Interactions among spruce beetle disturbance , climate change and forest dynamics captured by a forest landscape model

The risk of bark beetle outbreaks is widely predicted to increase because of a warming climate that accelerates temperature-driven beetle population growth and drought stress that impairs host tree defenses. However, few if any studies have explicitly evaluated climatically enhanced beetle population dynamics in relation to climate-driven changes in forest composition and structure that may alter forest suitability for beetle infestation. We synthesized current understanding of the interactions among climate, spruce beetles (Dendroctonus rufipennis) and forest dynamics to parameterize and further advance the bark beetle module of a dynamic forest landscape model (LandClim) that also integrates fire and wind disturbance and climate-driven forest succession. We applied the model to a subalpine watershed in northwestern Colorado to examine the mechanisms and feedbacks that may lead to shifts in forest composition and spruce beetle disturbance under three climate change scenarios. Simulation results suggest increased droughtand beetle-induced reduction of large Engelmann spruce (Picea engelmannii ) trees while Douglas-fir (Pseudotsuga menziesii ) and ponderosa pine (Pinus ponderosa) increased in dominance throughout the study area under all climate change scenarios. This shift in forest composition and structure counterbalances the enhancing effects of accelerated beetle population development and increased drought-induced susceptibility of spruce to beetles. As a result, we projected a long-term decrease in beetle-induced spruce mortality to below historical values under all climate scenarios at low elevations (,2800 m asl). Beetle-induced spruce mortality above 2800 m asl and under moderate climate change was slightly higher and more variable than under historical conditions but decreased to 36% and 6% of historical values under intermediate and extreme climate change, respectively. Because mechanisms driving beetle disturbance dynamics are similar across different bark beetle species, we argue that the depletion of host trees due to drought and beetle disturbance may also be important in other climatesensitive beetle-host systems. We advocate for the consideration of climate-driven shifts in forest and disturbance dynamics in devising adaptive management strategies.


INTRODUCTION
Bark beetles (Curculionidae: Scolytinae) are the most important biotic disturbance agents and key drivers of the dynamics of northern hemisphere conifer forests (Raffa et al. 2008).Populations of the mountain pine beetle (Dendroctonus ponderosae Hopkins) and the spruce beetle (Dendroctonus rufipennis Kirby) have recently erupted to region-wide outbreaks in western North America causing widespread tree mortality and affecting forest ecosystem services (Price et al. 2010, Meddens et al. 2012, Hansen 2014, Hart et al. 2014a, Negro ´n and Fettig 2014).The extent and severity of bark beetle outbreaks is determined by various interactions among forest composition and structure, climate, and other disturbance agents such as fire and blowdown (Raffa et al. 2008, Bentz et al. 2010, Jenkins et al. 2014).There is abundant quantitative information on the direct effects of weather and interannual climate variations that may lead to large-scale outbreaks (Sherriff et al. 2011, Chapman et al. 2012, Bentz et al. 2014, Hart et al. 2014a).In contrast, the indirect and long-term interactions and feedbacks mediated by changes in forest composition and structure are largely unknown, particularly under a warming climate (Bentz et al. 2010, Turner 2010).
Spruce beetles are the most widespread bark beetle species in subalpine spruce-fir (Picea engelmannii Parry ex Elgelm.-Abieslasiocarpa [Hooker] Nutt.) forests of western North America (Jenkins et al. 2014), and spruce beetle outbreaks are currently expanding in the Colorado Rocky Mountains (Colorado State Forest Service 2014).The development of spruce beetle broods is temperature dependent (Hansen et al. 2001, Bentz et al. 2010) with beetles requiring two years to complete larval, pupal and adult development when relatively cool summer temperatures prevail (semivoltine life-cycle).Under warmer conditions univoltine generations develop, which may accelerate beetle population growth and facilitate escape from parasites and predators (Jenkins et al. 2014).For extensive spruce beetle outbreaks to occur, favorable temperatures need to be maintained over several years and must coincide with susceptible host forests, i.e., older and drought-stressed forest stands dominated by large spruce (Schmid and Frye 1977, Hart et al. 2014b, Jenkins et al. 2014).In addition, an abundance of large, downed trees following wind disturbance provides an ideal breeding substrate for beetles, which has commonly been associated with outbreak initiation (Schmid andFrye 1977, Schmid 1981).
Given these interacting drivers, climate change may alter spruce beetle disturbance in several ways.Projected warmer temperatures will likely accelerate beetle population growth, allowing outbreaks to occur at higher elevations and latitudes than today (Bentz et al. 2010, DeRose et al. 2013).In the short term, increased droughtstress may weaken tree defenses and increase susceptibility to beetles (Bigler et al. 2006, McDowell et al. 2011, Hart et al. 2014a).Over longer time frames (decades to centuries) climate change may affect beetle disturbance indirectly (Bentz et al. 2010).For example, expanding populations of drought-adapted Douglas-fir (Pseudotsuga menziesii [Mirb.]Franco) and ponderosa pine (Pinus ponderosa Douglas ex Lawson) at the expense of cold and wet-adapted spruce and fir (Rehfeldt et al. 2006) may reduce the susceptibility of subalpine forests to spruce beetles in the long term.Increased fire and beetle disturbance due to climate warming may hasten such vegetation shifts, resulting in a negative feedback mediated by a depletion of susceptible hosts and a greater portion of the landscape in early seral stages (Veblen et al. 1991, Westerling et al. 2011, Hart et al. 2015).These stages are typified by high proportions of the non-host aspen (Populus tremuloides Michx) and subalpine fir, and younger, smaller trees that are less susceptible to subsequent bark beetle disturbance and blowdown.Thus, interactions among recurring disturbances may attenuate the potential for spruce beetle outbreak in the long term (Bigler et al. 2005, Kulakowski et al. 2012, Hart et al. 2014b).
While these short-term, direct and long-term, indirect effects of climate change on beetle disturbance have been documented individually, their relative importance under various scenarios of climate change in western North America is unclear.We hypothesize that forest composition and structure will be altered due to direct effects of climate on tree growth and survival and due to climate-sensitive disturbances by beetles and fire, and that these changes will reduce forest v www.esajournals.orgsusceptibility to subsequent beetle disturbance (Bebi et al. 2003, Rehfeldt et al. 2006, Kulakowski et al. 2012).Consequently, warming temperatures and increased drought-stress may not result in widespread spruce beetle disturbance in the long term, as a result of a reduced abundance of susceptible host trees (cf.Økland andBerryman 2004, Temperli et al. 2013).
To test this hypothesis and reveal the underlying mechanisms and their relative importance, a simulation approach that integrates climatesensitive forest succession and disturbance interactions over a wide range of spatial and temporal scales is required.Dynamic forest landscape models (FLM) that simulate forest and disturbance dynamics based on spatially explicit input of topographic, climate and other environmental data have been developed to meet such objectives (Scheller andMladenoff 2007, Seidl et al. 2011).Here, we used the climate-sensitive FLM LandClim (Schumacher et al. 2004) that integrates patch-scale forest succession with landscape-scale bark beetle, wind and fire disturbance (Schumacher et al. 2006, Temperli et al. 2013).Using the latest research we adapted the model to represent spruce beetle disturbance in the Colorado Rocky Mountains and applied it to project forest dynamics and bark beetle disturbance under a range of climate change scenarios.

LandClim model
LandClim is a spatially explicit stochastic landscape model that combines fine-grained forest dynamics with broad-scale bark beetle, fire and wind disturbance to simulate a dynamic forest landscape in response to climate, topography and soil properties (Schumacher et al. 2004).Recruitment, growth, mortality and competition among age cohorts of different tree species are simulated with a simplified gap-model (Bugmann 2001) in response to monthly mean temperature, drought (a function of temperature, precipitation, and soil water-holding capacity) and light-availability in 25 3 25 m cells at a yearly time step.Assuming trees within an age cohort to be the same size, the model tracks the number of stems, the above-ground whole tree biomass and derived variables such as diameter at breast height (DBH) for each age cohort.Seed dispersal and disturbances link individual cells across the simulated landscape (typically 10-100 km 2 ) and act on recruitment and mortality on a decadal time step.Here we provide a brief overview and refer to Schumacher (2004) and Temperli et al. (2013) for detailed descriptions of LandClim.
Beetle and fire disturbance regimes emerge from climate-forest interactions.The LandClim beetle module uses bioclimatic and forest stand structural variables to assess beetle population dynamics and forest susceptibility to bark beetle disturbance.These assessments are used to simulate the extent, occurrence and severity of beetle-induced tree mortality.While the beetle module was designed and structured to be generally applicable for northern hemisphere climate-sensitive bark beetle-host systems, it was originally parameterized to represent disturbances by the European spruce bark beetle (Ips typographus Linnaeus) in Norway spruce (Picea abies Karst.; Temperli et al. 2013).Here, we parameterized the beetle module for D. rufipennis in Engelmann spruce (P.engelmannii ).
Fire spread is simulated using a cell-to-cellpercolation model with spread probability dependent on the presence and moisture content of fuels in three size classes.Fire-induced mortality is a function of (1) bark thickness that depends on tree species and tree size and (2) crown scorch that in turn is a function of drought, available fuel biomass and tree size (Schumacher et al. 2006).Wind disturbance is simulated stochastically based on user-defined frequency and patch size distributions.In blowdown patches, the probability of wind-induced mortality increases with tree size.Blown down spruce trees increase the susceptibility of a forest patch (cell) to spruce beetle infestation in the decade wind disturbance occurs.For subsequent decades wind disturbance reduces susceptibility to beetle infestation via the reduction of large susceptible spruce (Temperli et al. 2013).LandClim has previously been applied to reconstruct and project forest and disturbance dynamics including historical fire regimes in the Colorado Front Range (Schumacher et al. 2006, Henne et al. 2013).

Model parameterization
Parameters that specify life history and responses of tree species to environmental drivers v www.esajournals.org(Schumacher et al. 2004) were available from Schumacher et al. (2006) for the five most common conifers in the Colorado Rocky Mountains (Engelmann spruce, subalpine fir, lodgepole pine, Douglas-fir and ponderosa pine).For trembling aspen we derived a new set of parameters from relevant literature (Appendix A: Table A1).
To adapt the LandClim beetle model to D. rufipennis we synthesized findings from recent spruce beetle research to (1) re-parameterize the relationships used to assess forest susceptibility to beetles, (2) assess temperature-dependent beetle population development and (3) parameterize an improved beetle dispersal routine.Here, we summarize the parameterization of these relationships and refer to Appendix A for additional details.
Following operational stand rating systems (Schmid and Frye 1976, Shore and Safranyik 1992, Reynolds and Holsten 1994), susceptibility to infestation at the level of individual cells (S cell ) is assessed based on drought-mediated tree resistance (S dr ), the DBH of the largest spruce cohort in the cell (S dbh ), the basal area share of spruce (S s ), and the amount of wind thrown spruce biomass (S w ).S dr , S dbh and S s were modeled with logistic functions.The assessment of S dr was based on the LandClim-inherent and cell-specific drought index (Bugmann andCramer 1998, Schumacher et al. 2004) and drought-dependent spruce beetle outbreak probabilities in northwestern Colorado (Hart et al. 2014a).S dbh was related to spruce DBH based on the DBH-dependent, tree-level infestation probabilities found by Hart et al. (2014b) and S s is based on Schmid and Frye's (1976) spruce sharespecific hazard rating scores.Because no quantitative data were available to relate S w to wind thrown spruce biomass we adapted the simple linear relationship of the original model (Temperli et al. 2013).We combined S dr , S dbh , S s and S w additively to obtain the cell-specific susceptibility index S cell .The effect of wind thrown spruce (S w ) was weighted four times more than the effects of the other susceptibility factors (Appendix A: Eq.A9; Dyer andTaylor 1971, Reynolds andHolsten 1994).A sensitivity test showed that the dynamics of simulated spruce beetle disturbance under climate change are robust under this weighting (Appendix A: Figs.A7-A9).Temperature-dependent beetle population development is modelled using index G, which is scaled between 0 and 1 and can be considered as a proxy for the potential number of bark beetles that are able to develop per season, contingent on the temperature sum (Temperli et al. 2013).We first calculated the proportion of univoltine (as opposed to semivoltine) spruce beetles (P u ) for each cell based on annual degree day sums using an approximation of the spruce beetle phenology model previously used in the BioSIM simulation model (Appendix A: Fig. A8; Hansen et al., unpublished manuscript;Bentz et al. 2010).We accounted for accelerated beetle population growth and concurrent increasing beetle mortality rates with increasing P u due to densitydependent competition and parasite pressure by relating G to P u using a logistic function (Hansen andBentz 2003, Wallin andRaffa 2004).
To parameterize beetle dispersal (Appendix A: Eq.A1) we relied on results of mark-recapture experiments that indicate that the yearly spruce beetle dispersal distance ranges between 90 and 300 m (Werner and Holsten 1997).Assuming that the probability of spruce beetles flying 300 m is 0.01 allowed us to calculate decadal (the time step of the beetle model), distance-dependent dispersal probabilities.
To estimate the importance of beetle pressure arising from beetle population dynamics in relation to forest susceptibility and to apply LandClim under climate change scenarios, we selected a case study landscape that is representative of subalpine forests of northwestern Colorado.The study landscape is a 39.7 km 2 watershed within the Routt National Forest of northwestern Colorado (Fig. 1) where fire and spruce beetle outbreak history was reconstructed in previous studies (Veblen et al. 1994, Eisenhart and Veblen 2000, Kulakowski and Veblen 2002).The watershed encompasses elevations between 2200 and 3340 m asl and is dominated by aspen and mixed-conifer (mostly Douglas-fir and lodgepole pine) forests at lower elevations (ca.,2800 m asl) and Engelmann spruce and subalpine fir at higher elevations (Fig. 1A, B; LANDFIRE, Ryan and Opperman 2013).Similar to many subalpine forests of northwestern Colorado, the lower elevation aspen, mixed conifer forests regenerated following fire in the late 19th century, while the higher elevation spruce-fir forests have no evidence of extensive fire in the past three centuries (Kulakowski and Veblen 2002).We used climate data that were gridded to 5 arc-minutes from PRISM (Daly et al. 2008, Coulson et al. 2010) for both the historical  and future (2010-2099) time series (Table 1) and calculated elevation lapse rates for monthly temperature and precipitation using the 1971-2000 climate normal data that were available from PRISM on a 30 arc-second grid.Elevation, aspect and slope were available from a digital elevation model with a resolution of 30 m (U.S. Geological Survey 2013).We derived cellspecific available soil water capacity (AWC) in the case study landscape by interpolating coarsescale AWC estimates from the STATSGO data-base using elevation and the compound topographic index (Appendix A: Fig. A9; Gessler et al. 2000, Henne et al. 2013).
The importance of beetle pressure (P bd ) relative to forest susceptibility is determined by the scaling coefficient C pb (Appendix A: Eq.A2).Because no direct empirical basis was available, we estimated C pb using an inverse modelling approach (Hartig et al. 2012, Temperli et al. 2013).We ran LandClim with varying estimates of C pb , compared simulated spruce beetle disturbance rotations with tree ring-derived spruce beetle disturbance rotations, and then inferred the C pb value that best matched the simulated and empirical rotations.To this end, we ran LandClim under the historical (1900-2009) cli- v www.esajournals.orgmate using the above topographic and soil input.These model runs were initialized with simulated data that approximate conditions in 1900 (Fig. 1C), as we describe and evaluate below.Simulation results from the latter 110 years were used to calculate spruce beetle disturbance rotations as the ratio between the respective time period (110 years) and the proportion of infested area during that time.A value of 0.08 for C pb resulted in a beetle rotation of 205 years, which best matched the empirically found ''turnover time'' of 259 years (Veblen et al. 1994) and ''recurrence interval'' of ,100 years (Eisenhart and Veblen 2000) close-by and partly within our case study landscape, respectively.
In these and subsequent simulations we included both fire and wind disturbances.Fire was simulated as in Schumacher et al. (2006), which resulted in a fire rotation of 678 years under historical climate and is consistent with previous research on fire rotation in Colorado's spruce-fir zone (Veblen et al. 1994, Buechling and Baker 2004, Sibold et al. 2006).To account for relatively frequent small and rare large windthrow events in Colorado's subalpine forests (Veblen et al. 1991, Kulakowski andBebi 2004), we parameterized stochastic wind disturbances in LandClim to a rotation of 489 years and a minimum, maximum and 95th-quantile wind throw patch size of ,0.1, 322.2 and 9.6 ha, respectively.

Model evaluation
We were not able to formally validate Land-Clim simulations because independent data that have not been used for model parameterization were not available.Nevertheless, we scrutinized the model for its applicability for projecting the dynamics of Colorado's subalpine forests and spruce beetle disturbance regimes by evaluating several aspects of the model.First we tested whether LandClim adequately represents the interactions among climate, soil, spruce beetle, fire and wind disturbances, and forest dynamics such that the expected distribution of tree species in the case study landscape can be reproduced.To this end we ran LandClim with the previously described parameterization for 3000 years attaining a pseudo-equilibrium state using a climate time-series that we constructed from 3000 random samples of the years 1900-2009.In pseudoequilibrium, fluctuations of forest characteristics such as forest biomass or species composition around a long-term equilibrium are driven by interannual climatic variations and disturbances.To approximate the vegetation that developed after the late 19th century fires at lower elevations of our case study, we first selected a decade when a series of large fires was simulated due to the simultaneous occurrence of conducive climate and fuel state.These simulated fires burned almost all forest below 2900 m asl.We then compared the simulated vegetation that developed 140 years after these simulated fires to the mapped distribution of the currently existing forest vegetation types from LANDFIRE.
A second test evaluated the sensitivity of the LandClim spruce beetle module to temporal variations in temperature and drought by comparing simulated with empirically documented temporal dynamics of spruce beetle disturbance in northwestern Colorado (Hart et al. 2014a).To this end, we simulated forest and disturbance dynamics on a generic 1 km 2 , flat landscape with 1900-2009 PRISM climate data that were averaged over the northwestern Colorado spruce-fir zone, i.e., the same data set that Hart et al. (2014a) used in their analyses of spruce beetleclimate interactions.To account for uncertainties regarding initial forest state and wind and fire Table 1.Historical (1961Historical ( -1990) ) annual and summer (Apr-Sept) mean temperature (T) and precipitation sums (P), and average (2070-2099) anomalies thereof based on output from three different global circulation models forced by the IPCC AR4 A2 emission scenario that were interpolated to a 5 arc-minute grid (Joyce et al. 2011).Data refer to a grid point within our case study area at 2754 m asl.In a third test we evaluated whether LandClim adequately captures long-term, forest-mediated disturbance interactions, i.e., the observed pattern of decreased beetle infestation in young forests that developed following beetle, fire and wind disturbance (Schmid and Frye 1977, Veblen et al. 1994, Bebi et al. 2003, Kulakowski and Veblen 2006, Hart et al. 2015).To this end we analyzed the output data from the spinup simulations under the randomized historical climate that we used to initialize the above climate-sensitivity test.We assessed the probability that a cell experiences beetle infestation in the 100 time steps (decades) following wind, fire and beetle disturbance by calculating the percentage of beetle infested cells that have been disturbed by beetles, fire and wind in 1-100 previous time steps.

Model application
To test our hypothesis that climate change may reduce forest susceptibility to spruce beetle in the long-term and that this may reduce spruce beetle-induced tree mortality even if climate becomes more favorable for the beetles, we applied LandClim to the study watershed (Fig. 1A) under three widely different climate change scenarios (Table 1; Appendix B: Figs.B1-B3).The climate scenarios are projections of three Global Circulation Models (CGCM31, CCSM3 and MIR-OC32) that were forced by the medium-high IPCC AR4 A2 emission pathway and for which the data were interpolated to a 5 arc-minute grid (Joyce et al. 2011).While the CCSM3 GCM is being considered as particularly suitable for the western United States (Cayan et al. 2009, Westerling et al. 2011), the outputs of the two other GCMs were used to encompass the extreme ends of climate change projection uncertainty, irrespective of their suitability for the western United States.The models respectively project 4.48C, 5.28C and 7.08C increases in annual mean temperature (AT) and À9%, þ12% and À29% changes in summer precipitation (SP) for the end of the century (2070-2099; Table 1).Hence we here use the acronyms þ4.48AT/À9%SP, þ5.28AT/ þ12%SP and þ7.08AT/À29%SP to refer to the climate change projections.In addition to these climate change scenarios we used a baseline climate scenario that we constructed by sampling years from the historical  record.We started all simulations from the year 1900, simulated through 2009 using the historical climate and then used the climate change and baseline scenario data to simulate through 2099.To assess vegetation development under end-ofcentury conditions we extended the simulations through 2199 using random samples from 2070-2099 temperature and precipitation anomalies based on the zero-order approximation that climate will stabilize towards the end of this century.
To determine the effect of bark beetles, we ran all climate change scenarios with and without bark beetle disturbance, but included fire and wind disturbance in all simulations.Below we focus on this set of simulations.To control for the effect of fire on forest succession under climate change, we ran additional simulations with fire switched off but including all climate scenarios and beetle and wind disturbance.We initialized the scenarios with the simulation data that were generated to evaluate the simulated distribution of tree species.To represent conditions in 1900, we used the data simulated 40 years after the simulated large fires (Fig. 1C).To account for the stochasticity in climate scenarios, fire ignitions and wind events, we replicated all simulation runs 20 times.
We aggregated all simulation results to low (2200-2800 m asl) and high (2800-3340 m asl) elevation bands reflecting the current transition to spruce-fir forests.We depict forest dynamics as developments in tree species-specific biomass and stem numbers aggregated over all replicates.Spruce beetle disturbance dynamics are depicted by the development of beetle-killed Engelmann spruce biomass over time.To reveal the mechanisms of beetle disturbance dynamics, we show the developments in beetle voltinism (i.e., the average number of generations spruce beetles may realize per year; Appendix A), droughtinduced susceptibility (S dr ), total susceptibility (S cell ) and total spruce biomass.
We isolated the effects of spruce beetle Negative and positive values for this interaction effect mean that climate change enhances or dampens, respectively, bark beetle disturbance as compared to historical conditions.If, for example, climate change enhances beetle-induced spruce mortality due to higher temperatures and drought, spruce biomass is lower than would be expected from the additive effects of beetle disturbance and climate change.Vice versa, in a situation where climate change dampens beetle disturbance due to increased drought-induced spruce mortality that in turn deprives the beetles of breeding material, the remaining spruce biomass will be higher than expected from the additive beetle and climate change effects.

Model evaluation
Vegetation patterns simulated to represent 140 years of post-fire forest development support the use of LandClim to reconstruct and project vegetation dynamics in response to climate and disturbance drivers in subalpine forests of Colorado (cf.Schumacher et al. 2006).Consistent with the distribution of existing vegetation types in LANDFIRE (Fig. 1B, D), simulated forests were dominated by spruce and fir at higher elevations .2800m asl.Lower elevations supported young post-fire aspen, lodgepole pine, and Douglas-fir forest with Douglas-firs becom-ing increasingly dominant as they grew larger.
Simulations driven by the averaged historical (1900-2009) climate of northwestern Colorado generally agree with documentary and tree ringbased evidence of northwestern Colorado's beetle outbreak history (Fig. 2A; Fig. 2 in Hart et al. 2014a).LandClim simulations reproduced the sharp increase in spruce beetle activity in the early 2000s and the increased spruce beetle activity in the 1930s and the 1950s.However, large outbreaks documented in the 1940s were not reproduced by LandClim.Possible explanations are that local climatic factors and factors other than climate, which have not been adequately represented by our generic simulations, were important in triggering the 1940s outbreaks.Schmid and Frye (1977), for example, noted that all past major outbreaks (including the 1940s outbreak) have been initiated by stand disturbances such as wind throw and logging and the ensuing increased abundances of susceptible coarse woody debris.Nevertheless, we concluded from this test that the LandClim beetle model adequately captured the general historical record of bark beetle outbreaks and its climate-driven trend, and can thus be used to assess beetle disturbance dynamics in response to climate variability.
In our assessment of LandClim's ability to represent expected patterns of long-term disturbance interactions we found that the probability of beetle infestation is reduced for respectively ca. 140, 100 and 120 years in previously beetle, fire and wind disturbed cells (Fig. 2B).In contrast to cells previously disturbed by fire and wind, the probability of re-disturbance in cells previously disturbed by beetles is elevated for one decade.This simulation result reflects the aggregation behavior of beetles (Byers 1989) and the fact that spruce beetles may take several years to attack and colonize a susceptible stand.Our assessment of beetle pressure accounts for this behavior by including the beetle killed biomass in the previous decade as a contributing factor (Eq. 3 in Temperli et al. 2013, Simard et al. 2011).The relatively low percentage of cells disturbed by beetles following fire reflects the relatively low area proportion disturbed by fire in each decade, which in turn results from the low fire spread probabilities and fuel flammability under the climate averaged over the northwestern v www.esajournals.orgColorado spruce-fir zone (Schumacher et al. 2006, Hart et al. 2014a).The 100-140 year time frame of reduced susceptibility to beetles following all disturbance types reflects the time required for spruce to grow to susceptible sizes (e.g., 0.9 S dbh at 36 cm spruce DBH; Appendix A: Fig. A3).This simulated 100-140 year time-lag corresponds with previous findings from tree ring reconstructions that suggest the susceptibility to spruce beetle is reduced for at least 70 years following stand-replacing fire (Veblen et al. 1994, Kulakowski et al. 2003, Kulakowski and Veblen 2006).Re-disturbance percentages and time-lags are highly contingent on the severity of previous disturbance (also see the discussion below on the possible underestimation of fire-severity simulated by LandClim), the spatial scale under consideration (forest patch vs. stand vs. landscape) and the local climate that influences spruce regrowth, which complicates comparisons across case studies and between simulated and empirical data.However, the important result of this test is that the implemented short-term relationships between forest state and forest susceptibility to beetles play out as the expected pattern of reduced spruce beetle disturbance following previous disturbances.Hence, this test confirms the suitability of LandClim to assess patterns of long-term, forest-mediated disturbance interactions.

Vegetation dynamics and wildfire
Simulations under climate change project drastic shifts in species composition in comparison to simulations under historical climate (Fig. 3).The simulations under historical climate show Douglas-fir becoming increasingly dominant in the initially aspen and lodgepole pine dominated forests at lower elevation (,2800 m asl).This aspen-lodgepole pine forest was severely impacted by a simulated fire in the 1910s that reduced forest biomass to ca. 2/3 of the initialized biomass.The spruce-fir dominated forest at higher elevations (.2800 m asl) experienced only limited fire in the 1910s and remained approximately constant with respect to biomass and stem numbers.The total tree density at the end of the simulation period at high elevation was projected to be 2135 stems/ha with 6% of spruce and 2% of fir trees .25 cm DBH and with these larger trees contributing 53% and 27% to total spruce and fir biomass, respectively.
The lower elevation mixed aspen-conifer forest is projected to be replaced and dominated by the more drought-tolerant ponderosa pine and Douglas-fir by the year 2100 under all climate change scenarios.These species may realize faster growth rates with warmer temperatures and may out-compete similarly sized spruce and fir trees (Appendix A: Table A1).Aspen, spruce, fir, and lodgepole pine retain small portions of v www.esajournals.orgforest biomass and stem count under the moderate þ4.48AT/À9%SP and intermediate þ5.28AT/þ12%SP climate scenarios.However, if a drastic precipitation decrease is projected (þ7.08AT/À29%SP), aspen, spruce, fir and lodgepole pine are eliminated on shallow soils and at elevations ,2800 m asl.Douglas-fir and ponderosa pine are projected to become part of the species mixture at elevations .2800m asl towards the end of this century and by the end of the simulation (year 2199) these species start to replace spruce and fir.The magnitude of this projected species shift increases with projected temperature change.Ponderosa pine stem counts accrue to 30% (188 stems/ha) by 2199 at high elevations and under extreme climate change (þ7.08AT/À29%SP), while biomass shares rise up to 71% (207 Mg/ha).These simulated ponderosa pines grow to relatively large sizes with ponderosa pines .40cm DBH comprising 66% of total forest biomass.
Fire activity increased under climate change (Appendix B: Fig. B4).The yearly percent area burned at lower elevations averaged across 20 replicates over the 2100-2199 period was 0.032% under the historical climate and increased to 1.6%, 4.7% and 1.8% under the moderate þ4.48AT/À9%SP, the intermediate þ5.28AT/ þ12%SP and the extreme þ7.08AT/À29%SP scenario, respectively.The yearly percent area burned in the 2100-2199 period at higher elevations .2800m asl was 0.026% under the historical climate and 0.6%, 3.0% and 0.25% under the three climate scenarios, respectively.Biomass burned developed similarly, with the highest increase being at lower elevations and under the þ5.28AT/þ12%SP scenario.Under this climate scenario inter-annual temperature and precipitation variability was highest resulting in the most extreme drought years (Appendix B Figs. B1-B3).Increased fire-related tree mortality favored tree regeneration due to decreasing Fig. 3. Development of species-specific live biomass (left) and stem count (right) at low ( 2800 m asl) and high (.2800 m asl) elevation and under four climate scenarios that are labeled at the right side of panels (see also Table 1).Biomass and stem counts were averaged over 20 replicate simulations and refer to simulations including beetle, fire and wind disturbance.
v www.esajournals.orgcompetition for light.In comparison to simulations that did not include fire (Appendix B: Figs.B5 and B6), this resulted in higher stem densities and a ca.20-30 year earlier establishment of Douglas-fir and ponderosa pine in formerly spruce and fir dominated areas.

Drivers of spruce beetle disturbance
Historical temperatures were high enough for beetles to realize a complete generation per year at elevations ,2800 m asl in most decades (Fig. 4A).At elevations .2800m asl between 0.60 and 0.72 generations were completed per year (voltinism) in the first half of the 20th century, indicating that a two-year life cycle was predominant.Thereafter, voltinism gradually increased to 0.81 in the 2000s and reached 1 in the first half of this century under all climate scenarios, meaning that beetles reproduced within one year.These seemingly high voltinism estimates are averages over the two elevation bands (mean elevation: 2571 and 3077 m asl, respectively) and thus represent the lower and warmer end of the historical spruce-fir zone and potential spruce beetle habitat in Colorado (2850-3500 m asl; Peet 1981).Drought-induced host susceptibility to bark beetle infestation increased under all climate scenarios with the highest increase under the extreme scenario þ7.08AT/À29%SP.The variation among replicates in estimates of drought-induced susceptibility was relatively large, which highlights the sensitivity of drought-induced susceptibility to random differences in the climate time series.
The increases in temperature-and droughtrelated drivers of beetle disturbance were counteracted by developments in forest composition and structure (Fig. 4A).Spruce biomass started to decrease early in the 21st century at low elevations under all climate change scenarios and at higher elevations under the two more severe scenarios (þ5.28AT/þ12%SP and þ7.08AT/ À29%SP).These decreases in spruce size-and spruce share-induced susceptibility overrode the outbreak-promoting effects of increased drought and caused total forest susceptibility to decrease in the long-term.

Development of spruce beetle disturbance
The simulated beetle-induced spruce mortality integrated the developments in the driving factors of beetle disturbance (Fig. 4B).Beetlekilled spruce biomass averaged over the 20 replicates fluctuated between 0.03 and 0.44 Mg ha À1 yr À1 at elevations ,2800 m asl and under historical climate.The maximum simulated beetle-killed biomass realized by one of the 20 replicates under these conditions was 1.3 Mg ha À1 year À1 .Climate change decreased the abundance of susceptible spruce at these elevations, which reduced beetle disturbance to nearly zero by the end of this century.Beetle disturbance at elevations .2800m asl gradually increased during the historical period and peaked at 0.85 Mg ha À1 year À1 between 2000 and 2009.Assuming a continuation of the historical climate, average beetle-killed biomass fluctuated around 0.30 Mg ha À1 year À1 for the remainder of the simulation period.Under moderate climate change (þ4.48AT/À9%SP) beetle-killed biomass remained at the level of the 2000-2009-decade for another two decades and then fluctuated between 0.25 and 0.50 Mg ha À1 year À1 , i.e., slightly higher than under historical climate, for the rest of the simulation period.If both increasing temperatures and precipitation were simulated (þ5.28AT/þ12%SP), spruce biomass remained high (Fig. 4A) and supported high beetle disturbance until the middle of this century.Thereafter, beetle-induced spruce mortality decreased to 0.12 Mg ha À1 year À1 , i.e., to 36% of the average of the historical period (0.34 Mg ha À1 year À1 ).Under extreme climate change (þ7.08AT/À29%SP) beetle-killed biomass peaked at .1.25 Mg ha À1 year À1 in the 2020s and then decreased below the historical period by 2070 reaching 6% of the historical average by the end of the simulation period.In sum, under all climate scenarios and elevations beetle-induced spruce mortality decreased compared to the historical period in the long term, except for the high elevations under moderate climate change (þ4.48AT/À9%SP),where beetle disturbance was slightly higher and more variable than under historical conditions.

Net effects of climate change on spruce beetle disturbance
Climate change dampened the effect of beetle disturbance on Engelmann spruce biomass under most conditions, because the combined effect of climate change and beetle disturbance decreased v www.esajournals.orgthe availability of susceptible spruce in the long term (Fig. 5).Comparing spruce development without beetle disturbance between historical and climate change conditions isolated the effect of climate change (Appendix B: Fig. B7).Climate change resulted in a long-term decrease in spruce biomass ,2800 m asl under all climate scenarios.However, this decrease did not start until year 2050 if summer precipitation was projected to increase under the þ5.28AT/þ12%SP scenario.Above .2800m asl climate change even enhanced spruce growth under this scenario until 2070.Thereafter spruce biomass started to decline and reached values below the baseline scenario by the end of the simulation period (60.5 vs. 128.5Mg/ha, respectively).With a moderate (þ4.48AT/À9%SP) reduction in summer precipitation higher elevation spruce biomass remained approximately constant, whereas severe climate change (þ7.08AT/À29%SP) reduced spruce biomass at high elevations at the beginning of this century.
Including spruce beetle disturbance in simulations under historical climate reduced spruce v www.esajournals.orgbiomass at lower elevations by 35.0 Mg/ha (49.3%) by the end of the simulation period (Fig. 5).At higher elevations this beetle effect was even greater with a 60.0 Mg/ha (46.7%) spruce biomass reduction.
The combined effect of beetle disturbance and climate change on spruce biomass was stronger than the effects of either climate change or beetle disturbance individually under all climate scenarios and at all elevations (Fig. 5).However, beetle disturbance and climate change did not affect spruce biomass in a simple additive way.The interaction between climate change and beetle disturbance was negative at elevations ,2800 m asl under all climate change scenarios, and thus had a positive, stabilizing effect on spruce biomass.This important finding illustrates that the combined effect of climate change and beetle disturbance on spruce biomass was less than the simple additive effect of climate change and beetle disturbance.The underlying mechanism for this negative effect of climate change on beetle-induced tree mortality is the depletion of spruce due to both climate change and beetle disturbance.In contrast, climate change and beetle disturbance interacted synergistically at higher elevations .2800m asl, under the relatively moderate þ4.48AT/À9%SP scenario over the whole simulation period and under the more severe þ5.28AT/þ12%SP and þ7.08AT/ À29%SP scenarios until ca.2070.Climate change intensified beetle disturbance under these conditions such that the climate change-beetle interaction effect led to a further reduction in spruce biomass.The dominant mechanism for this synergistic, reinforcing interaction is the positive effect of climate change on beetle brood development and drought-induced susceptibility of spruce to beetles.Beetles and climate change began to interact negatively also .2800m asl.after ca.2070 and under more severe climate change (þ5.28AT/þ12%SP and þ7.08AT/À29%SP) when the depletion of susceptible spruce became increasingly important.While individual replicates of the beetle and climate change effects, i.e., spruce biomass differences, showed large departures from the mean, the interquartile ranges suggest robust trends in these effects (Fig. 5).

DISCUSSION
Assessment of future bark beetle disturbance dynamics requires integration of at least three mechanisms by which climate change acts upon bark beetles (Temperli et al. 2013): (1) changes in temperature directly affect beetle phenology and population development (Hansen et al. 2001, Bentz et al. 2010), (2) changes in drought affect the defense mechanisms of the host trees (McDowell et al. 2011, Hart et al. 2014a), and (3) long-term (decades to centuries) changes in climate and disturbance regimes affect forest species composition and stand structure (Rehfeldt et al. 2006, Turner 2010), thereby acting v www.esajournals.orgupon beetle disturbance dynamics indirectly through changes in the availability of spruce hosts (Økland andBerryman 2004, Temperli et al. 2013).We used a simulation modeling approach and integrated these three mechanisms to provide novel insights into their relative importance in driving the dynamics of spruce beetle disturbance under climate change in a case study watershed representative of Colorado's subalpine forests.
We found that with increasing magnitude and advancement of climate change, drought-and spruce beetle-induced spruce mortality and concomitant shifts in species composition decrease the availability of the beetles' breeding material.The reduction in spruce abundance is important enough that in the long term and under more severe climate scenarios spruce beetle disturbance decreases below historical values even though increasing temperatures favor beetle population growth and drought is projected to increase host tree susceptibility to beetles.
This finding is an important addition to previous studies that projected the potential for future spruce beetle disturbance over large regions based on climate and static forest habitat variables (Bentz et al. 2010, DeRose et al. 2013).These studies did not take into account shifts in the distribution and structure of spruce forests or the effects of future beetle or fire disturbance on forest structure and composition.While conducted at a much smaller scale (watershed vs. western U.S. and Canada) and using different realizations of the IPCC AR4 A2 emission scenario, our projections of spruce beetle voltinism are consistent with those of Bentz et al. (2010): by the end of this century we expect a high probability of univoltine rather than semivoltine life cycles in most of the spruce-fir zone.Also, our assessment confirms DeRose et al.'s (2013) predictions of the expansion of suitable spruce beetle habitat due to projected temperature increases.
The importance of forest composition and structure in determining spruce beetle disturbance in our simulations corroborates DeRose et al.'s (2013) finding that spruce beetle habitat variables such as spruce basal area are more important than temperature variables in modeling spruce beetle's realized environmental niche.This empirically-based finding is reflected in the design of our simulation model that gives high weight to habitat variables in assessing forest susceptibility to beetles, beetle population pressure and the resulting cell-specific infestation risk (Appendix A: Eq.A9; Eq. 4 in Temperli et al. 2013).Since beetle habitat variables are dynamically coupled with climate and beetle disturbance in our model, they shift with time, unlike the static niche model of DeRose et al. (2013).This dynamic coupling of climate, beetle phenology and beetle habitat variables, as well as negative feedbacks from future disturbance by fire and bark beetles are the key differences between our approach and most previous assessments (e.g., Bentz at al. 2010, DeRose et al. 2013).Our results suggest that these dynamic feedbacks are pivotal in understanding spruce beetle disturbance dynamics under climate change as their explicit consideration reverses the expectation of increased beetle disturbance.
Recent projections of shifts in forest composition and structure and therefore in the spruce beetle's habitat under climate change generally agree in direction.The shift from spruce-fir to mixed conifer forest that we simulated under end-of-century climate is consistent with climate envelope and process models that predict upward elevational shifts in the distribution ranges of ponderosa pine, Douglas fir and Engelmann spruce (Rehfeldt et al. 2006, Crookston et al. 2010, Notaro et al. 2012).These vegetation shifts suggest that climatic conditions for Engelmann spruce will become unsuitable in much of its current geographic range.Whether spruce-fir forests and spruce beetles may expand beyond current tree-lines and elevations that our study landscape covers is difficult to project and likely depends on geomorphic and non-climatic factors and whether higher precipitation at higher elevations may compensate for increased moisture stress due to warmer temperatures (Villalba et al. 1994, Hu et al. 2010, Macias-Fauria and Johnson 2013).
Our projected vegetation shifts should be interpreted cautiously given the study's limitations.The simulated increases of Douglas-fir and ponderosa pine result first from warmer projected temperatures and concomitant increases in fire and spruce beetle disturbance.Higher temperatures and increased light-availability at the forest floor promote Douglas-fir and ponderosa pine establishment and growth.Second, Douglas-fir and ponderosa pine were parameterized to attain larger maximum tree size (i.e., biomass) than Engelmann spruce or subalpine fir (Appendix A: Table A1).While the improved establishment and growth conditions for Douglas-fir and ponderosa pine at higher elevations under warming temperatures is well supported (Burns andHonkala 1990, Rehfeldt et al. 2006), we may have overestimated Douglas-fir and ponderosa pine tree-size.Notably, the present version of LandClim does not account for tree mortality due to mountain pine beetle and Douglas-fir beetle (Dendroctonous pseudotsugae Hopkins).Higher mortality rates due to beetles may prevent Douglas-fir and ponderosa pine from attaining the large sizes that we simulated under climate change.
Additionally, we may have underestimated increased fire-induced tree mortality.Even though our finding of increased fire activity agrees with other future projections (Appendix B: Fig. B4; Westerling et al. 2011), the ability of large numbers of Douglas-fir and ponderosa pine trees to reach large sizes under future fire regimes is difficult to validate.Two factors may have led us to overestimate their survival and growth.First, Douglas-fir and ponderosa pine were parameterized to be less fire-sensitive than smaller fir and spruce trees (Appendix A: Table A1; Schumacher et al. 2006) and thus experienced less fireinduced mortality.Second, the LandClim fire module does not consider increased fire spread rates under crown fire conditions that may occur with high fuel loads.Hence, we may have underestimated large, high-severity stand replacing fires and concomitant mortality of large trees (Schumacher et al. 2006, Cruz andAlexander 2010).
The underestimation of Douglas-fir and ponderosa pine mortality due to beetles and large stand-replacing fires may have two contrasting implications for the simulated spruce abundance under climate change.First, beetle-induced mortality of Douglas-fir and ponderosa pine may lead to reduced competition from these species and higher resource availability for Engelmann spruce.This in turn could result in a slower spruce decline then our simulations suggest under climate change.In contrast, increased crown-fire activity would lead to increased mortality in all tree species, which in turn would accelerate the decline of spruce.In sum, projections of future forest composition and structure are still afflicted with large uncertainties related to the choice of climate change scenarios and tree species and disturbance responses to climate change (e.g., Xu et al. 2009), but they largely agree in the decline of the area suitable for future Engelmann spruce growth and thus a decline in spruce beetle habitat.
Whether the declining abundance of large spruce due to climate change and beetle disturbance offsets the effects of warming on brood development and the increased drought-induced spruce susceptibility to beetles depends on drought severity.We demonstrated that climate change enhances or dampens beetle disturbance depending on whether or not drought surpasses a threshold.At this threshold, drought not only results in high susceptibility of Engelmann spruce to beetles but also leads to drought stress-induced mortality.In our model this threshold is reached locally, i.e., in a simulation cell, when drought-stress reduces the Engelmann spruce growth rate for at least 3 consecutive years to ,10% of its potential maximum under optimal, stress-free environmental conditions (Schumacher et al. 2004).Such conditions are met by a LandClim drought index value of 0.347, which corresponds to a summer vapor pressure deficit of 8.567 mbar (Appendix A: Fig. A2).
Both this study in Colorado and Temperli et al. (2013) in Central Europe showed that the range of drought conditions where host-tree susceptibility to beetles is high but not yet lethal to the host trees is relatively narrow.In our model this narrow range is determined by the two relationships that describe drought-induced susceptibility to beetles and the drought-dependent growth response and thus mortality of the host tree species (Appendix A: Fig. A2).These two relationships implicitly account for a complex set of interdependent mechanisms that determine photosynthesis, tree growth, pools of non-structural carbon, water transport, mechanisms to defend against beetles and many other biotic pathogens and ultimately tree mortality (Bigler et al. 2006, McDowell et al. 2011).Further research on the nature and relative importance these mechanisms is critical to more accurately disenv www.esajournals.orgtangle and model ecosystem-specific timing and severity of drought-and beetle-induced mortality under climate change (e.g., Jo ¨nsson et al. 2012).While our simulations showed that drought may mask the importance of blowdown under climate change conditions (cf.Temperli et al. 2013), a priority for further research is also to corroborate this finding empirically by more accurately quantifying the blowdown effect on forest susceptibility to spruce beetles (Jenkins et al. 2014).
We validated the sensitivity of the LandClim spruce beetle model to variations in the climate of the past century.To this end we compared the output of LandClim simulations on a generic, flat 1 km 2 landscape to Hart et al.'s (2014a) regional spruce beetle disturbance reconstructions.That procedure, of course, limits the validation to broad-scale and long-term climatic influences.A more spatially precise validation including the influences of local topography and climate, soil properties and disturbance history at the same spatial scale and resolution as the LandClim model would have been desirable.However, the necessary reconstructions of past local climate, forest succession and disturbance history for multiple watersheds in northwestern Colorado are currently not available.Such data would be extremely valuable, not only for model validation, but also to better understand the mechanisms that drive bark beetle disturbances.
The challenges to validate LandClim simulations are counterbalanced with LandClim's relatively generic approach to integrating climate drivers, forest succession and disturbances.This generality facilitates the identification of ecological mechanisms that operate over wide environmental gradients and across ecosystem boundaries.Hence, LandClim may be applied to other conceptually similar beetle-host systems to test, for example, whether host tree depletion may also be important in determining the longterm dynamics in mountain pine beetle and Douglas-fir beetle disturbance under climate change (Aukema et al. 2008, Raffa et al. 2008).These and further LandClim applications may inform anticipatory adaptive management that needs to account for projections and uncertainties of tree species-specific growth conditions and disturbance regimes (Millar et al. 2007, Temperli et al. 2012, Six et al. 2014).

CONCLUSIONS
We adapted the climate-sensitive forest landscape model LandClim and further developed the model's representation of spruce beetle (Dendroctonus rufipennis) disturbance dynamics in Colorado's subalpine forests.The resulting model advances previous work in that it explicitly accounts for changes in forest development due to climate change and feedbacks from fire and beetle disturbance (Bentz et al. 2010, DeRose et al. 2013).It integrates the climate and forest structural drivers of spruce beetle disturbance from tree to landscape scales and from years to centuries to identify emergent patterns in forest landscape and disturbance dynamics.
Our simulations under three climate change scenarios corroborate previous assessments that project future climate to become increasingly conducive to spruce beetle outbreaks.Concomitantly, we project an increase in Engelmann spruce mortality due to both drought and spruce beetle disturbance, which is likely to shift tree species dominance to more drought-adapted Douglas-fir and ponderosa pines.This decrease in host tree availability offsets the positive climate change effects on beetle population growth to the degree that beetle disturbance decreases in the long term at low elevations (,2800 m asl) and also at higher elevations under more severe climate change scenarios.Whether climate change enhances or dampens spruce beetle disturbance depends on whether drought surpasses a threshold, above which not only stress-induced susceptibility to beetles but also drought-induced host tree mortality is increased.Further research is necessary to disentangle the complex interactions among drought-stress, bark beetles and other pathogens that ultimately cause tree mortality.We suggest that host tree depletion due to drought-and beetle-induced tree mortality may also be important for disturbances by other climate-sensitive bark beetle species (e.g., mountain pine beetle and Douglas-fir beetle).Incorporating these beetle species into the scenarios investigated in this study is thus a high priority for future work.Further applications of climate-sensitive forest landscape models should focus on the role of wildfire and wind disturbances, test varying beetle and fire mitigation treatments and assess their impact on forest ecosystem service provision.Climate impact assessments such as the here depicted climatedriven shifts in forest and disturbance dynamics need to be considered in long-term adaptive management.

Fig. 1 .
Fig. 1.Study area: (A) Topography, longitude and latitude of Buffalo Pass case study watershed.The inset map shows the location (black dot) of the case study watershed on the western slope of the continental divide (grey line) in northern Colorado (CO); (B) Existing vegetation types (EVT) per LANDFIRE.Spruce-fir refers to both dry-mesic and mesic original EVT, Aspen-Mixed conifer includes all aspen, aspen-mixed conifer and mixed conifer EVT, Lodgepole refers to the EVT ''Rocky Mountain Lodgepole Pine Forest'', and other refers to EVTs that occupied ,20 cells and include limber, ponderosa and pinyon pine-juniper.Not forested refers to all grass, shrub and barren lands; (C) Dominant species (highest biomass share) simulated by LandClim 40 years after a simulated fire that burned most of the area below ca.2900 m asl.This represents the data that were used to initialize scenario runs starting from year 1900; (D) LandClim simulated dominant species 140 years post-fire representing possible current forest vegetation.Note that ponderosa pine was included in the simulations but 140 years after fire is only dominant in two cells at the western, low-elevation tip of the case study area.

v
www.esajournals.orgdisturbance and climate change (i.e., the combined effect of a change in temperature and drought) to reveal their relative importance in determining the development of Engelmann spruce host biomass.FollowingTemperli et al. (2013), we calculated the effect of beetle disturbance and climate change as the spruce biomass difference of simulations that included beetle and climate change, respectively, to reference simulations that included neither beetles nor climate change.Accordingly, the combined effect of climate change and beetle disturbance was assessed by the spruce biomass difference between simulations that included both beetles and climate change and the reference simulations.To measure how beetle disturbance and climate change interact on spruce biomass, we subtracted the beetle and climate change effects individually from the combined beetle-climate change effect.

Fig. 2 .
Fig. 2. Model evaluation.(A) Boxplots showing the distribution of 20 replicates of simulated beetle-killed Engelmann spruce biomass by decade.(B) Temporal development of the proportion of cells that have been redisturbed by beetles after beetle (top), fire (middle) and wind (bottom) disturbance.

Fig. 4 .
Fig. 4. Panel (A) shows the development of beetle model parameters by elevation (low: 2800, high: .2800m asl) and climate scenario.Voltinism (range: 0.5-1) indicates the temperature-driven number of generations the spruce beetle population can realize per year.Spruce biomass is the live standing biomass of Engelmann spruce on the landscape in 100 Mg/ha.Total susceptibility is an index (range: 0-1) describing the susceptibility induced by the combined influence of spruce size, spruce abundance and drought-stress in spruce; susceptibility drought indicates the drought component of total susceptibility.Light shaded uncertainty ribbons indicate minima and maxima and darker shaded ribbons lower and upper quartiles of 20 replicate simulations.(B) Development of beetle-killed Engelmann spruce biomass by elevation (low: 2800, high: .2800m asl) and climate scenario.The black line indicates the mean, light shaded uncertainty ribbons indicate minima and maxima and darker shaded ribbons lower and upper quartiles of 20 replicate simulations.

Fig. 5 .
Fig. 5. Effect of beetles, climate change, the combined effect of beetles and climate change (Beetle þ climate change) and the interaction effect of beetle and climate change (Beetle 3 climate change).Effects are expressed in differences in Engelmann spruce biomass to baseline simulations that do not include beetles and climate change.Light shaded uncertainty ribbons indicate minima and maxima and darker shaded ribbons lower and upper quartiles of 20 replicate simulations.