock, with a reduction of 48%, compared with a reduction of 24% in eastern hemlock (Fig. 4c,d). However, because of the stronger effect of HWA on T than on GPP in eastern hemlock, and the opposite in Carolina hemlock, WUE increased in eastern hemlock and decreased in Carolina hemlock following infestation (Fig. 5).

Figure 4.

Modeled (soil-plant-atmosphere, SPA) tree transpiration (T) (a, b) and gross primary productivity (GPP) (c, d) for both hemlock species (EH, eastern hemlock; CH, Carolina hemlock) under control or infested conditions. In (a), the yearly T rates were 608 and 401 mm yr−1 for control and infested EH trees, respectively, and, in (b), 400 and 265 mm yr−1 for control and infested CH trees, respectively. In (c), the yearly GPP rates were 1800 and 1380 gC yr−1 for control and infested EH trees, respectively, and, in (d), 1600 and 835 gC yr−1 for control and infested CH trees, respectively. Values of whole-stand annual water use efficiency (WUE) are also given.

Figure 5.

The relationship between the average water use efficiency (WUE) of the two species (EH, eastern hemlock; CH, Carolina hemlock) and the mean 13C natural abundance (δ13C) of leaves from winter and growing season periods. The line represents the fitted linear regression line (R2 = 0.57, = 0.011). Error bars represent standard errors (SE).


Consistent with our first hypothesis, HWA infestation impaired xylem hydraulic characteristics as a result of the production of abnormal xylem cells. However, contrary to our second hypothesis, the negative effect of HWA infestation on sapwood hydraulic conductivity (ks) did not impair whole-tree hydraulic conductance (Ktree), and thus cannot be the direct cause of tree mortality. The increase in Ktree was the consequence of the expected decline in leaf to sapwood area ratio (AL : AS) and of the unexpected increase in leaf hydraulic conductance (Kleaf). In terms of resistance to water flow, HWA decreased by half the contribution of the leaves (1/Kleaf) to whole-tree resistance (1/Ktree; Table 2). During the spring season, this increase in Ktree may have rapidly depleted water in the root zone, thereby increasing the resistance to water uptake at the soil–root interface, as seen by the more negative predawn water potentials (ΨPD) and the increase in the hydraulic resistance of the root to stem compartment. This root hydraulic dysfunction probably increased drought stress in infested trees during subsequent reductions in water availability, leading to reduced minimum leaf water potentials (Ψleaf) and stomatal conductance (gs; Fig. S1). The increase in carbon isotope ratios (δ13C) for the infested trees also suggest that the trees were experiencing drought-like water stress (Farquhar et al., 1982; Zhang et al., 1997).

The δ13C tissue values of both species infested by HWA were less negative than those of healthy trees, and so infestation increased WUE, as demonstrated by the larger reduction in whole-tree water use relative to GPP (Fig. 5). In eastern hemlock, increased WUE was also documented by the significant increase in foliar nitrogen accompanying HWA infestation (Table 1; Fig. 2). Because the majority of leaf nitrogen is bound in photosynthetic enzymes, this implies that photosynthetic capacity may also have increased following infestation. The robust correlations between leaf δ13C and foliar nitrogen implied that high leaf nitrogen contents correspond to increased photosynthetic capacities and decreased internal leaf CO2 concentrations.

Although HWA infestation did not decrease Ktree, in accordance with our second hypothesis, ΨL and gs were influenced. In general there is a strong relation between gs and the aboveground efficiency of the plant water transport system (Meinzer, 2002; Domec et al., 2010), and yet high Ktree in the infested trees was not supported by the development of higher gs. By combining Darcy's law and the Penman–Monteith equation, water relations within the tree can be connected to each other according to Ohm's law analog (Hubbard et al., 2001):

display math(Eqn 7)

(ks, xylem specific conductivity; LSC, leaf specific conductivity; ΔΨ/H, water potential gradient; D, time-averaged vapor pressure deficit of air; c, coefficient representing air and water thermodynamic properties). Under the same evaporative demand, (Eqn 7) predicts that, within a species, as Ktree increases with HWA infestation, so should gs, unless ΔΨ compensates (Fig. S1). High Ktree in the infested trees was not supported by the development of higher gs, because ΔΨ declined significantly, largely as a result of the lower ΨPD (Fig. 1). Predawn water potential measurements are assumed to be an indicator of available water in the soil (Ritchie & Hinckley, 1975). The more negative ΨPD of infested trees from this study indicated that infestation must have caused a form of resistance in the roots, because these trees were growing on the same site as the healthy trees. A recent study by Nuckolls et al. (2009) found that the very fine root biomass of eastern hemlock is indeed reduced when infested with HWA. It has been shown that, if roots are compromised at the soil–root interface, soil–root resistance will increase and the ability of the tree to take up water from the soil will be diminished (Barataud et al., 1995; Sperry et al., 1998). During a drought event, infested trees would be further compromised because of this added resistance. This increase in root resistance was also seen in the decrease in the percentage of resistance represented by the leaves of the infested trees (Table 2), and by the concurrent increase in the root to stem hydraulic resistance from c. 50% of tree resistance in the healthy trees to almost 70% in the infested trees. The increase in root to stem resistance was probably also the consequence of embolism-induced loss of root xylem conductivity, because roots have been shown to be very sensitive to declining xylem water potentials (Domec et al., 2009a,b). Although the ΨPD values of infested trees were not sufficiently negative to cause root death, they were sufficiently low to induce large rates of root embolism.

Our modeling results were consistent with our third hypothesis and illustrated a reduction in whole-stand water use and carbon uptake, which could lead to mortality through the failure to defend against further biotic attack because of a negative carbon budget. Although we did not have estimates of stand-level T and GPP specifically for the site studied, direct estimates of transpiration for eastern hemlock trees of similar sizes were comparable with our predictions (600–700 mm yr−1; Vose & Swank, 1994; Daley et al., 2007), as were modeling estimates of GPP for the region (1600–1800 gC m−2 yr−1; Coops et al., 2009). The lower rates of carbon assimilation and growth in the infested trees explained the abnormal wood production, composed of narrow growth rings with smaller hydraulic tracheid diameters and thus with reduced ks. On attack by HWA, other distinct anatomical changes appeared, including the activation of parenchyma cells and the formation of traumatic resin canals in the xylem (Hudgins et al., 2004; Gonda-King et al., 2012). Eastern and Carolina hemlock may be reacting to infestation in a manner similar to the response of Fraser fir to infestation by BWA. Balch et al. (1964) found that BWA induced hormonal changes in the xylem of infested firs by injecting saliva into the tree for feeding. The saliva stimulated the tree to exhibit the characteristics of abnormal wood that resembled compression wood. Perhaps in a similar hormonal reaction to the saliva of HWA, infested eastern and Carolina hemlock trees produce more latewood in the form of false rings, more longitudinal parenchyma and, in eastern hemlock, traumatic resin canals (Fig. S1). Expansion of these parenchyma cells might be a secondary response induced by growth substances liberated from wounded dying cells (Radville et al., 2011), or might result from changes in turgor pressure caused by an accumulation of osmotically active substances (Hudgins et al., 2004). Inducible secondary xylem parenchyma may provide greater resistance capacity as parenchyma cells are involved in the synthesis and storage of phenolics and resin. These parenchyma cells are not part of the normal developmental program of the xylem and resemble the polyphenolic parenchyma cells, which, in some conifers, are seen after massive fungal infestation (Krekling et al., 2000; Krokene et al., 2008). Although we did not closely examine these putative polyphenolic parenchyma cells, we suggest that they are produced in both hemlock species following HWA infestation.

By feeding on starch reserves, HWA creates a sink for photosynthates produced by the leaves (Nuckolls et al., 2009) and, in the case of eastern hemlock, trees compensate by increasing the capacity for photosynthesis, as seen by the increase in leaf nitrogen. However, this compensation was insufficient to limit the decline in carbon uptake (Fig. 4). The direct effect of reduced GPP has a carry-over effect on carbohydrate reserves, which has important implications for the resilience of already infested trees (Sala et al., 2010; McDowell, 2011). Extended periods with limited photosynthesis may cause mortality because, although carbon allocation to wood growth may be reduced without death, carbon allocation to maintain respiration (Ryan et al., 1995; Bosc et al., 2003) must continue at some basal level, or mortality is inevitable. Assuming that the autotrophic respiration (Ra) of the healthy trees represented 47% of GPP (Waring & Running, 1998), and that this fraction can be applied to the infested trees as Ra is mainly a function of living tissue biomass, it can be calculated that respiration would represent 64% and 91% of GPP in infested eastern and Carolina hemlock, respectively. After taking into account the reduction in leaf area in infested trees, and by assuming that leaf maintenance Ra of hemlock trees corresponds to 18% of total Ra (Waring et al., 1998), respiration would still represent a calculated 60% and 86% of GPP in infested eastern and Carolina hemlock, respectively. For eastern hemlock, the exact respiration cost in infested trees may in fact lie between these numbers as the effect of the decline in leaf area on the respiration cost would be offset by the increase in respiration rates per leaf area because of higher leaf nitrogen content (Reich et al., 2008). The higher proportion of respiration cost would induce a sharp decrease in net carbon productivity (GPP – Ra) which would lower carbon reserves, especially in Carolina hemlock trees. As a consequence, carbon allocation to defense compounds against new insect attack would also be reduced, thus further impairing the resistance of infested trees to subsequent attacks (Coley et al., 1985; McDowell et al., 2007). In addition, insufficient carbohydrate reserves will limit the production of new leaves, and therefore will further reduce GPP and the synthesis of defense compounds (Waring & Pitman, 1985). We therefore hypothesize that, once trees become infested, carbon limitation remains a predisposing mechanism of further vulnerability to HWA attack, and would worsen under water stress (Fig. S1).

A recent modeling study estimated that 40 000 km2 of hemlock forest became infested between 1951 and 2008, with an average of 1520 km2 yr−1 for the last 20 yr (Fitzpatrick et al., 2012). At that rate, our model predicts that a reduction of 23% in GPP would represent a loss of 632 500 tC yr−1 (tC = tonnes of carbon). Knowing that the total GPP of eastern forests represents c. 1 giga tC yr−1 (Xiao et al., 2010), HWA infestation would reduce overall carbon uptake by < 1‰. Overall, we conclude that, at the local scale, HWA can have a large effect on water balance (Ford & Vose, 2007; Guswa & Spence, 2012) and GPP, but that, at the regional scale, HWA infestation is unlikely to have a significant impact on water and carbon fluxes, as hemlock forests represent a small fraction of eastern forests.


We would like to acknowledge Dr Brian Joyce and Montreat College, NC, USA for providing the study site, and we thank Professor David D. Ackerly for valuable insights and improvements to the manuscript. Support for the work of J.S.K., J.-C.D. and L.N.R. was provided by Grant no. 08-JV-11330147-038 from the USDA Forest Service (EFETAC), by the NC Agricultural Research Service via the Christmas Tree Genetics Program and by a research donation from Bayer Corporation. Additional support to J.-C.D. was provided by the DOE – BER Terrestrial Ecosystem Sciences program (11-DE-SC-0006700 – ER65189).