Drought supersedes warming in determining volatile and tissue defenses of piñon pine (Pinus edulis)

Trees are suffering mortality across the globe as a result of drought, warming, and biotic attacks. The combined effects of warming and drought on in situ tree chemical defenses against herbivory have not been studied to date. To address this, we transplanted mature piñon pine trees—a well-studied species that has undergone extensive drought and herbivore-related mortality—within their native woodland habitat and also to a hotter-drier habitat and measured monoterpene emissions and concentrations across the growing season. We hypothesized that greater needle temperatures in the hotter-drier site would increase monoterpene emission rates and consequently lower needle monoterpene concentrations, and that this temperature effect would dominate the seasonal pattern of monoterpene concentrations regardless of drought. In support of our hypothesis, needle monoterpene concentrations were lower across all seasons in trees transplanted to the hotter-drier site. Contrary to our hypothesis, basal emission rates (emission rates normalized to 30 °C and a radiative flux of 1000 μmol m−2 s−1) did not differ between sites. This is because an increase in emissions at the hotter-drier site from a 1.5 °C average temperature increase was offset by decreased emissions from greater plant water stress. High emission rates were frequently observed during June, which were not related to plant physiological or environmental factors but did not occur below pre-dawn leaf water potentials of −2 MPa, the approximate zero carbon assimilation point in piñon pine. Emission rates were also not under environmental or plant physiological control when pre-dawn leaf water potential was less than −2 MPa. Our results suggest that drought may override the effects of temperature on monoterpene emissions and tissue concentrations, and that the influence of drought may occur through metabolic processes sensitive to the overall needle carbon balance.


Introduction
Trees are suffering mortality related to drought, warming and biotic attacks from pests and pathogens across the globe (Allen et al 2010, 2015, Hartmann et al 2018. Pines and other tree species use defensive secondary metabolites, such as volatile terpenes, to defend against herbivory and other biotic attacks Dudareva 2007, Unsicker et al 2009). Monoterpenes (C 10 H 16 ) in particular help plants resist herbivory by acting as oviposition and feeding deterrents (Hummelbrunner andIsman 2001, Abdelgaleil et al 2009), negatively affecting larval performance and survival through toxicity (Lerdau et al 1994, Thoss and Byers 2006, Ilse and Hellgren 2007, decreasing insect immunocompetency , and aiding in parasitoid host location (Thaler et al 2002, Kant et al 2004, Mithöfer et al 2005. Terpenes are multifunctional: on the one hand, high concentrations of particular compounds are toxic to bark beetles and fungal symbionts, but bark beetles can also use volatile monoterpenes to identify preferred host trees and as precursors for aggregate pheromone production (Raffa et al 2005, Seybold et al 2006. Despite the acknowledged importance of monoterpenes in affecting higher trophic level interactions and contributing to tree resistance, the effect of warming with drought on their synthesis and emissions has not been evaluated despite observed and predicted increases in temperature and drought conditions. The composition and concentration of monoterpenes in conifer oleoresin is dynamic and can be altered by biotic and abiotic factors, including drought (Zulak andBohlmann 2010, Keefover-Ring et al 2016). The interactive effect of these stressors is expressed through shifts in monoterpene synthesis and emission (or volatilization). Monoterpene emission rate is generally unaffected by mild drought (Staudt et al 2002b, Lavoir et al 2009, Niinemets 2010, Eller et al 2016 and decreases only during more severe droughts (Llusià and Peñuelas 1998, Staudt et al 2002a, Lavoir et al 2009, Trowbridge et al 2014. The internal concentration of monoterpenes in the needles-and seasonal variations therein-generally increase under drought stress (Blanch et al 2009), but this can vary with drought severity (Niinemets 2015) and may not reflect the composition of released volatiles Peñuelas 1998, Trowbridge et al 2014). In other words, drought often reduces monoterpene emission rate without a decrease-or possibly an increase-in the rate of monoterpene synthesis (Lavoir et al 2009) and is one of the major uncertainties in models of biogenic volatile organic compound (BVOC) emissions (Seco et al 2015). Although drought is expected to reduce plant defensive emissions, the effect of additional warming with drought has not been evaluated. An increase in monoterpene emissions under warming could mitigate drought-related decreases in emissions with implications for plant defense and herbivore success in a changing climate. So, while temperature is commonly thought to be the dominant control over emission rate, which in turn can affect foliar concentrations, plant water status during drought may mask its impact. Unfortunately, we lack a basic understanding of how these two variables interact to affect monoterpene production and emission rate in mature conifers in the field.
While several studies have documented the interaction between drought, increased temperatures, and bark beetles in semi-arid and arid areas (e.g. Negrón and Wilson 2003, Breshears et al 2005, Mueller et al 2005, Gaylord et al 2013, the interactive mechanisms underlying an increased susceptibility to insects and pathogens remain unknown. Retrospective studies show trees that succumbed to drought stress tend to have smaller resin ducts and a lower resin duct area (as a percent of xylem area) relative to trees that lived (Gaylord et al 2013(Gaylord et al , 2015. However, contrary to other conifer systems (e.g. Blanche et al 1992, Kane andKolb 2010), resin flow and resin duct parameters are not correlated in piñon pine (Gaylord et al 2013), which complicates our ability to generalize how drought-induced changes in tree defenses influence destructive pests. Furthermore, recent work on lodgepole pine (P. contorta) and whitebark pine (P. albicaulis) shows that constitutive and induced terpene concentrations are unrelated to resin duct size and abundance (Mason et al 2018), suggesting an uncoupling of anatomical and chemical anti-herbivore traits that may also be present in other species of pines in semi-arid areas.
Tree defense against biotic agents affects whether trees die or survive during drought, but defense is rarely studied or integrated into vegetation models (McDowell et al 2011, Dietze andMatthes 2014). Furthermore, studies often assume that plants are in a 'steady-state' (Anderegg et al 2013), failing to account for seasonal changes in growth or defense despite seasonal changes in their metabolic demand for carbon and subsequent influence against herbivory (Trowbridge et al 2014). It is thus critical to account for seasonality and phenology when evaluating how changing environmental conditions interact with plant primary and secondary metabolism to promote susceptibility to biotic agents. Here, we used a field transplant experiment to quantify monoterpene foliar concentrations and emission rates under temperature and moisture conditions that are consistent with global change projections across the vegetative growing season. We focused our study on the piñon pine Pinus edulis, a widespread, well-studied species that has undergone extensive drought-herbivore-related mortality (Breshears et al 2005(Breshears et al , 2018. Trees were transplanted downhill to a hot and dry site outside the species' range (hereafter, 'Hotter-Drier'), and to a site near their origin which served as an ambient temperature control (hereafter, 'Ambient') (Law et al 2019). Within each site, piñon pine water status was altered by watering trees and by adding impervious barriers to prevent water from entering the soil as precipitation or leaving through soil evaporation. Foliar monoterpene concentrations and emission rate-in addition to common meteorological drivers and leaf-level physiological variables-were measured throughout the growing season at both sites and across all treatments. We expected that trees transplanted to the Hotter-Drier site would exhibit lower needle monoterpene concentrations due to volatilization and assumed that trees would have limited capacity for investment of C toward secondary metabolism that could offset volatilized losses. We further hypothesized that emission rate will be determined primarily by atmospheric meteorological factors rather than foliar physiological processes as previously determined in Trowbridge et al (2014).

Materials and methods
Field study sites Field campaigns took place monthly during the 2010 May through September growing season at two experimental sites in northern Arizona, USA: A piñon pinejuniper woodland at 35.49°N 111.85°W (Ambient), and a hotter-drier site at 35.45°N 111.50°W (HFotter-Drier) ca. 35 kilometers east of the Ambient site (Law et al 2019). Air temperature at the Hotter-Drier site was 1.5°C warmer on average than the Ambient site when measurements were made. In October 2008, piñon pine trees of similar size and age were randomly selected from the Ambient site and, using a 2.3 m tree spade, were immediately transplanted into open areas of the site as well as at the lower elevation Hotter-Drier site. The transplanted trees were regularly watered until August 2009 when the drought treatments began, resulting in 100% survival following the transplant at both sites (Law et al 2019). All selected trees had nearest neighbor canopy-tocanopy distances of at least one meter and were transplanted in a grid of squares with a spacing of at least 10 m. Eighteen trees at each site were randomly assigned to three moisture treatment groups: watered (n=3), ambient (n=6), and barrier (n=9). Ambient trees were left exposed to natural conditions, watered trees received 25 gallons of water one week prior to sampling on a monthly basis, and barrier trees were fitted with a 4.3 m×4.3 m below-canopy rainout shelter consisting of a UV resistant tarp elevated just above the soil surface on a PVC frame to prevent direct vertical exchanges of water between the soil and atmosphere of the transplanted trees. These treatments created variability in pre-dawn water potential (Ψ) within each site. Pre-dawn water potential measurements were made within one week of gas exchange and monoterpene flux measurements, which are described below. See appendix A in Law et al (2019) for additional details on the field transplant experiment.
Field measurement techniques: gas exchange Branch gas exchange was measured using a portable photosynthesis system with a transparent conifer chamber (LI-6400, LiCor Inc., Lincoln, NE, USA) to obtain rates of net CO 2 assimilation (A) and stomatal conductance (g s ). During each field campaign, we conducted gas exchange measurements over a fourday period alternating each day between the two sites. The trees we measured throughout the day at each site were randomly selected during each measurement period so as not to introduce diurnal biases in temperature and light over the course of the experiment. All measurements were taken at 400 ppm CO 2 with a flow rate of 500 μmol s −1 . Repeated measures were performed on the same trees each month by placing ca. 4 cm of the terminal portion of each study branch in the conifer chamber and sealing with a silicone polymer (Silly Putty ® , Crayola LLC, Easton PA, USA) to minimize leaks. Because there was no controllable light source in the conifer chamber, measurements were only made once the light, CO 2 , and humidity measurements within the chamber were stable on sunny days. Photosynthetic photon flux density (PPFD), leaf temperature (T L ), and air temperature outside of the chamber (T a ) were also recorded. After gas exchange measurements were made and following BVOC sampling, the entire branch was harvested, and the smaller gas exchange portion versus larger area used for BVOC measurements were separated to obtain needle dry weight and leaf area using methods for volume displacement according to Chen et al (1997) (see also appendix B in Trowbridge et al 2014).
Monoterpene volatile emissions sampling A dynamic headspace branch-level enclosure was used to measure the flux of monoterpenes emitted from piñon pine branches (see supplemental materials is available online at stacks.iop.org/ERL/14/065006/ mmedia and Trowbridge et al 2014). All monoterpene flux measurements were conducted over a four-day period during each field campaign in the same manner as the gas exchange measurements described above. Monoterpenes were sampled onto custom-made glass adsorbent cartridges (7.6 cm in length, 0.635 cm OD, Allen Scientific Inc., Boulder, CO, USA) packed with 25 mg Tenax ® GR adsorbent (20/35 mesh, Alltech Associates Inc., Deerfield, IL, USA) between plugs of quartz wool at a flow rate of 150 ml min −1 for 10 min. Inlet and sample flow rates were set and controlled using mass flow controllers and a four-channel power supply readout box (MKS Instruments Inc., Methuen, MA, USA) to enable two branches to be measured simultaneously. Three branches were measured per tree plus one control (an empty chamber). Immediately after sampling, the glass tube was disconnected from the outlet and both ends were capped with Swagelok fittings, cooled to 0°C, transported back to the lab, and stored in a −20°C freezer for chemical analysis. Sample branches were then immediately cut and dry weight was measured after drying the needles at 60°C for 48 h.

Foliar samples and monoterpene chemical analysis
At the time of collection, one-year old needles were flash frozen and stored in liquid nitrogen, transported to the laboratory, and placed in a −80°C freezer. Details for monoterpene extraction are provided in the Supplemental Material.

Statistical analyses
We fit an exponential model between Ψ and: (1) net assimilation rates (A), and (2) stomatal conductance rates (g s ), across all treatments and sites using a nonlinear least square curve fitting procedure. Total and individual monoterpene tissue concentrations and emission rate observations were log transformed to meet assumptions of normality. To determine the effect of site on emission rate and foliar monoterpene concentrations over time, total and individual emission rate and monoterpene concentration data from the two sites were analyzed using a repeated measured ANOVA (SAS v 9.3; PROC MIXED statement) with tree listed as a random effect nested within each water treatment and applying a Bonferroni correction. We used Welch's two-sample t-test to detect any differences in average environmental and physiological variables between the two sites.
To quantify the variables responsible for the seasonal variability of emission rate and monoterpene concentrations in piñon pine needles over the growing season, we constructed linear models of meteorological variables (temperature, vapor pressure deficit [D], PPFD), plant physiological variables (namely A, g s , Ψ, and leaf internal CO 2 concentration, C i , as a surrogate for photosynthesis that excludes g s ), and sitelevel information (tree, drought treatment, and site). Models were fit for each month using all available observations, and for the entire growing season using month as an explanatory variable. Concentration was included as an input for the emission rate models, and emission rates were included as an input for concentration models. The parsimonious model with the minimum value of Akaike's Information Criterion corrected for finite sample sizes (AICc) (Akaike 1974, Burnham and Anderson 2002) was selected using the dredge command in the 'MuMln' package (Bartoń 2018) that follows the model selection routines of Burnham and Anderson (2002) using R (R Core Team 2017). In brief, dredge explores all combinations of linear models and selects the one with the lowest AICc to penalize models with additional parameters (k) and favor models with high values of the likelihood function L while accounting for sample size n. We also relaxed the assumption of linearity between driver and response by creating linear models for the natural logarithm of emission rate and monoterpene concentrations. Observations from May lacked leaf-level ecophysiological observations and were excluded from the modeling analysis.

Results
Environmental and plant physiological differences between sites During the sampling periods, the Hotter-Drier site was, on average, 1.5°C warmer than the Ambient site (28.7°C versus 27.2°C; P<0.03) as noted and sampling chamber measurements were 1.7°C warmer (29.4°C versus 27.7°C; P<0.003). Ψwas significantly lower at the Hotter-Drier site (−1.80 MPa versus −1.54 MPa; P<0.0001). A and g s exhibited exponential relationships to Ψ (figure 1) and as a consequence were ∼2.5-fold greater, on average, at the Ambient site versus the Hotter-Drier site (P<0.0001). Values for A and g s averaged 0.26 μmol m −2 s −1 and 0.0031 mol m −2 s −1 at the Hotter-Drier site, and 0.64 μmol m −2 s −1 and 0.0078 mol m −2 s −1 at the Ambient site, respectively.

Seasonal patterns of monoterpene emissions
The total monoterpene emission rate did not differ between the Ambient and Hotter-Drier sites but did differ as a function of time (P<0.0001). Total emission rate was significantly greater in June at both sites relative to other months (figure 2) due to increases in emissions of α-pinene, β-pinene, β-myrcene, and limonene ( figure 3). Month was also a significant main effect (P<0.0001) for the emission rates of all individual compounds, and the relative contribution of different monoterpene compounds to total emissions changed throughout the growing season because the emission rate of each compound followed different seasonal patterns ( figure 3).

Models of the seasonal variability of monoterpene emission rates
The linear model with the lowest AICc value explained only ∼10% of the variance across all emission rate observations regardless of the exclusion of the statistical outliers, considered here to be >1 μg C gDW −1 h −1 (i.e. five standard deviations greater than the mean emission rate), that were frequently observed in June. This model included temperature, month, treatment (dry, ambient, or watered), and intercellular CO 2 concentration (C i ) as a surrogate for leaf physiology (table 1). The linear model for the logarithm of emission rate for the entire season explained nearly 20% of the variance of observations and included similar variables (table 2).
Linear models with the lowest AICc values explained ∼30% of the variance of emission rate during June and August, but only 12%-15% during July and September (table 1). Models for July and September included only T (and in the case of July also D), but not leaf-level physiological variables or experimental treatments (table 1). Models for June and August suggested that leaf-level variables C i and g s , as well as total leaf monoterpene concentrations, and D (for the case of June) and experimental treatment (for the case of August), should not be excluded from an emission rate model. Models for the logarithm of monoterpene emission rate included similar variables but tended to explain a higher amount of the observed variance, notably in August when a model that included temperature, D, C i , g s , site, treatment, and needle monoterpene concentration explained nearly 50% of the observed variance in emission rate (table 2). Regardless of month, emission rate was not related to stomatal conductance, photosynthesis or any observed micrometeorological variable when Ψwas below −2 MPa. The model with the lowest AICc under these conditions only contained an intercept value of 0.19 mg C g DW −1 h −1 .
Ψ, an integrated measure of plant water status, was not included in the selected models for emission rate during any time period despite its significant exponential relationship to both g s and A (figure 1). Emission rates in excess of 1 μg C gDW −1 h −1 , however, were Table 1. Linear models of monoterpene basal emission rates (μg C gDW −1 h −1 ) selected using the minimum value of the adjusted Akaike's information criterion (AICc) found using the dredge function in R. F: F-statistic, df: degrees of freedom, Adj. r 2 : adjusted coefficient of variation, P: P-value, T: air temperature within the sampling chamber (°C), D: vapor pressure deficit (kPa), g s : stomatal conductance (mol m −2 s −1 ), C: total leaf monoterpene concentration (mg gFW −1 ), M: month, E: experimental treatment (1: Dry, 2: Ambient, 3: Wet).  Table 2. Linear models of the logarithm of monoterpene basal emission rates (μg C gDW −1 h −1 ) selected using the minimum value of the adjusted Akaike's information criterion found using the dredge function in R. F: F-statistic, df: degrees of freedom, Adj. r 2 : adjusted coefficient of variation, P: P-value, T: air temperature within the sampling chamber (°C), D: vapor pressure deficit (kPa), g s : stomatal conductance (mol m −2 s −1 ), C: total leaf monoterpene concentration (mg gFW −1 ), M: month, E: experimental treatment (1: Dry, 2: Ambient, 3: Wet), S=site (1: Hotter-drier, 2: Ambient).

Seasonal patterns of foliar monoterpene concentrations
The total monoterpene concentration was greater in needles at the Ambient site than at the Hotter-Drier site (P<0.05) due to higher concentrations of αpinene and β-pinene (figures 4(B), (C), and (F)). Total monoterpene concentration and all individual compounds, except R-(+)-limonene, differed significantly as a function of time. Unlike total emission rate, total monoterpene concentrations did not show a pronounced change during June (figures 2 and 4).

Models of the seasonal variability of monoterpene concentrations
Selected models for monoterpene concentration for the entire growing season and for all months (table 3) always included site (Ambient or Hotter-Drier) as a variable but explained only 9%-14% of the observed variance with the exception of the model for June, which explained 29% of the observed variance and also included C i , g s , and emission rate. Models of the logarithm of monoterpene concentrations (not shown) did not increase the percent of variance explained. Notably, Ψ was included in models for monoterpene  Table 3. Linear models of monoterpene concentrations (mg gDW −1 ) selected using the minimum value of the adjusted Akaike's information criterion (AICc) found using the dredge function in R. F: F-statistic, df: degrees of freedom, Adj. r 2 : adjusted coefficient of variation, P: P-value, D: vapor pressure deficit, g s : stomatal conductance (mol m −2 s −1 ), C: total leaf monoterpene concentration (mg gFW −1 ), D vapor pressure deficit (kPa), E: experimental treatment (1: Dry, 2: Ambient, 3: Wet), S=site (1: Hotter-drier, 2: Ambient), emission rate: total monoterpene emissions rate (μg C gDW −1 h −1 ), M: month, Ψ: pre-dawn water potential (MPa).

Experimental hypotheses
As predicted, we observed lower needle monoterpene concentrations from trees transplanted to the Hotter-Drier site relative to Ambient site ( figure 4). While we hypothesized that this decrease in concentration would result from a loss via volatilization at higher temperatures, emission rates did not differ between sites consistent with an overriding effect of increased water limitation on emissions, especially as Ψdecreased below a limiting value determined to be −2 MPa. The environmental and phenological factors that determined monoterpene concentrations and emission rate were dynamic throughout the study period (tables 1-3) as discussed in greater detail in the following.

Seasonal patterns of monoterpene emissions
Few studies have explored the consequences of interactions between altered water availability and temperature on emission rate and monoterpene concentrations in situ across the vegetative growing season (Staudt et al 2002a, Geron et al 2016. A striking feature of our observations is the significant increase of emission rate during June, when nearly 25% of the emission rate measurements exceeded the growing season mean plus four times the variance of emission rate observations (figures 2a and 3). When considering emission rates from trees across the water stress gradient in June (−1 to −2.5 MPa), none of the high emission rate measurements occurred when Ψ was below ca. −2 MPa, the approximate zero-assimilation point of piñon pine (Lajtha and Barnes 1991). Although we did not measure the timing of leaf and shoot growth in our study, others have observed that needle emergence (bud break) in piñon occurs in June (Grossiord et al 2017), which could have driven the high emission rates we measured in this month. Multiple lines of evidence suggest that periods of leaf differentiation and growth are of particular importance to BVOC flux. For example, monoterpene emissions were greatest in the early growing season for both P. rigida and P. koraiensis (Son et al 2015). Furthermore, peaks in atmospheric particle formation rates during the early growing season in boreal ecosystems have been associated with springtime increases in BVOC fluxes from evergreen conifers (Dal Maso et al 2009). Models based on summertime observations underestimated BVOC emissions during earlier times of year (Holzinger et al 2006), suggesting that unique springtime processes or relationships between BVOC emissions and environmental drivers are important (Bäck et al 2005). Monoterpene emission bursts have been demonstrated in Pinus sylvestris due to new foliage growth (Aalto et al 2014), the recovery of photosynthesis (Aalto et al 2015), and possibly the refilling of embolized tracheid elements (Vanhatalo et al 2015). These observations are also consistent with recent findings that the presence of exposed resin in the axils of needles-especially when sap pressure is relatively high-is associated with high BVOC emission rates (Eller et al 2013).
We cannot exclude resin exposure as a potential contributor to the transient increases in emission rate observed in the present study, which can occur during shoot extension and needle emergence. Resin exposure is related to positive xylem pressure potentials that cause exudation (Eller et al 2013), and emission rate spikes were not observed when Ψ was below −2 MPa as noted. Our results demonstrate the need to further study different states of plant water relations and physiological function-including those that increase the likelihood of resin exposure-on BVOC emissions from piñon pine shoots and needles.
Models for monoterpene emission rates are determined by water thresholds Monoterpene basal emission rate was not related to stomatal conductance, photosynthesis, or any measured micrometeorological variable when Ψ was below −2 MPa. A and g s are strongly limited in piñon pine when Ψ is lower than −2 MPa (figure 1, Lajtha and Barnes 1991), and the best model of total monoterpene basal emission rate under these conditions is a constant 0.19 mg C g DW −1 h −1 . Above the −2 MPa threshold, a model with T, C i , monoterpene concentration, and water treatment had the lowest AICc, but explained only 14% of the observed variance in emission rate. The model for the logarithm of emission rate with the lowest AICc included similar variables and explained 5% more of the observed variance. Combined, these results suggest that emission rate is controlled by measured physiological and environmental factors above the −2 MPa Ψthreshold, but not below. These results contrast the findings of Eller et al (2016), who found that BVOC emissions are correlated with leaf physiology below a plant physiological threshold, namely a net CO 2 assimilation rate A<ca. 2 μmol m −2 s −1 and stomatal conductance g s <0.02 mol m −2 s −1 in the needles of mature P. ponderosa trees. These results point to the importance of studying physiological controls over BVOC efflux across different Pinus species to understand why differences emerge.
The role of temperature and leaf level monoterpene emissions Monoterpene emission rate tends to increase 2-3 fold for every 10°C increase in temperature (Lerdau et al 1994, Lerdau et al 1997, and this relationship is used to construct global atmospheric models of forest monoterpene emissions (Guenther et al 1991, 1993, Keenan et al 2009. Thus, we hypothesized that trees in the hotter-drier site would exhibit higher emission rate due to increased volatility. However, we did not observe a significant difference in emission rate between sites despite a 1.5°C average air temperature increase (and 1.7°C average chamber temperature increase) at the Hotter-Drier site, where Ψwas on average 0.3 MPa lower. This lack of relationship also held when emission rate was not normalized by temperature (i.e. when raw emission rate data rather than basal emission rate data were analyzed), emphasizing that temperature increases did not result in emission rate increases. Temperature was included as an explanatory variable in most models of emission rate, but the relationship, when present, was negative (tables 1 and 2). Previous studies have demosnstrated that BVOC efflux from piñon needles responds to temperature during certain parts of the growing season, but also found that water stress was a more important control over emissions in this semi-arid system (Trowbridge et al 2014).
Past studies, in addition to citing temperature as a primary driver of emission rate from conifers, have also shown monoterpene emission rate to be a function of foliar monoterpene concentrations through a Henry's Law relationship (Lerdau et al 1994, Lerdau et al 1997. Other studies have found no relationship between monoterpene concentrations and emission rate (Constable et al 1999, Trowbridge et al 2014, Eller et al 2016. In addition to our inability to find the expected positive temperature-emissions relationship in this study, we also failed to verify a relationship between monoterpene concentrations and emission rate across most time periods studied except for June (table 1). Observations point to a dynamic system where monoterpene concentrations are coupled to emission rate during periods of more positive plant water status during the growing season but are otherwise unrelated as leaf monoterpene concentrations do not factor into the most parsimonious model for emission rate across the rest of the study period.
Combined, our observations are in contrast to other studies and atmospheric models that have derived algorithms based solely on the effects of foliar monoterpene concentrations and temperature for estimating monoterpene emission rate. Our observations are consistent with empirical and modeling studies demonstrating that monoterpene emissions are not under stomatal control when averaged over longer periods of time (Harley 2013, Grote et al 2013, Eller et al 2016; g s rarely entered models of emission rate (tables 1 and 2). As noted, when Ψwas above −2 MPa, a model with T a , C i (a surrogate for A apart from g s ), needle monoterpene concentration, and experimental water treatment were necessary to explain observed emission rate variability. These observations suggest that plant water status via Ψ acts as a switch between monoterpene emission rates that are not under environmental or biotic control (below −2 MPa) and those that are, although we cannot exclude unmeasured physiological factors when interpreting these results.

The impact of drought and heat on individual compounds
The compound α-pinene was the primary constituent of total monoterpene concentrations and emission rate (figure 2, table 1), in agreement with past studies on piñon pine (e.g. Smith 2000, and drove many of the patterns that we observed for both total emissions and concentrations. Although many individual compounds exhibit similar concentration and/or emission patterns to α-pinene (e.g. β-pinene and β-myrcene, figures 3 and 4), the magnitude and percent change over time, between sites, and in response to water status, were often different among compounds, likely due to variation in physiochemical properties and controls over their synthesis (Niinemets et al 2004). Furthermore, some compounds show little or no change in either concentration or emissions across treatments or across time (e.g. β-phellandrene, figures 3 and 4), perhaps due to lower relative volatility and/or deterministic gene expression patterns that are little affected by environmental variability.
Piñon pine mortality and herbivory Piñon pine populations have declined in mass mortality events across the southwestern US, and herbivore damage to compromised trees is often attributed as the cause of tree death (Cobb et al 1997, Gaylord et al 2013. Synergistic effects among monoterpenes and the ratio of compounds to one another can influence herbivores directly by affecting growth and immunocompetence (Dyer et al 2003, but also indirectly by attracting parasitoids (e.g. Havill and Raffa 2000). However, over time and under more severe and consistent drought stress, we may expect to see a shift in monoterpene concentrations as piñon pines become more carbon limited, which may increase tree susceptibility to insect pests. The dynamics of monoterpene emission rate and concentrations under prolonged drought is uncertain and must be studied further to understand the interplay between drought, defense, and herbivory.

Conclusions
Our study represents an initial step to add tree defensive chemistry to a comprehensive understanding of piñon pine forest response to drought and temperature stress. We find that the Ψthreshold associated with zero net photosynthetic carbon assimilation in piñon pine trees, −2 MPa, also acts as a threshold below which monoterpene emission rate is not under apparent control by measured plant physiological or micrometeorological variables. Plant water status thus plays an important role in emission rate and should be incorporated into regional and global models of monoterpene efflux. We failed to observe increased emission rate with increased temperature, likely due to the overriding impact of limited water availability, indicating that such relationships need to be factored into global models. Research should focus on the underlying phenological mechanisms responsible for the frequent spikes in emission rate during conditions in which water was not yet limiting (e.g. budburst, resin exposure, or other factors) to improve our inventories of emissions under predicted global change. Future studies should also explore the dynamics of plant secondary compounds as carbon limitation and water stress reach values sufficient to cause tree death to understand how plants allocate resources toward maintenance and defense during periods of acute and/or prolonged stress, and how these changes alter herbivore dynamics. In summary, our results suggest that drought may override the effects of temperature on monoterpene emissions and tissue concentrations, and that the influence of drought may occur through processes sensitive to overall needle carbon balance. Consequently, added warming does not worsen drought-induced suppression of defensive pine emissions-findings that need to be incorporated into global biogeochemical and biogeographic models.