Introduction
Bark beetles are important agents of forest disturbance that are influenced by climate change in ways that are expected, in most cases, to cause outbreaks to be more frequent, severe or move into new regions (Berg et al. 2006; Raffa et al. 2008, Bentz et al. 2010). Recent climate change, including warmer, drier conditions throughout much of the range of Picea engelmannii (Easterling et al. 2000; Rehfeldt et al. 2006; Seager et al. 2007) are thought to have escalatedDendroctonus rufipennis populations from endemic to outbreak (DeRose and Long 2007, 2012a). Drought conditions have continued to be linked with recent D. rufipennis outbreaks (Hart et al. 2014a, 2017). However, the mechanistic underpinnings of why bark beetle populations tend to transition from endemic to outbreak during drought conditions is lacking for most species of bark beetles.
D. rufipennis outbreaks may be influenced both directly and indirectly by climate (Fig. 1). During the transition from endemic to outbreak beetle population levels, warmer winter and summer temperatures promote an amplified lifecycle; beetles reproduce sooner and more often (Fig. 1; Hansen et al. 2001; Hansen and Bentz 2003; Raffa et al. 2008). Moreover, as long as beetles do not lack for cold-hardening, they are expected to have larger broods due to increased overwinter survival (Massey and Wygant 1954; Bentz et al. 2010). Indirect effects of climate on bark beetle outbreaks are thought to include drought stress that increases susceptibility of host trees that have less non-structural carbon available for inducible defenses (Fig. 1; Raffa and Berryman 1983; Waring and Pitman 1985; Herms and Mattson 1992; Coops et al. 2009; Bentz et al. 2010; Anderegg et al. 2015). Indeed, tree vigor, often interacting with tree size, is important for determining tree susceptibility to D. rufipennis (Massey and Wygant 1954; Hard 1983, 1985; Holsten 1984; Doak 2004). Two intensive experiments have demonstrated how drought can weaken trees and make them more susceptible to bark beetle attack (Gaylord et al. 2013; Kolb et al. 2019) but these findings may not simply translate to endemic beetle populations transitioning to incipient or full irruptive outbreak conditions across landscapes, and to other bark beetle and host species.
Retrospective tree-ring studies can help decipher landscape-level influences on beetle outbreaks and have linked regional droughts toD. rufipennis outbreaks (DeRose and Long 2012a; Hart et al. 2014a). In these studies, meteorological droughts were associated with warmer temperatures, reduced snowpack, and longer growing seasons (Winchell et al. 2016; Pederson et al. 2011; Williams et al. 2013; Diffenbaugh et al. 2015; Belmecheri et al. 2016). Due to the covariance of warmer temperatures and drought conditions, bark beetle populations could have been impacted directly by life-cycle amplification during a drought without hosts experiencing drought stress that significantly impacted their ability to marshal resin and other chemical defenses used to ward off bark beetle attacks (Fig. 1). Indeed, P. engelmanniigrowing at high elevation are exposed to summer temperatures that are much cooler and, in many locations, summer monsoon or convective rainfall is more common when compared to the middle- or low-elevation forests. It is exactly these low-elevation trees that are sought in sampling to reconstruct past regional droughts (Williams et al. 2013). Overall, experimental studies provide mechanistic knowledge and retrospective studies have provided patterns linking drought to bark beetle outbreaks. However, it remains unclear whether past D. rufipennis outbreaks were directly driven by warmer temperatures with little role for host resistance, or whether drought conditions have played a role in altering host physiology to the extent that resistance to beetle attacks was compromised (Fig. 1).
Tree-ring carbon isotope discrimination (∆13C) can often provide unique insights on drought stress among trees and across landscapes. In areas with abundant growing season sunlight, tree-ring ∆13C records the degree to which stomatal conductance constrains canopy-integrated leaf gas exchange, which in turn records a shorter-term atmospheric water deficits and longer-term soil moisture deficits integrated across weeks to months in most cases (McCarroll and Loader 2004; Saurer and Voelker 2020). Where soil moisture is abundant in snowy montane forests, tree-ring ∆13C primarily records summer drought stress imposed by past variation in temperatures (Ratcliff et al. 2018). More specific to D. rufipennis outbreaks, tree-ring ∆13C had correlations with temperature that were stronger in beetle-killed Picea glauca as compared to surviving conspecifics in Alaska (Csank et al. 2016). Altogether these findings provide support for the use of ∆13C to provide a canopy-integrated record of past leaf gas exchange responses to summer drought stress.
Most retrospective studies of bark beetle outbreaks have compared various aspects of live and dead trees without regard for when dead trees were attacked and died and/or whether the outbreak was over (McDowell et al. 2010; Knapp et al. 2013; Csank et al. 2016; DeRose et al. 2017), and this could present opportunities or problems for interpreting results. For example, if >95% of overstory trees died during an outbreak, sampling of survivors would represent an extreme tail in the distribution of tree characteristics that could yield insights on how host genetics or physiology impacted beetle outbreaks if the only trees surviving outbreaks had exceptional capacity for defense. However, sampling the <5% of overstory trees that survived may dilute those signals where trees grew in locations where substantial beetle populations did not disperse, and/or microsite conditions benefitted tree defenses. Sampling of the >95% of trees that died during an outbreak would likely include a small proportion of trees that died early and had lower resistance; whereas a much larger proportion would have had stronger resistance, but ultimately not enough to successfully defend against mass attacks during the peak of a bark beetle outbreak. To avoid comparisons that are potentially disparate by way of demography, we developed an iso-demographic approach based on carbon isotope signals and the timing of tree death to contrast trees that died early vs late during an outbreak (Fig. 2; see Appendix S1: Fig. S1 in Supporting Information).
For our P. englemannii system, if host drought stress was an important driver advancing D. rufipennis outbreaks, we hypothesized that the drought sensitivity of ∆13C in early-dying trees should be greater than late-dying trees (Fig. 1). Supported by theory (Herms and Mattson 1992) and evidence that tree vigor was one of the major drivers of past beetle attack success (Massey and Wygant 1954; Hard 1983, 1985; Holsten 1984; Doak 2004), this approach implicitly assumes that early-dying trees, would have had fewer resources to defend themselves relative to tree size, whereas late-dying trees would have had more resources for induced defenses relative to tree size, but were eventually overwhelmed by outbreak beetle populations. In contrast, if the drought sensitivity of ∆13C were found not to differ among early- and late-dying trees, this would provide evidence that drought stress did not significantly alter host physiology and that the direct effect of temperature on beetle populations was primarily responsible for the transition from endemic levels to an irruptive outbreak. Hence, this study uses a well-characterized D. rufipennis outbreak that killed 95% of overstory P. engelmannii between ca . 1990 and 2000 (DeRose and Long 2007) to identify early- and late-dying trees and thereafter test the relative influence of drought stress versus warmer temperatures in the development of that outbreak.