Drought predisposes piñon–juniper woodlands to insect attacks and mortality


Author for correspondence:

Monica L. Gaylord

Tel: +1 928 523 3079

Email: Monica.Gaylord@nau.edu


  • To test the hypothesis that drought predisposes trees to insect attacks, we quantified the effects of water availability on insect attacks, tree resistance mechanisms, and mortality of mature piñon pine (Pinus edulis) and one-seed juniper (Juniperus monosperma) using an experimental drought study in New Mexico, USA.
  • The study had four replicated treatments (40 × 40 m plot/replicate): removal of 45% of ambient annual precipitation (H2O−); irrigation to produce 125% of ambient annual precipitation (H2O+); a drought control (C) to quantify the impact of the drought infrastructure; and ambient precipitation (A).
  • Piñon began dying 1 yr after drought initiation, with higher mortality in the H2O− treatment relative to other treatments. Beetles (bark/twig) were present in 92% of dead trees. Resin duct density and area were more strongly affected by treatments and more strongly associated with piñon mortality than direct measurements of resin flow. For juniper, treatments had no effect on insect resistance or attacks, but needle browning was highest in the H2O− treatment.
  • Our results provide strong evidence that ≥ 1 yr of severe drought predisposes piñon to insect attacks and increases mortality, whereas 3 yr of the same drought causes partial canopy loss in juniper.


A positive relationship between drought and insect attacks on trees is widely accepted (e.g. Mattson & Haack, 1987). Much of the evidence for this positive relationship is based on correlative observations of insect activity and tree mortality after drought. Experimental manipulations of water availability to mature trees that include measurements of insect defense mechanisms and tree mortality are rare. Furthermore, drought stress in trees may have a negative or no effect on insect populations and effects often vary by insect feeding guild, damage agent status (i.e. primary or secondary), and intensity and duration of drought (Larsson, 1989; Huberty & Denno, 2004; Jactel et al., 2012). An improved understanding of drought impacts on tree resistance to insect attacks is critical because climate models forecast more frequent and severe droughts in the future across many subtropical regions (Seager et al., 2007), and forests globally appear susceptible to increases in mortality during drought (e.g. Allen et al., 2010). Here, we report the first experimental evidence regarding the role of drought in insect resistance mechanisms, insect attacks, and mortality of mature trees of the two dominant trees of woodlands of the southwestern USA, piñon pine (Pinus edulis) and one-seed juniper (Juniperus monosperma).

Bark beetles (Coleoptera: Curculionidae), which include the most lethal insects to trees in western forests (Raffa et al., 2008), often attack and kill drought-stressed trees (Huberty & Denno, 2004; Jactel et al., 2012). Drought has well-established physiological impacts on trees, including compromised hydraulic function, reduced carbohydrate production, and possible disruption of carbohydrate supply to sinks (McDowell et al., 2008; McDowell & Sevanto, 2010; Sala et al., 2010; Ryan, 2011). Reduced carbon uptake as a result of stomatal closure and diminished transport to sinks could diminish tree bark beetle defense by reducing carbon available for resin production (McDowell et al., 2008; Sala et al., 2010), which is acknowledged as the primary defense against bark beetles (Berryman, 1972; Christiansen et al., 1987; Herms & Mattson, 1992; Strom et al., 2002; Franceschi et al., 2005). Furthermore, the growth-differentiation-balance hypothesis (GDBH) predicts that moderately water-stressed trees may actually be more resistant to bark beetle attacks because moderate water stress constrains growth sinks more than defensive compound sinks, leading to a relative increase in resin production (Loomis, 1932; Lorio, 1986; Herms & Mattson, 1992; Dunn & Lorio, 1993; Stamp, 2003). The framework also predicts that long-term, severe droughts eventually reduce defense because chronic depletion of carbohydrates reduces carbon availability for resin production (McDowell et al., 2011). Similarly, reduced phloem transport during drought could reduce carbon available for resin production (Dunn & Lorio, 1992; Sala et al., 2010). Despite substantial research on carbon allocation within trees (e.g. Litton et al., 2007), the role of drought in resin defense of trees is poorly understood.

Mortality of piñon was unusually high throughout the southwestern USA during a severe drought in 2000–2003 (Breshears et al., 2005; Shaw et al., 2005; Kleinman et al., 2012). In 2002, Arizona and New Mexico had one of the driest and warmest years on record and by 2003 c. 774 711 ha of piñon woodlands had evidence of bark beetle activity (United States Department of Agriculture, aerial surveys (USDA 2002–2010); Kleinman et al., 2012). Regional (Arizona, Utah, Colorado and Nevada) piñon mortality increased at least two-fold during this drought (Shaw et al., 2005; Williams et al., 2012) and certain areas had > 90% mortality (Breshears et al., 2005; Floyd et al., 2009). Mortality was lower for juniper than for piñon during this drought (Shaw et al., 2005; Floyd et al., 2009; Koepke et al., 2010).

Juniper and piñon have been useful model organisms for studies of tree mortality during drought because they often grow together, but their different survival and hydraulic strategies are representative of a broad range of species in semi-arid regions globally (Mueller et al., 2005a; McDowell et al., 2008; Adams et al., 2009; Koepke et al., 2010). Piñon is more isohydric than juniper (McDowell et al., 2008; Plaut et al., 2012) and minimizes xylem cavitation during drought via stomatal closure to reduce water loss. Juniper, by contrast, can withstand negative xylem pressures with little cavitation during drought and continues carbon assimilation (Linton et al., 1998).

Trees can be attacked by many pathogens and guilds of herbivores. The most lethal or abundant herbivores can determine which components of tree defense are most crucial to tree survival. The most lethal insect to piñon in recent years is Ips confusus (LeConte), the piñon ips bark beetle, which generally attacks stressed or recently dead trees (Wood, 1982b; Rogers, 1995; Raffa et al., 2008). In contrast to piñon, little is known about the insect guild of junipers. Insects that can damage junipers include bark beetles and wood borers (Coleoptera: Buprestidae & Cerambycidae) (Furniss & Carolin, 1977; Itami & Craig, 1989), but data are scarce regarding the amount of tree mortality attributable to these agents.

Our study evaluated the impacts of water availability on insect attacks, resin defenses, canopy condition, and mortality of co-occurring piñon and one-seed juniper over 3 yr via a large-scale precipitation manipulation experiment performed in situ on mature trees. Based on differences in mortality during nonexperimental drought (Shaw et al., 2005; Koepke et al., 2010) and physiological differences in carbon uptake and water relations between these species (e.g. McDowell et al., 2008), we hypothesized that 3 yr of experimental drought would decrease resin defense and increase bark beetle attacks and mortality of piñon, but have little impact on juniper. Moreover, we compared treatment effects on the carbon isotope ratio (δ13C) of leaf sugars and bole resin of piñon to assess the role of recent photosynthate in resin synthesis. Finally, we examined the relationship between resin volume and other parameters of defense/vigor in piñon to understand possible tradeoffs.

Materials and Methods

Study site

Our study site was located in a mature piñon–juniper woodland (Pinus edulis Engelm. and Juniperus monosperma (Engelm.) Sarg.) ≈ 100 km south of Albuquerque, New Mexico in the Los Pinos Mountains in the Sevilleta National Wildlife Refuge and the Sevilleta Long-Term Ecological Research Area (34°23′11″N, 106°31′46″W; elevation 1911 m). Mean temperature (20-yr) ranges from a minimum of −3.3°C (December) to maximum 31°C (July). Mean precipitation (20-yr) at the site is 362.7 mm (Sevilleta Long-Term Ecological Monitoring, 2011), over half of which arrives as late-summer rain. Piñon and juniper basal area (at 30 cm stem height) averaged 20.0 m2 ha−1 across the study site, with piñon basal area comprising 2.9 mha−1 of this total (Pangle et al., 2012). Pangle et al. (2012) and Plaut et al. (2012) provide more detailed site information, including a photo of the study site, information on the block design, assignment of treatments to plots, and criteria for determining plot size.

The experimental design consisted of four treatments and three blocks dispersed across ≈ 25 ha. Installation of treatment infrastructure was completed in August 2007 and consisted of a water removal treatment (H2O−), a water addition treatment (H2O+), an ambient treatment (A), and an additional control (C), which was designed to assess effects of the plexiglass covers used in the H2O− plots. Replicates of each treatment were assigned to 40 × 40 m plots that contained several mature piñons and junipers. In the H2O− and the C treatments, 29 parallel plexiglass gutters were attached to metal poles at ≈ 1 m above ground level and covered ≈ 45% of ground area of plots. In the H2O− plots these gutters intercepted precipitation and transported it off-plot to collection troughs. In the C plots, the gutters were inverted and precipitation dripped to the ground. The gutters increased average maximum soil and ground-level air temperature by 1–4°C during the growing season (Pangle et al., 2012). Water was applied to the H2O+ plots from May to October using 16 equally spaced sprinkler nozzles placed 2–3 m higher than mean tree height. Precipitation amounts in the H2O+ plots averaged 125% of those in the ambient plots and average tree predawn water potentials (Ψpd) were significantly higher in the H2O+ treatment than in the H2O− treatment for both species in all post-treatment study years (Supporting Information Tables S1, S2). More details on site infrastructure and methods for physiological measurements are available in Plaut et al. (2012) and Pangle et al. (2012).

At the beginning of the study (2007), five trees of each species were identified per plot as target trees. Target trees were all centrally located on plots and their stem diameters at a height of 30 cm ranged from 9.0 to 40.9 cm for piñon and from 9.0 to 74.8 cm for juniper (Pangle et al., 2012; Plaut et al., 2012). There was no significant difference in tree diameter among treatments (= 0.26 for both piñon and juniper).

We present methods separately for piñon and juniper because different measurements were performed on each species. Specifically, our measurement of insect resistance mechanisms on juniper was limited to bole resin because our preliminary work showed little resin was exuded from clipped twigs of juniper, and juniper increment cores could not be reliably cross-dated for measurement of sapwood resin ducts (Esper, 2000; Braüning, 2001; Sass-Klaassen et al., 2008).

Piñon resistance and insect measurements

Piñon bole resin

We measured piñon resin flow from bole wounds once before the onset of treatments (3 July 2007) and on five subsequent dates (18 August 2008, 24 May 2009, 1 August 2009, 2 June 2010, and 16 August 2010) using methods first described by Lorio (1993). Briefly, we removed most bark from a section of the bole at ≈ 1 m above the ground to create a smooth surface. We used a 1-cm-diameter arch punch to remove a cylinder of the remaining bark and phloem. Funnels were placed under the wound (one per tree), and resin was collected in 15-ml tubes for 24 h and volume recorded to the nearest 0.5 ml. For the 2007, 2008 and May 2009 measurements, we collected resin from a subset of the previously described target trees (two or three trees/plot; = 8–9 trees/treatment). By August 2009, piñon mortality in the H2O− treatment reduced sample size for this treatment to between three and five trees total.

Piñon twig resin

We collected twig resin on the same piñons used for bole resin sampling between 12:00 and 16:00 h once before the onset of treatments (3 July 2007) and on seven subsequent dates (6 September 2007, 19 June 2008, 18 August 2008, 24–26 May 2009, 31 July 2009 to 1 August 2009, 1–3 June 2010, and 16–17 August 2010). When, because of time constraints, twig resin collection was spread over several consecutive days, we structured our sampling so that complete treatment blocks would be sampled on the same day. We collected twig resin using methods described by Mopper et al. (1991). Briefly, we clipped one terminal twig segment from two, opposite aspects of each tree directly behind the previous year's needles (May, June and early July sample dates) or current-year growth (late July, August and September sample dates). Aspect varied by growth form of the tree. After 10 min we used a toothpick to scrape the resin bubble that formed on the proximal side of the cut branch into a pre-weighed microcentrifuge tube. We recorded resin flow as the difference between the initial mass (g) of the vial and toothpick and the mass after collection.

Piñon resin ducts

We used an increment borer (5 mm diameter) to sample sapwood from the target trees initially identified in 2007 for resin collection (= 8–9 per treatment). One core was extracted from the lower bole (0.5 m above ground) of each tree in August 2010. Xylem resin duct characteristics have been associated with pine survival during drought and bark beetle outbreaks (Kane & Kolb, 2010; Kläy, 2011). We measured resin ducts using methods modified from Kane & Kolb (2010). Briefly, cores were allowed to air-dry for 1 wk before mounting on wood blocks. We then sanded the cores using progressively higher grit sand paper (120–400). We scanned each core to create a high-resolution image (1200 dpi), which was analyzed with WinDendro software (Regent Instruments Inc, 2009) to assign annual ring boundaries and measure ring widths. We assigned calendar years to the last 10 yr of growth before the onset of treatments (1996–2007) via identification of narrow and wide marker years from a previously constructed master chronology for the site. Visual cross-dating was confirmed using the statistical software cofecha (Holmes, 1983; Grissino-Mayer, 2001). We did not include post-treatment years in our cross-dating correlation because the treatments probably altered the annual precipitation signals on which cross-dating was based. Next, we imported the core images into the public domain image-processing program Image J (Rasband, 1997-2011), and we measured resin duct characteristics for each year from 2003 to 2010, or until year of death, along a 2-mm-wide core section. Resin duct characteristics are described in detail by Kane & Kolb, (2010) and Kläy (2011) and included production (number of ducts), average duct area (mm2), resin duct density (number of ducts per area of ring (mm2)), and per cent area (mm2 of resin ducts mm−2 of growth). For two trees in the H2O− treatments, a missing ring was noted between treatment onset and year of death. Because we were interested in yearly allocation to growth and defense, we assigned values of zero for tree growth and resin duct production and area to the missing rings. We assumed growth declined as water stress increased and assigned the missing ring to the most recent year before death, or the end of our study. Because of low sample numbers in the H2O− treatment in 2009 and 2010 (six and three trees for 2009 and 2010, respectively), a sensitivity analysis to confirm our assumption was not possible.

Piñon insect attacks and tree condition

We surveyed the five target piñons on each plot (= 15 per treatment) once before the onset of treatments (2 July 2007) for canopy condition, mortality, and insect attacks. We surveyed the trees once or twice in each post-treatment year (19–20 June 2008, 23–26 May 2009, 30 July to 2 August 2009, 1–5 June 2010, and 16–18 August 2010). Once a tree was recorded as dead (defined as 100% needle browning), it was dropped from subsequent surveys. Sample sizes for the post-treatment surveys ranged between five and 17 trees per treatment because of tree mortality and the addition of replacement survey trees. For each survey we recorded the presence or absence of new (since previous survey) twig beetle attacks (as evidenced by needle browning, feeding activity, or presence of twig beetles) and bark beetle attacks (presence/absence based on entrance holes, galleries, or presence of beetles), and classified trees as live or dead. Bark beetle attacks were further quantified by counting successful attacks (defined as entrance holes with visible frass) or unsuccessful attacks (resin only) on the bottom ≈ 1.5 m of the main bole. Bark beetles were excavated from galleries of a subset of trees and were identified to species based on morphological characteristics (Chansler, 1964; Wood, 1982b), and comparison with voucher specimens from the Rocky Mountain Research Station, Flagstaff, Arizona. Twig beetles were identified to species and voucher specimens preserved by Jim LaBonte (Oregon Department of Agriculture, Salem, OR, USA). Other insects noted during piñon surveys (damage or actual insect) included the piñon stem-boring moth (Dioryctria albovittella, Lepidoptera: Pyralidae), aphids (Hemiptera: Aphididae), piñon needle scale (Matsucoccus acalyptus Herbert, Hemiptera: Margarodidae), piñon spindle gall midge (Pinyonia edulicola, Diptera: Cecidomyiidae), pitch moth (Synanthedon spp., Lepidoptera: Sesiidae), and piñon sawfly (Hymenoptera: Diprionidae). Data on these species were not analyzed because of low presence and presumed low tree impact.

Piñon resin and leaf carbon isotope ratios

To better understand the source of carbon used for resin production, we compared the δ13C of piñon bole resin and recently assimilated leaf sugars among treatments. Leaf sugar isotopes were analyzed because they reflect carbon isotope discrimination during recent photosynthesis and thus would indicate whether the water treatments had an immediate impact on leaf gas exchange. In arid systems, a higher (less negative) δ13C generally indicates lower stomatal conductance and photosynthesis (Ehleringer, 1991).

Bole resin was collected from piñon as described above for pi\xF1on bole resin collection and frozen at ≈ −2°C in sealed tubes until processed for isotope analysis. We were unable to analyze some samples because of the small volume collected. For 2007, 2008, and May 2009, we measured bole resin δ13C on seven to nine trees per treatment. For the remaining dates (August 2009, 2010, and June 2010), sample sizes were between eight and nine trees for the C, A and H2O+ treatments, but reduced to three to five trees for the H2O− treatment. Between 1.5 and 3.0 mg of resin was weighed into small tin capsules and analyzed for δ13C in a Costech ECS 4010 elemental analyzer (Costech Analytical Technologies, Valencia, CA, USA) coupled with a Thermo Electron Delta Plus Advantage mass spectrometer (Bremen, Germany) operated in continuous flow at the Colorado Plateau Stable Isotope Laboratory, Northern Arizona University, Flagstaff, AZ, USA. Isotope ratios are expressed relative to Vienna Standard Peedee Belemnite. A total of 202 resin samples were run and overall precision for δ13C was 0.070/00 (= 60).

We sampled soluble sugars from new, healthy foliage from five trees per treatment in one sample block (= 20) on the south side of the tree in the afternoon on 13 dates between October 2006 and August 2008. Trees sampled for resin (as described above) were a subset of these same trees within the respective sample block. Samples were collected in the afternoon, 6–8 h after sunrise. To extract carbohydrates from leaf samples, we used procedures described by Brugnoli et al. (1988) and adapted to piñon by West et al. (2007). Briefly, we clipped leaf tissues into coin envelopes which were immediately buried in dry ice (CO2) in a cooler and kept frozen for transportation. In the laboratory these samples were kept frozen at −80°C until analysis. Before analysis, samples were freeze-dried and then 150 mg of sample was ground in a ball mill and placed in a 50-ml centrifuge tube with 35 ml of deionized water. The tube was boiled in a water bath for 30 min, cooled, and then centrifuged for 15 min at 12 100 g. We collected the supernatant and passed it through a C-18 Sep-Pak cartridge (Waters, Milford, MA, USA) to remove large organic molecules. Next, we passed the sample through an ion-exchange column, consisting of Dowex-50 and Dowex-1 (Sigma Aldrich) and then filtered with 45 μM Acrodiscs (Pall Corp., Port Washington, NY, USA). We then evaporated the sample and weighed 2 mg of the remaining sugars into tin capsules for stable isotope analysis. Carbohydrate extracts were analyzed at the Department of Earth and Planetary Sciences, University of New Mexico in a Costech elemental analyzer (EA 1108) coupled with a Delta Plus Mass Spectrometer. Isotope ratios are expressed relative to Vienna Standard Peedee Belemnite with an overall precision for δ13C of 0.030/00 (= 261).

Juniper resistance and insect measurements

Juniper bole resin

For juniper, we measured bole resin produced after wounding using methods similar to those described for piñon, except that resin was collected using pre-weighed toothpicks and vials because of the low flow. Briefly, on each date we wounded a subset of the target trees (two or three junipers per plot; eight to nine trees per treatment; one wound per tree) to the face of the xylem and, after 24 h, scraped resin from the face of the xylem using a pre-weighed toothpick into a pre-weighed microcentrifuge tube. We recorded resin flow as the difference between the initial weight of the vial and toothpick and the weight after collection.

Juniper insect attacks and tree condition

For juniper we recorded needle browning (% of total canopy) at each survey date on the five target trees per plot (= 15 trees per treatment). Juniper mortality was recorded when 100% of needles had browned. Insect attack data on juniper, such as branch flagging from juniper twig pruners and Buprestid and Cerambycid emergence holes, were not analyzed because of the low number of emergence holes and, in the case of twig pruners, because it was difficult to conclusively determine if twig dieback was from twig pruners or water stress in the H2O− treatment.

Statistical analysis

We evaluated treatment impact on juniper and piñon bole resin, piñon twig resin, resin duct parameters, bole resin δ13C, and leaf sugar δ13C using separate repeated measures MANCOVA for each characteristic with individual trees as the experimental unit, date as the repeated factor, and treatment as the independent variable. Blocks were not formally included as a factor for all analyses in order to increase power. Values were rank-transformed for juniper and piñon bole resin, piñon twig resin, and resin duct density to correct for normality of residuals as indicated by the Shapiro–Wilk W test. In the case of juniper bole resin, values were still not normal, and thus P values close to zero should be treated with caution. Pretreatment values were a covariate for each respective variable to account for any pretreatment differences. For piñon, because of tree mortality and the use of replacement trees, the pretreatment value assigned to each tree for twig and bole resin and bole resin δ13C was the respective pretreatment plot average. For piñon leaf sugar δ13C and resin duct parameters, the average pretreatment value (seven measurements between October 2006 and August 2007 for piñon leaf sugar δ13C; years 2003–2007 for resin duct parameters) of each individual tree was the covariate. The covariate for juniper bole resin was the single pretreatment value of each tree. We used Student's t-tests with Bonferroni-adjusted alpha levels for post hoc comparisons of treatment and date means.

We used a chi-squared analysis with a null hypothesis of even distribution among treatments at the end of the study (2010) to analyze the effect of treatment on cumulative twig beetle attacks (presence/absence) and cumulative piñon mortality. We used an ANOVA to analyze the effect of treatment on cumulative piñon bark beetle attacks (number of successful attacks/unit bark area).

We used simple linear regression to evaluate the relationships (1) between piñon tree diameter and bark beetle attacks (density and presence/absence); (2) between piñon twig resin and tree diameter growth; and (3) between piñon bole resin volume and the following variables: twig resin volume, tree diameter, tree diameter growth, and all measured resin duct parameters. We also examined the relationship between piñon bole resin and all predictor variables using multiple regression; however, this did not result in a better model fit. We used ANOVA to analyze the difference in resin volume and resin duct parameters (mean values of 2007 to either 2010 or the last measurement before mortality/attack for both resin volume and resin ducts) between attacked and unattacked piñon trees. We also used ANOVA to analyze the difference in Ψpd between attacked and unattacked piñon trees both before and after attacks. The Ψpd measurements were taken 3 or 4 wk before, and within 2 wk after bark beetle attacks were first recorded on trees.

The effect of treatment on juniper needle browning was analyzed with repeated measures MANOVA with date, treatment and their interaction as independent variables. Cumulative needle browning was calculated using the pretreatment year of 2007 as the base year of 100% live crown and subsequent percentages were subtracted from this initial total. Needle browning was square root-transformed to correct for normality and heteroscedascity as indicated by the Shapiro–Wilk W test. Juniper mortality was not analyzed because only one juniper (H2O− treatment) had died by the end of the study in 2010. All statistics were performed using jmp 9 (SAS Institute Inc, 2010). For ease of interpretation we present untransformed data in figures and tables.


Piñon insect resistance mechanisms

Piñon bole resin flow was not significantly affected by the interaction of treatment and date (Wilks' lambda, = 0.49; Table S3) or treatment (= 0.84). Date had a significant effect (P = 0.0380) (Fig. 1), as did the covariate pretreatment resin flow (= 0.0017). For piñon twig resin flow, the interaction of treatment and date was significant (Wilks' lambda, = 0.0210; Table S3), and thus each date was analyzed separately. Treatments had a significant influence on piñon twig resin on two dates (May 2009, F ratio = 4.56, df = 3, = 0.0096; August 2009, F ratio = 7.64, df = 3, = 0.0007), when trees in the H2O− (May and August) and H2O+ (August) treatments had significantly lower resin flow than trees in the A treatment (Fig. 2). The pretreatment covariate was significant for August 2009 (F ratio = 5.11, df = 1, = 0.0321).

Figure 1.

Mean ± SE of resin volume collected over a 24-h period after phloem wounding of piñon boles at the Sevilleta, NM field site by date and treatment (H2O+, water addition plots; H2O−, water removal plots; C, cover control; A, ambient; = 30–35 on each date). Treatments were initiated in August 2007 as indicated by the arrow on the graph. MANCOVA, with plot-mean pretreatment values as the covariate, showed no significant effect of the treatment and date interaction (Wilks' lambda, = 0.49) or treatment (= 0.84). Date had a significant effect (= 0.0380). Dates labeled with different letters are significantly different based on Student's t-test with Bonferroni-corrected α = 0.005.

Figure 2.

Mean ± SE weight of piñon twig resin flow collected 10 min after twigs were cut at the Sevilleta, NM field site, by date and treatment (H2O+, water addition plots; H2O−, water removal plots; C, cover control; A, ambient; = 30–36 on each date). Treatments were initiated in August 2007 as indicated by the arrow on the graph. MANCOVA, with plot-mean, pretreatment values as the covariate, showed a significant interaction of treatment and date for post-treatment twig resin flow (Wilks' lambda, = 0.0210), and thus each date was analyzed separately. Treatment was significant on two dates (May 2009, = 0.0096; and August 2009, = 0.0007). Treatments within dates not labeled by the same letter are significantly different as determined by Student's t-tests with Bonferroni-corrected α = 0.0083.

Piñon resin duct production, area, density, and per cent area were all significantly affected by the treatment and time interaction (Wilks' lambda,  0.0149; Table S4), and thus each year was analyzed separately using ANOVA. Treatment had no significant effect on any resin duct characteristic in 2008 or 2009 (Table S5; Fig. 3a–d). In 2010, resin duct production, area, density, and per cent area were consistently lowest in the H2O− treatment, and significant differences among treatments occurred for production, density, and per cent area. The pretreatment covariate was significant ( 0.0181) for some years/variables.

Figure 3.

Mean ± SE of piñon resin duct characteristics from increment cores (= 28–35) collected from the lower 0.5 m of the main bole of trees at Sevilleta, NM. Characteristics are: (a) production (number of ducts per 2-mm width of ring), (b) average duct area (mm2), (c) density (number of ducts per area of ring mm−2), and (d) per cent area (mm2 of duct area mm−2 of ring area). Results are presented by year and treatment (H2O+, water addition plots; H2O−, water removal plots; C, cover control; A, ambient). Treatments were initiated in August 2007 as indicated by arrows on graphs. MANCOVA, with individual tree pretreatment values (average of years 2003–2007) as the covariate, indicated that there was a significant treatment and year interaction for each characteristic (Wilks' lambda,  0.0149, for all parameters), and thus each year was analyzed separately. In 2008 and 2009, there was no effect of treatment on resin duct characteristics. In 2010, treatment significantly affected production, density, and per cent area, as indicated by * on graphs, but did not affect duct area. Treatments in 2010 labeled with different letters are significantly different as determined by Student's t-tests with Bonferroni-corrected α = 0.0083.

Piñon insect attacks and tree mortality

By August 2010, treatment significantly (χ2 = 10.37, df = 3, = 0.0157) impacted attacks by twig beetles (Pityophthorus opaculus LeConte) on piñon, with the highest percentage of attacks occurring in the H2O− treatment (Fig. 4a). Bark beetle (Ips confusus (LeConte)) attack densities were significantly higher (F ratio = 3.85, df = 3, = 0.0134) in the H2O− treatment than in the other treatments (Fig. 4b). Treatment significantly affected piñon mortality (χ2 = 28.04, df = 3, < 0.0001), which was greatest in the H2O− treatment (Fig. 4c). At the end of our measurements in 2010, piñon mortality was c. 70% in the H2O− treatment and 10% or less in the other treatments.

Figure 4.

For the post-treatment study period (2008–2010): (a) cumulative per cent of piñons attacked by twig beetles (Pityophthorus opaculus LeConte) by treatment, (b) mean ± SE density of successful bark beetle attacks (successful attacks per m2 of piñon bark) of bark beetles (Ips confusus (LeConte)), and (c) cumulative piñon tree mortality by treatment (= 69 for all graphs). Treatment significantly affected cumulative tree attacks by twig beetles (= 0.0157; chi-squared), density of bark beetle attacks (= 0.0134; ANOVA), and cumulative tree mortality (P < 0.0001; chi-squared). Significant differences in bark beetle attack densities by treatment, based on Bonferroni-corrected α = 0.0083, are indicated by different letters on the graph. Treatments are: H2O+, water addition plots; H2O−, water removal plots; C, cover control; A, ambient.

Piñon stem diameter had no linear relationship with density of bark beetle attacks (= 0.28, R2 = 0.02, = 59) and no logistic relationship with presence/absence of bark beetle attacks (= 0.21, χ2 = 1.57, = 59). There was no linear (= 0.24, R2 = 0.01, = 145) or nonlinear (P ≥ 0.40 for quadratic and cubic) relationship between piñon twig resin and bole resin. Similarly, no other relationship was found between piñon bole resin volume and tree diameter, or resin volume and measured resin duct parameters ( 0.0543 for all; = 171–179) (Fig. S1a,c–f). Bole resin had a weakly significant (= 0.0479, R2 = 0.03, = 180) quadratic relationship with tree diameter growth (Fig. S1b). There was no relationship (linear, quadratic, or cubic;  0.68; = 176) between stem diameter growth and twig resin flow in piñon (Fig. S2).

Average resin duct size during the experiment was significantly (= 0.0022; Table S6) smaller (37%) in piñon trees that died than in those that survived, whereas there was no significant difference in any other resin duct parameter (production, density or per cent area), bole resin, or twig resin between trees that survived and those that died (Table S6). In addition, there was no significant difference in Ψpd either pre- or post-attack between attacked and unattacked trees (Table S7).

Piñon resin and leaf sugar carbon isotope ratios

The δ13C of piñon bole resin averaged −24.0 ± 0.04‰ (SE) over all treatments and dates and was not significantly affected by the interaction of treatment and date (Wilks' lambda, = 0.53), treatment (= 0.94), or date (= 0.97) (Table S3; Fig. 5a). Piñon leaf sugar δ13C averaged −23.6 ± 0.06 ‰ over all treatments and dates and was not significantly affected by the treatment and date interaction (Wilks' lambda, = 0.09) or date (= 0.80) (Table S3). Leaf sugar δ13C in the H2O+ treatment was consistently lower (= 0.0263) than in other treatments throughout the first growing season (2008) after the onset of treatments (Fig. 5b).

Figure 5.

Mean ± SE carbon isotope ratio (δ13C) of (a) piñon bole resin (= 28–32/date) and (b) leaf sugars (= 12–20/date) collected at the Sevilleta, NM field site by date and treatment (H2O+, water addition plots; H2O−, water removal plots; C, cover control; A, ambient). Treatments were initiated in August 2007 as indicated by arrows on graphs. MANCOVA, with pretreatment values (plot average for bole resin δ13C; average of seven measurements between October 2006 and August 2007 for leaf sugar δ13C) as the covariate, indicated no significant interaction of treatment and date (Wilks' lambda, = 0.53), and no significant effect of treatment (= 0.94) or date (= 0.97) on tree bole resin δ13C. For leaf sugar δ13C there was a significant treatment effect on post-treatment δ13C (= 0.0263), but no effect of the treatment and date interaction (Wilks' lambda, = 0.09) or date (= 0.80). Trees in the H2O+ treatment had significantly lower δ13C leaf sugar than the other treatments post-treatment, as determined by Student's t-tests with Bonferroni-corrected α = 0.0083.

Juniper insect resistance mechanisms and canopy condition

Juniper bole resin flow was low for most measurements (< 0.01 g) and was not significantly affected by the treatment and date interaction (Wilks' lambda, = 0.12) or treatment (= 0.84). The covariate pretreatment resin flow (= 0.0003) and date (= 0.0012) had significant effects (Table S3; Fig. S3).

Juniper cumulative needle browning was significantly influenced by the treatment and date interaction (approx. F = 6.37, Numerator df = 9, Denominator df = 131.6, Wilks' lambda, < 0.0001); therefore, each date was analyzed individually using ANOVA (Table S8). Treatment did not significantly affect needle browning in 2008 (= 0.18); however, in 2009 and 2010 junipers in the H2O− treatment had significantly more needle browning (< 0.0001, for both) than trees in other treatments (Fig. 6). By the end of the study in 2010, ≈ 50% of juniper foliage was brown in the H2O− treatment compared with 15–20% in the other treatments.

Figure 6.

Mean ± SE cumulative needle browning on juniper trees at the Sevilleta, NM field site by date and treatment (H2O+, water addition plots; H2O−, water removal plots; C, cover control; A, ambient; = 60/date). Treatments were initiated in August 2007 as indicated by the arrow on the graph. MANOVA showed that there was a significant treatment and date interaction (Wilks' lambda, < 0.0001), and therefore each date was analyzed individually. Treatment was not significant in 2008 (= 0.13), but was significant in 2009 and 2010 (< 0.0001 for both, as indicated by * on graph). Treatments within each date labeled with different letters are significantly different as determined by Student's t-tests with Bonferroni-corrected α = 0.0083.


Our study provides the first experimental evidence that drought predisposes mature piñons to bark beetle attacks and improves understanding of mechanisms of drought impacts on insect resistance of piñons. One or 2 yr of drought with 45% reduction in ambient precipitation increased bark and twig beetle attacks and tree mortality for piñon. In addition, we found that junipers were stressed, but not killed, by 3 yr of drought.

Interestingly, experimental drought had variable impacts on the insect resistance characteristics we measured on piñon. Based on previous reports of reduced photosynthesis of piñon during drought (McDowell et al., 2008; Breshears et al., 2009) and evidence of 7 months of near-zero gas exchange of piñon in the drought treatments at our study site (Plaut et al., 2012), we hypothesized that piñons in the H2O− treatment would be carbon limited and thus have lower resin flows in response to wounding than in other treatments. Support for this hypothesis differed for bole and twig resin. For bole resin, the H2O− treatment had little impact on flow. We should note that, similar to other studies of southwestern pines (Gaylord et al., 2007, 2011), piñon bole resin flow was extremely variable in our study (coefficient of variation = 127%), and thus a larger sample size would be needed to detect treatment effects. In addition, our sampling methodology did not allow us to examine within-tree variation of bole resin flow.

Our finding of no treatment effect on piñon bole resin δ13C in combination with lower δ13C for leaf sugars in the H2O+ treatment suggests a lag between drought impacts on carbon assimilation and resin synthesis, because most bole resin was constitutive and formed from previously assimilated carbon. A larger difference in δ13C between pre- and post-treatment values of leaf sugars (≈ 2.5‰) than resin (≈ 0.6‰) provides additional support for this interpretation. Our interpretation that pine bole resin is largely constitutive and formed from old assimilates has been previously reported for seedlings (Lewinsohn et al., 1991; Guérard et al., 2007), but this is the first report for mature trees.

The results for twig resin, that is, less flow in both the H2O− and the H2O+ treatments than in other treatments on some dates, were more consistent with our hypothesis and with predictions of the GDBH framework of a bell-shaped relationship between carbon-based defense and drought stress (Herms & Mattson, 1992; Stamp, 2003). The bell-shaped relationship is expected because severely water stressed trees, such as in the H2O− treatment, are not able to supply photosynthate for resin production, whereas well-hydrated trees, such as in the H2O+ treatment, preferentially allocate photosynthate to growth rather than defense. This tradeoff results in the moderately water-stressed trees (the A and C treatments in our study, which experience seasonal water stress from the ambient precipitation cycle) having the greatest resin production and intermediate growth. Contrary to our prediction, however, we found no relationship between twig resin volume and tree radial growth (Fig. S2). Interestingly, we found a weakly significant, bell-shaped relationship between bole resin volume and tree radial growth (Fig. S1b). In conclusion, our findings of no relationship between twig resin flow and bole resin flow, and differing treatment effects on resin flow from the bole and twigs, strongly suggest that resin synthesis and flow are compartmentalized within piñon trees, and argue against attempts to infer the defense capability of an entire tree from resin flow measured on a single location or tissue.

The greater bark and twig beetle attacks on piñons in the H2O− treatment than in other treatments may be explained by treatment effects on tree characteristics other than resin flow. For instance, I. confusus, the attacking bark beetle in our study, may select hosts via olfactory or auditory cues from drought-stressed trees (Kimmerer & Kozlowski, 1982; Wood, 1982a; Tyree et al., 1984; Mattson & Haack, 1987; Stumpf & Johnson, 1987; Tadege et al., 1999; Seybold et al., 2006). Furthermore, resin quality (e.g. terpenes and crystallization rate) or production of induced defenses, both of which could be impacted by tree water stress, may be more important than constitutive resin to beetle attack success (Raffa & Berryman, 1982; Lieutier, 2002). The importance of induced defenses for piñon, which could include the production of additional resin after constitutive resin has been drained by initial attacks, is not known, although studies of another southwestern pine (Pinus ponderosa) have shown little induction of resin flow by traumatic bole wounding or inoculation with beetle-vectored fungi (Wallin et al., 2003; Gaylord et al., 2011). Moreover, other studies have reported a positive association between sapwood resin duct abundance and pine survival during drought and bark beetle attacks (Kane & Kolb, 2010; Kläy, 2011; A. K. Macalady, unpublished data). Our study found that trees that died had smaller resin ducts than those that survived, and most duct parameters were reduced by the H2O− treatment by the third year, but not earlier (Fig. 3). In contrast to other studies (Blanche et al., 1992), our study did not find a significant relationship between constitutive resin flow and resin duct parameters (Fig. S1c-f).

We measured piñon twig resin because of previous reports that twig resin is important for tree resistance to stem-boring moths (Whitham & Mopper, 1985; Mopper et al., 1991; Brown et al., 2001; Mueller et al., 2005b). Tree damage from stem moths was minor in our study (ranging from 0 to 5% foliage impacted on a tree for any year); however, twig beetle attacks were common in dying piñons and may have contributed to mortality. To our knowledge, no studies have assessed the effect of twig resin on twig beetle performance or host selection. We found no difference in twig resin flow between trees that were attacked by twig beetles and those that were not (= 0.32). This result suggests little functional importance of twig resin flow to tree resistance to twig beetles, although more research is needed on this topic.

Extremes of water availability in our study had no effect on juniper resin flow, which is consistent with current understanding that juniper maintains gas exchange during drought (e.g. McDowell et al., 2008; Plaut et al., 2012). We also found no evidence of bark beetles in junipers, suggesting that susceptibility to insect attacks was unchanged by water availability. Damage associated with the juniper twig pruner (twig dieback) occurred at the beginning of our study before the onset of treatments, but this damage did not increase during the study and few junipers died over the 3 yr of the study, consistent with reports that this Cerambycid is generally not associated with tree death (Furniss & Carolin, 1977). In addition, there was some evidence of Buprestids on dead junipers, as indicated by ovoid as opposed to circular exit holes; however, no western cedar borers (a Buprestid associated with juniper mortality) were collected in Lindgren funnel traps (Lindgren, 1983), which we deployed at the study site in 2010 for monitoring purposes (= 4, two with I. confusus lure (cis-verbenol, 50/50 ipsdienol and 50/50 ipsenol), and two with a general wood borer lure (ethanol ultra high release (UHR) and alpha pinene gel UHR)) (Contech Enterprise, Victoria, BC, Canada). Although we know little about juniper resistance to insect attacks, our study and others (e.g. Floyd et al., 2009) have concluded that insect damage is not a strong contributor to juniper decline during drought.

Greater needle browning of junipers in the H2O− treatment than in other treatments over 3 yr clearly shows sublethal impacts of drought on juniper. Our finding of drought-induced canopy loss in juniper, but not whole-plant mortality, is consistent with previous nonexperimental field observations in the southwestern USA after severe drought in 2002 (Koepke et al., 2010). While our results show that mature one-seed juniper can survive at least 3 yr with 45% of ambient precipitation, the intensity and duration of drought required for substantial juniper mortality are not yet known.


Climate predictions for the southwestern USA suggest that droughts will be more frequent and severe as a consequence of global warming (e.g. Seager et al., 2007). Our results provide strong experimental evidence that 1 or more years of drought with 45% less than ambient precipitation will kill many piñons, and 3 yr of the same drought treatment will cause substantial canopy loss of one-seed juniper and suggest increased mortality in the future. Differences in mortality among species could lead to fundamental ecosystem shifts in dominant plants of woodlands and also changes in associated communities dependent on these plants (Allen & Breshears, 1998, 2007; Brown et al., 2001). Furthermore, biotic agent impacts may lag drought, which may provide opportunities for future investigators to examine the correlation between droughts and bark beetle outbreaks more fully (e.g. Williams et al., 2012). Although our study cannot address whether bark beetles were the ultimate agent of tree mortality during drought, our study shows that drought predisposes piñon trees to bark beetle attacks. Furthermore, increased temperatures, which occur often with drought in the southwestern USA, could increase bark beetle populations and tree mortality via increases in insect reproductive rate and geographic range (Ayres & Lombardero, 2000; Tran et al., 2007; Bentz et al., 2010). Increased availability of susceptible hosts for bark beetle attacks during warming and drought could lead to increased bark beetle outbreaks and further landscape-scale tree mortality.


Thanks to K. Barrett, J. Hockersmith, M. McKinney, N. Gehres, and J. Kane for assistance with laboratory and field work. Thanks to the Northern Arizona University Statistical Consulting lab and Chonggang Xu (Los Alamos National Labs) for assistance with statistical analyses. We also thank Matt Ayres and two anonymous reviewers for valuable comments. Support for this research was provided by grants from the National Institute of Climatic Change Research (NICCR), Department of Energy (DOE) Terrestrial Carbon Program, and Drought Impacts on Regional Ecosystems Network (DIREnet via NSF). The precipitation manipulation experiment was funded by grants to N.G. McDowell (LANL) and W.T. Pockman (UNM) by the Office of Science (BER), US DOE.