spruce beetle outbreak: a new study from cu-boulder
TRANSCRIPT
Sarah J. Hart et al.
1
Running Head: Drought and spruce beetle outbreaks
Title: Drought induces spruce beetle (Dendroctonus rufipennis) outbreaks across northwestern
Colorado
Authors: Sarah J. Hart1, Thomas T. Veblen1, Karen S. Eisenhart2, Daniel Jarvis3, Dominik
Kulakowski3
1 - Department of Geography, University of Colorado, Boulder, CO
2 - Department of Geosciences, Edinboro University of Pennsylvania, Edinboro, PA
3 - School of Geography, Clark University, Worcester, MA
Sarah J. Hart et al.
2
ABSTRACT
This study examines influences of climate variability on spruce beetle (Dendroctonus
rufipennis) outbreak across NW Colorado during the CE 1650–2011 period. Periods of broad-
scale outbreak reconstructed using documentary records and tree-rings were dated to 1843 to
1860, 1882-1889, 1931-1957, and 2004-2010. Periods of outbreak were compared with seasonal
temperature, precipitation, vapor pressure deficit (VPD), the Palmer Drought Severity Index
(PDSI), and indices of ocean-atmosphere oscillation that include the El Niño Southern
Oscillation (ENSO), Pacific Decadal Oscillation (PDO), and Atlantic Multidecadal Oscillation
(AMO). Classification trees showed that outbreaks can be predicted most successfully from
above average annual AMO values and above average summer VPD values, indicators of
drought across Colorado. Notably, we find that spruce beetle outbreaks appear to be predicted
best by interannual to multidecadal variability in drought, not by temperature alone. This finding
may imply that spruce beetle outbreaks are triggered by decreases in host tree defenses, which
are hypothesized to occur with drought stress. Given the persistence of the AMO, the shift to a
positive AMO phase in the late 1990s is likely to promote continued spruce beetle disturbance.
Keywords: bark beetle, disturbance, climate, Atlantic Multidecadal Oscillation, tree ring
INTRODUCTION
Over the past 30 years, severe and extensive bark beetle outbreaks have caused dramatic
tree mortality from Alaska to the Southwestern US (Berg et al. 2006, Bentz et al. 2009). Broad-
scale tree mortality has been linked to changes in regional carbon dynamics and thus may
feedback into future global climate change (Kurz et al. 2008). Severe bark beetle outbreaks are
dependent upon the presence of a large population of mature host trees, often determined by
history of natural disturbances or past land-use practices (Schmid and Frye 1977, Veblen et al.
Sarah J. Hart et al.
3
1994, Safranyik and Carroll 2007). Given a susceptible landscape, outbreaks are often incited
by events that either decrease tree defenses, including drought or pathogens (Christiansen et al.
1987), or positively influence beetle population growth, including periods of warm temperature
(Werner and Holsten 1983), windthrow (Schmid 1981), or decreased predation (Berryman
1982). Weather can thus have both direct and indirect effects on beetle population success at
multiple points throughout the course of an outbreak (Raffa et al. 2008). While many studies
have illustrated the importance of weather variability for beetle population dynamics
(McCambridge and Knight 1972, Bentz et al. 1991, Logan and Bentz 1999, Hansen et al. 2001,
Hansen and Bentz 2003), few studies have examined the association between infestation and
climate across multiple outbreaks (Campbell et al. 2007, Hebertson and Jenkins 2008, Sherriff
et al. 2011). Yet future predictions of outbreaks will require a better understanding of how
changing climatic processes affect outbreaks through combined effects on both tree defenses
and beetle populations.
Spruce beetle (Dendroctonus rufipennnis) is one of the most destructive forest insects in
North America, where it can lead to mortality of greater than 90% of the mature spruce within a
stand (Hopkins 1909, DeRose and Long 2007). In the Pacific Northwest and Rocky Mountains,
the spruce beetle is found in high elevation spruce-fir forests where it predominantly feeds upon
Engelmann spruce (Picea engelmannii Parry ex Engelm.) (Schmid and Frye 1977). The spruce
beetle inhabits the inner bark and feeds on the tree’s phloem tissues. Heavy colonization and
reproduction within the inner bark interrupts the flow of water and nutrients throughout the tree
and can cause tree death. Endemic populations typically live in weakened trees. Outbreaks
occur as beetle populations grow and start attacking apparently healthy spruce (Massey and
Wygant 1954, Schmid et al. 1977). Conifer defense against bark beetles includes resin flow,
Sarah J. Hart et al.
4
which can create a physical barrier, the formation of necrotic tissues, which deprive beetles of
living tissues for food, and constitutive and induced chemicals, which are toxic to the beetles
and their eggs and inhibit fungal growth (Christiansen et al. 1987).
Previous research based on laboratory experiments has shown that warm temperatures
favor the growth of spruce beetle populations through direct effects on larval development and
survival (Hansen et al. 2001, Hansen and Bentz 2003). Spruce defense against beetle outbreaks
may also be mediated by weather (Hard 1985). Hot, dry weather is expected to decrease tree
defense (Mattson and Haack 1987), and there is evidence that the extensive spruce beetle
outbreak that initiated in the 1990s in Alaska may have been triggered by above-average
summer temperatures (Berg et al. 2006). Likewise, 20th-century spruce beetle outbreaks in
Colorado and Utah are associated with warm, dry years based on instrumental climate records
from 1906 to 1996 (Hebertson and Jenkins 2008). In an area of 800 km2 in southeastern Utah,
spruce beetle outbreaks in the 1990s also have been shown to be associated with higher
maximum summer temperatures and higher minimum winter temperatures (DeRose and Long
2012b). In the same study area, tree-ring reconstructions of spruce beetle outbreak were
associated with prolonged drought (DeRose and Long 2012a).
While local patterns of drought and temperature affect spruce beetle outbreaks, the
atmospheric mechanisms affecting drought and temperature may be global or regional in scale.
Indeed, spruce beetle outbreaks in Alaska have been linked to negative phases of the Pacific
Decadal Oscillation (PDO), which brings cooler winter temperatures and decreased winter
precipitation (Sheriff et al. 2011). However, the importance of hemispheric-scale ocean
atmosphere circulation patterns for spruce beetle outbreaks in the Rocky Mountains, where
drought is related to oscillations in both the Atlantic and Pacific Oceans (McCabe et al. 2008), is
Sarah J. Hart et al.
5
poorly understood. To contribute to this emerging understanding of how climate factors
influence the occurrence of spruce beetle outbreaks, in the current study we develop a 300+ year
history of outbreaks over a 60,000 km2 study area in Colorado in relation to records of both
persistent (annual to multidecadal ocean-atmosphere oscillations) and short-term (monthly and
seasonal temperature, precipitation and drought) climate.
This study relies on a multiproxy approach to reconstruct spruce beetle outbreak. Both
historical documents and tree-ring records were used to identify periods of spruce beetle
outbreak; this has advantages over other multidecadal spruce beetle-climate studies (e.g.
Hebertson and Jenkins 2008, Sherriff et al. 2011). While tree rings have been reliably used to
reconstruct spruce beetle outbreak in spruce-fir forests in Colorado (Veblen et al 1991), other
disturbances including windthrow and stand-replacing fire must be excluded. Typically,
coincident and widespread increased growth rates (known as releases) of subcanopy and nonhost
trees following outbreak are used to reconstruct periods of beetle outbreak, particularly in
combination with synchronous dates of host tree mortality (Veblen et al 1991). Release events
distinguish bark beetle outbreaks from defoliator outbreaks that result in periods of reduced
radial growth (Swetnam and Lynch 1989). Following spruce beetle outbreaks stands are
characterized by mixed ages and radial growth rates that are initially slow and then dramatically
increase. This can be distinguished from fire, which is characterized by even-aged populations
and rapid initial growth (Veblen et al. 1991). Spruce beetle outbreak can also be differentiated
from windthrow, which typically results in a uniform orientation of fallen logs, multiple species
of fallen logs, and uprooting (Veblen et al. 2001).
The overall aim of this paper is to determine the history and synchrony of spruce beetle
outbreaks from 1650-2010 in NW Colorado and to examine the climatic conditions under which
Sarah J. Hart et al.
6
these outbreaks occur. We employ a novel technique combining historical documentary records
and two types of tree-ring data, tree mortality and growth release dates, to more accurately
reconstruct widespread and severe spruce beetle outbreak. This record is then used to examine
the influence of interannual to multidecadal variability in temperature, precipitation, and drought
on the occurrence of spruce beetle outbreak.
STUDY AREA
We analyzed documentary and 18 tree-ring records of spruce beetle outbreak across NW
Colorado (Fig. 1; Appendix A: Table A1). Spruce-fir forest is found in Colorado’s subalpine
zone (~2750 to 3350 m; Peet 2000), which is characterized by cold, wet winters and warm, dry
summers (Ray et al. 2008). Peak precipitation in NW Colorado typically occurs in April or May
and is followed by a period of reduced precipitation extending into early fall (Ray et al. 2008).
Drought in Colorado has well documented relationships with ocean-atmosphere
oscillations. The El Niño Southern Oscillation (ENSO) is a complex ocean-atmosphere
interaction that causes cyclical warming and cooling of sea surface temperatures in the tropical
Pacific Ocean on a time scale of 5-7 years (Ware 1995). During La Niña years, cooler-than-
average ocean temperatures are found across the equatorial eastern Pacific, often driving changes
in storm tracks that result in drought across NW Colorado (Ray et al. 2008). The PDO is an
ocean-atmosphere interaction in the North Pacific that describes ENSO-like variability at both
interdecadal and decadal scales (Newman et al. 2003). During the negative phase of the PDO,
Colorado often experiences drought (McCabe et al. 2008). The Atlantic Multidecadal Oscillation
(AMO) is an ocean-atmosphere circulation that causes cyclical warming and cooling of sea
surface temperatures in the Atlantic Ocean on a time-scale of 50 to > 70 years (Gray et al. 2003).
Sarah J. Hart et al.
7
Positive phases of the AMO are associated with high temperatures and low precipitation,
resulting in drought across most of the United States (McCabe et al. 2008).
DATA AND METHODS
Dendroecological analyses
Time series of initiation years of spruce beetle outbreaks over the past 350 years were
constructed by processing previously collected tree-ring data (Fig. 1, Appendix A: Table A1). At
each site, the largest 20-80 live spruce and fir (Abies lasiocarpa) were sampled for releases and
at 13 of the 18 sites, the 10-20 largest dead spruce with evidence of spruce beetle infestation
(galleries, blue stain) were also sampled. In addition to the tree-ring samples collected for
reconstructing outbreaks, we obtained tree-ring datasets used in analyses of spruce radial growth
and climate from the International Tree-Ring Data Bank
(http://www.ncdc.noaa.gov/paleo/treering.html) (Fig. 1, Appendix A: Table. A1). All tree-ring
data were previously processed using standard dendrochronological methods (Stokes and Smiley
1968).
To create site-level time series of initiations of spruce beetle outbreaks, we classified
years of major growth releases, where the mean ring width of years t to t+9 was 200% or
greater than the mean ring width of years t-1 to t-10 (Veblen et al. 1991). The first ten years of
annual radial growth were trimmed from each series to eliminate the misclassification of rapid
initial (i.e. seedling establishment) tree growth as beetle outbreak. We also excluded the 10
years following the first year of each release event to ensure independence of release dates
(Berg et al. 2006, Sherriff et al. 2011). To create site histories of spruce beetle outbreak, we
then summed the number of trees exhibiting releases in each year over the time period where
the sample depth was at least 20 trees, the minimum number required for statistical testing.
Sarah J. Hart et al.
8
Release and mortality tree-ring dates are typically accurate within a few years (Veblen et al.
1991), thus we modeled the probability of observing a release event in five-year periods at each
site using a binomial model and determined significant (p < 0.05) release events (Berg et al.
2006, Sherriff et al. 2011). We then defined outbreak initiation as the earliest year within a
significant release event period that exhibited dramatic release.
Regional synchrony
Given that other disturbances may cause tree mortality that could cause a release in
surviving trees, the synchrony of beetle-related tree death dates and outbreak initiation dates
was evaluated across the 13 sites with both data types. We used a modified Ripley’s K function
called bivariate event analysis (BEA) using K1D software (Gavin unpublished). BEA identifies
the synchrony of events in one dimension (time) within a defined window (± t years) by
comparing the timing of events within the two records (for details see Gavin et al. 2006). To
determine if outbreaks are occurring synchronously across NW Colorado, we used the
multivariate extension of BEA, multivariate event analysis (MEA; Gavin unpublished).
Association of climate with spruce beetle outbreak
Multiproxy identification of outbreak periods. – First, we used reports in the scientific
literature, newspapers, and government reports (Table 1) to develop a time series of spruce beetle
outbreak initiation (1850 – 2011). In this documentary record of outbreaks, we defined the
initiation of a spruce beetle outbreak as the first time an outbreak was mentioned for a given
USFS forest administrative region. We also recorded the collapse date of each spruce beetle
outbreak, which was defined as the latest date each outbreak was reported as still ongoing.
Second, we used intervention analysis (Rodionov 2004) to test for statistically significant
shifts in the frequency of outbreaks in the combined documentary and tree-ring records of SB
Sarah J. Hart et al.
9
outbreaks. Periods of outbreak were determined from the number of tree-ring release events,
death dates, and documented outbreaks for the time period where at least 10 tree-ring sites were
recording. We used Rodionov’s Sequential Regime Shift Detection program, which computes
sequential t-tests to determine the timing of regime shifts (Rodionov 2006). We used a default
90% significance level and 10-year cut-off period to identify significant shifts. To eliminate the
importance of single events in producing shifts associated with spruce beetle outbreak, periods of
spruce beetle outbreak were conservatively classified as periods when the average number of
records per year was at least three.
Climate datasets. – To analyze climate and spruce beetle outbreaks over the instrumental
meteorological record we obtained 4 by 4 km grids of monthly minimum and maximum
temperature, total precipitation, and mean dew point temperature for the entire study area over
the time period from January 1900 to December 2011 (PRISM database;
www.prism.oregonstate.edu). A 30 by 30 m grid of spruce-fir forest cover presence for the study
area was then created from the USFS Landfire existing vegetation type data set
(www.landfire.gov). Mean monthly time series (1900 to 2011) of precipitation, minimum and
maximum temperature, and vapor pressure deficit (VPD), a measure of the evaporative demand
of the air, were then created using all PRISM grid squares within the study area’s spruce-fir
zone.
To examine the association of spruce beetle outbreaks with more slowly changing
climate, we also obtained instrumental and tree-ring based reconstructions of the Palmer Drought
Severity Index (PDSI), AMO, ENSO, and PDO. We obtained monthly instrumental PDSI values
for 1900 to 2011 from the US Climatological Divisions in our study area (the Colorado River
Drainage Basin and Platte River Drainage Basin) from the National Climate Data Center website
Sarah J. Hart et al.
10
(www.ncdc.noaa.gov). Data for the two climate divisions were then averaged to produce a time
series from 1900 to 2011. The Hadley Centre Sea Ice and Sea Surface Temperature (HadISST;
www.metoffice.gov.uk /hadobs/) monthly instrumental time series of ENSO, PDO, and AMO
indices from 1900 to 2011 were also obtained. We used tree-ring based reconstructions of AMO
(Gray et al. 2003), PDSI (Cook et al. 2004), and PDO (MacDonald and Case 2005), which were
obtained from the NOAA Paleoclimatology Program (www.ncdc.noaa.gov/paleo/recons.html).
Climate reconstructions were then extended to 2011 using the following procedures. First, we
scaled the mean of each climate reconstruction to the mean of the detrended instrumental record,
where the detrended values are the residuals from a linear regression of the index values versus
time during the 1950-2011 period. We then adjusted the standard deviation of the reconstructed
series to 1 and replaced the 1950-2011 period with scaled values from the instrumental record
(cf. Schoennagel et al. 2007).
Outbreak-climate analyses. - Periods of spruce beetle outbreak and non-outbreak
identified from the intervention model were compared with seasonal temperature, precipitation,
PDSI, and VPD data over the period of common overlap (1900-2011). A Mann-Whitney test was
used to statistically assess if mean climatic parameters in outbreak periods were different than
non-outbreak periods over the past 100 years. We used a conservative Bonferroni correction to
account for multiple comparisons (Gotelli and Ellison 2004). To assess if significant
relationships were stable through time we compared reconstructed outbreaks with tree-ring
reconstructions (1650-2011) of PDSI and ocean-atmosphere oscillations. No stability assessment
was done for monthly and seasonal climate because pre-1900 regional tree-ring reconstructions
of temperature, precipitation, or VPD do not exist for Colorado’s subalpine zone.
Sarah J. Hart et al.
11
A model of the probability of outbreak was then constructed using a Random Forests
(RF) framework using the package randomForest (Liaw and Wiener 2002) in R (R Development
Core Team 2010). The RF method builds on classification and regression tree methods, where
trees are constructed by repeatedly splitting the data into two mutually exclusive groups
(Breiman 2001). In RF analysis, many trees are fit to the data and then combined. RF provides
high classification accuracy and has been shown to identify ecologically meaningful
relationships (Cutler et al. 2007). We then constructed a classification tree from five variables
that explained the most variability in terms of the mean decrease in accuracy statistic (a measure
of how much inclusion of a variable reduces classification error).
RESULTS
Regional synchrony
Using the combined documentary and tree-ring records we were able to reconstruct a
total of 39 spruce beetle release events at 18 sites across NW Colorado between 1650 and 2011
(Fig. 2). Based on this 18-site record, the median number of years between outbreaks at a site
was 75 years. Most sites exhibited 1-2 outbreak events, while 2 sites exhibited 3 outbreaks. The
Ouzel Lake site in Rocky Mountain National Park exhibited 4 outbreak events. During the 1900s
the tree-ring evidence (releases and tree deaths) coincide with documentary records of outbreaks,
but also appear to indicate the incipient and collapse phases (Fig. 2). In contrast, outbreaks
during the 1800s show stronger peaks and more unimodal distributions of evidence (Fig. 2).
Documentary evidence of outbreaks during the 1800s did not have as precise information about
the start or collapse of outbreaks. This may contribute to the apparent lack of incipient and
collapse phases for these outbreaks. Evidence of outbreaks during the 1700s was limited to
Sarah J. Hart et al.
12
release events. Due to low sample sizes, outbreaks during the 1700s are typically represented as
a single event (Fig. 2).
Across the 13 sites that have mortality data, BEA indicated that release dates were
synchronous with death dates, where release dates occurred more often than not within 5 years
after a death date (Fig. 3a; 95% confidence level; 23 release events, 86 death dates). This
statistically confirms that growth releases recorded since 1800 can be used to reconstruct spruce
beetle-related mortality. Additionally, climate-sensitive spruce chronologies (Appendix A; Table
A1) did not show sustained periods of abnormally high radial growth confirming that climate
alone does not result in radial growth releases. This suggests that growth releases in subcanopy
and nonhost trees are more likely to occur in response to spruce beetle mortality than in response
to climate.
MEA of the synchrony of spruce beetle outbreak dates across CO showed outbreaks were
more likely than not to occur within 17 years of another outbreak (28 release events 1800-1990;
Fig. 3b). Confidence intervals in MEA approach 0 as the lag approaches 0 because more
outbreak events occur in the same year, or within a few years, than at longer time periods (Fig.
3b). Intervention analysis of combined spruce beetle-attributed tree release, spruce beetle-
attributed mortality, and documentary data showed statistically significant (p≤0.1) periods of
outbreak and non-outbreak. From 1800 to 2011, we found 4 periods of spruce beetle outbreak,
with the longest period of outbreak lasting from 1931 to 1957 (Fig. 2).
Association of climate with spruce beetle outbreak
Wilcox rank-sum tests comparing PRISM climate data during the period 1900 to 2011
showed that previous summer, previous fall and current summer VPD in the spruce-fir zone was
significantly higher in outbreak years than non-outbreak years (Appendix B: Table B1). We also
Sarah J. Hart et al.
13
found that PDSI across the region was significantly lower in outbreak periods than non-outbreak
periods. Previous fall and winter temperatures were higher in outbreak years than non-outbreak
years, but temperature throughout the rest of the year was similar in outbreak and non-outbreak
periods. Mann-Whitney U tests also showed that annual AMO values were significantly different
during outbreak periods (Appendix B: Table B1). AMO index values in outbreak periods
indicate that sea surface temperatures in the North Atlantic are much warmer in outbreak periods
than non-outbreak periods (Appendix B: Table B1). We did not find any significant differences
in precipitation, PDO, or ENSO between outbreak years and non-outbreak years (Appendix B:
Table B1).
The associations between spruce beetle outbreaks and PDSI and AMO detected over
1900 to 2011 were found to be consistent over the past 350 years. Since 1650, outbreaks have
tended to initiate during periods of positive AMO (i.e. drought) and negative PDSI (i.e. drought).
Of the 21 five-year periods exhibiting spruce beetle related tree-ring releases over the time
period from 1650-2011, 11 occurred during periods of negative PDSI and 11 during periods of
positive AMO. Six periods of spruce beetle outbreak occurred during times of both negative
PDSI and positive AMO (Fig. 2).
Similar results for climate predictors were obtained by random forests analysis. The RF
out-of-bag (OOB) error estimate was 24.6%, indicating that accurate classification occurred
about 75% the time. We found that the most important variable for classifying outbreak years
and non-outbreak years was the AMO index (Fig. 4a). Partial dependence, the dependence of the
probability of outbreak on one predictor after averaging out the effects of other predictor
variables, confirmed the importance of both high AMO and summer VPD values in driving
outbreak (Fig. 4b-c). The classification tree constructed from the top five predictors. AMO,
Sarah J. Hart et al.
14
summer and spring PDSI, and previous fall and summer VPD, provided insight into the climatic
thresholds that characterize spruce beetle outbreak. We found that AMO index values above 1.6
co-occur with outbreaks, particularly when in combination with summer VPD values greater
than 7.3 (Fig. 4d).
DISCUSSION
This study uses a unique multiproxy approach to reconstruct severe spruce beetle
outbreaks across NW Colorado. Outbreaks were identified from tree-ring release and mortality
data, and historical accounts. The accuracy of tree-ring release dates was confirmed, as release
events were more likely than not to occur within 5 years following a spruce beetle-induced
mortality event. Periods of broad-scale outbreak were identified as having occurred from 1843 to
1860, 1882-1889, 1931-1957, and 2004-2010. The synchronous occurrence of spruce beetle
outbreak suggests a broad-scale environmental driver. Spruce beetle outbreaks could occur
because outbreaks cause the mortality of mature spruce and thus the likelihood of a subsequent
outbreak remains low until spruce becomes a canopy dominant species again (Schmid and Hinds
1977). Although severe outbreaks, such as those documented here, can result in drastic reduction
in the abundance of large spruce lasting for more than 150 years (Hart et al. unpublished),
synchronous occurrence of 4 periods of outbreak over a ca. 200 year period points to low-
frequency climate variability as the major driver of outbreaks.
Periods of outbreak were significantly related to positive summer VPD and annual AMO
(Figure 4). High values of VPD occur during periods of drought and high temperature because
the saturation capacity of the air parcel increases with temperature, but the actual moisture in the
air parcel remains low. Thus variability in VPD describes high frequency variability in
atmospheric drought. The AMO index indicates decadal to multidecadal drought variability
Sarah J. Hart et al.
15
(McCabe et al. 2008). Variability in the AMO is likely the most important predictor of spruce
beetle outbreak in NW Colorado because high AMO index values reflect persistent drought
(McCabe et al 2008).
Drought likely influences outbreaks by affecting both tree defenses and beetle population
dynamics (Mattson and Haack 1987). For plants, drought stress may reduce photosynthesis
because stomates close to reduce transpiration (Pallardy and Kozlowski 2008). During drought
trees may also reduce leaf area, which will affect carbohydrate reserves in both the drought year
and following years (Mattson and Haack 1987). Carbon availability is important for the
production of terpene compounds that provide important conifer resistance to bark beetle attack
(Bohlmann 2012). Sustained low levels of carbohydrate production that are hypothesized to
occur during persistent drought, such as during the positive phase of the AMO, may also
predispose trees to bark beetle outbreak (Christiansen et al. 1987).
Drought may also increase spruce beetle populations because periods of drought across
southern Rocky Mountains have typically occurred during periods of high temperatures (Salzer
and Kipfmueller 2005), which directly affects larval development and survival (Hansen et al.
2001, Hansen and Bentz 2003). Above average temperatures indicated by the AMO index may
allow for the continued growth of endemic populations over multiple years, leading to outbreak
levels.
CONCLUSIONS
Here we use a novel multiproxy approach combining dendroecological data and historical
records to create a 300+ year history of spruce beetle outbreaks across NW Colorado. After
identifying the synchronous occurrence of outbreak during this time period, we use weather and
climate data to confirm the interannual association between outbreak and drought, as has been
Sarah J. Hart et al.
16
previously documented for Colorado and Utah (Hebertson and Jenkins 2008, DeRose and Long
2012a). Significantly, our results also document the importance of multidecadal variability in
drought as indicated by periods of positive AMO on spruce beetle outbreak.
The identification of interannual to multidecadal variability in drought, as opposed to
temperature alone, as an important predictor of spruce beetle outbreak, suggests that the
importance of climate in conditioning outbreaks may occur through drought-induced decreases
in tree defenses. If the primary effect of climate on broad-scale spruce beetle outbreaks was
beetle population success, we would expect temperature to be a more important predictor of
outbreak. Supporting this inference we find a reduced period of spruce beetle outbreak during the
1976-1998 warm/wet period, when beetle population success and tree defense were both likely
high. Thus future spruce beetle outbreaks across Colorado may depend on the likelihood of
future drought.
Models of future spruce beetle outbreak, driven by laboratory studies on beetle
populations (Hansen et al. 2001, Hansen and Bentz 2003), suggest increased risk as temperatures
increase and beetles become more successful (Bentz et al. 2010). The role of climate in inciting
outbreak by decreasing tree defenses may heighten this risk of outbreak. The expected increase
in global temperatures coupled with little change in precipitation will lead to higher VPD, and
likely decreased tree defenses that may also lead to spruce beetle outbreaks. Furthermore,
although climate modelers are not yet capable of deterministically predicting switches in the
phase of AMO, it is expected that the current period of positive AMO, which began in the 1990s,
may continue for decades. Given the importance of AMO in driving spruce beetle outbreak
identified in this study, the likelihood of future spruce beetle outbreak may be greater than
expected based on previous studies that consider only the anticipated changes in temperature.
Sarah J. Hart et al.
17
ACKNOWLEDGEMENTS
The manuscript was greatly improved by comments from K. Raffa, J. Negron, J. Pitlick, and two
anonymous reviewers. This research was supported by National Science Foundation awards
1203204, BCS 1262691, and DEB 0743351 and National Geographic award 8927-11.
LITERATURE CITED
Bentz, B., J. Logan, and G. Amman. 1991. Temperature-Dependent Development of the
Mountain Pine Beetle (Coleoptera : Scolytidae) and Simulation of its Phenology. Canadian
Entomologist 123:1083–1094.
Bentz, B., and others. 2009. Bark beetle outbreaks in Western North America: causes and
consequences. Bark Beetle Symposium. University of Utah Press, Snowbird, UT.
Bentz, B. J., J. Régnière, C. J. Fettig, E. M. Hansen, J. L. Hayes, J. A. Hicke, R. G. Kelsey, J. F.
Negrón, and S. J. Seybold. 2010. Climate change and bark beetles of the Western United
States and Canada: direct and indirect effects. BioScience 60:602–613.
Berg, E. E., J. David Henry, C. L. Fastie, A. D. De Volder, and S. M. Matsuoka. 2006. Spruce
beetle outbreaks on the Kenai Peninsula, Alaska, and Kluane National Park and Reserve,
Yukon Territory: Relationship to summer temperatures and regional differences in
disturbance regimes. Forest Ecology and Management 227:219–232.
Berryman, A. A. 1982. Biological control, thresholds, and pest outbreaks. Environmental
Entomology 11:544–549.
Bohlmann, J. 2012. Pine terpenoid defenses in the mountain pine beetle epidemic and in other
conifer pest interactions: specialized enemies are eating holes into a diverse, dynamic and
durable defense system. Tree Physiology 32:943–945.
Breiman, L. 2001. Random forests. Machine learning 45:5–32.
Sarah J. Hart et al.
18
Campbell, E. M., R. I. Alfaro, and B. Hawkes. 2007. Spatial distribution of mountain pine beetle
outbreaks in relation to climate and stand characteristics: A dendroecological analysis. Journal
of Integrative Plant Biology 49:168.
Christiansen, E., R. H. Waring, and A. A. Berryman. 1987. Resistance of conifers to bark beetle
attack: Searching for general relationships. Forest Ecology and Management 22:89–106.
Ciesla, W. 2011. 2010 Report on the Health of Colorado’s Forests. Pages 1–40. Colorado State
Forest Service, Fort Collins, CO.
Colorado State Forest Service. 2005. 2004 Report on the Health of Colorado’s Forests. Pages 1-
34. Colorado State Forest Service, Fort Collins, CO.
Colorado State Forest Service. 2008. 2007 Report on the Health of Colorado’s Forests. Colorado
State Forest Service, Fort Collins, CO.
Cook, E. R., C. A. Woodhouse, C. M. Eakin, D. M. Meko, and D. W. Stahle. 2004. Long-Term
Aridity Changes in the Western United States. Science 306:1015–1018.
Cutler, D. R., T. C. Edwards Jr., K. H. Beard, A. Cutler, K. T. Hess, J. Gibson, and J. J. Lawler.
2007. Random forests for classification in ecology. Ecology 88:2783–2792.
DeRose, R. J., and J. N. Long. 2007. Disturbance, structure, and composition: Spruce beetle and
Engelmann spruce forests on the Markagunt Plateau, Utah. Forest Ecology and Management
244:16–23.
DeRose, R. J., and J. N. Long. 2012a. Drought-driven disturbance history characterizes a
southern Rocky Mountain subalpine forest. Canadian Journal of Forest Research 42:1649–
1660.
Sarah J. Hart et al.
19
DeRose, R. J., and J. N. Long. 2012b. Factors Influencing the Spatial and Temporal Dynamics of
Engelmann Spruce Mortality during a Spruce Beetle Outbreak on the Markagunt Plateau,
Utah. Forest Science 58:1–14.
Doesken, N., R. A. Pielke Sr, and O. A. P. Bliss. 2003. Climate of Colorado. Climatography of
the United States 60.
Gavin, D. G. unpublished. K1D: Multivarite Ripley’s K-function for one-dimensional data.
University of Oregon, Department of Geography.
Gavin, D. G., F. S. Hu, K. Lertzman, and P. Corbett. 2006. Weak climatic control of stand-scale
fire history during the late Holocene. Ecology 87:1722–1732.
Gotelli, N. J., and A. M. Ellison. 2004. A Primer Of Ecological Statistics, 1st Ed. Sinauer Assoc.
Gray, S. T., J. L. Betancourt, C. L. Fastie, and S. T. Jackson. 2003. Patterns and sources of
multidecadal oscillations in drought-sensitive tree-ring records from the central and southern
Rocky Mountains.
Hansen, E. M., and B. J. Bentz. 2003. Comparison of reproductive capacity among univoltine,
semivoltine, and re-emerged parent spruce beetles (Coleoptera: Scolytidae). Canadian
Entomologist 135:697–712.
Hansen, E. M., B. J. Bentz, and D. L. Turner. 2001. Temperature-based model for predicting
univoltine brood proportions in spruce beetle (Coleoptera: Scolytidae). Canadian
Entomologist 133:827–841.
Hard, J. S. 1985. Spruce beetles attack slowly growing spruce. Forest Science 31:839–850.
Hart, S.J., Veblen, T.T., and Kulakowski, D. Unpublished. Do tree and stand-level attributes
determine susceptibility of spruce-fir forests to spruce beetle outbreaks?
Sarah J. Hart et al.
20
Hebertson, E. G., and M. J. Jenkins. 2008. Climate factors associated with historic spruce beetle
(Coleoptera: Curculionidae) outbreaks in Utah and Colorado. Environmental Entomology
37:281–292.
Hinds, T. E., F. G. Hawksworth, and R. W. Davidson. 1965. Beetle-Killed Engelmann Spruce its
Deterioration in Colorado. Journal of Forestry 63:536–542.
Hopkins, A. D. 1909. Practical information on the scolytid beetles of North American forests. I.
Bark beetles of the genus Dendroctonus. Govt. print off.
Knight, F. B. 1953. Engelmann spruce beetle conditions: Routt, Arapaho, and White River
National Forests, Colorado. Unnumbered Report, USDA, Agricultural Research
Administration, Bureau of Entomology and Plant Quarantine, Division of Forest Insect
Investigations, Forest Insect Laboratory, Fort Collins, CO.
Knight, F. B., and W. F. McCambridge. 1952. Engelmann spruce beetle conditions: Routt,
Arapaho, and White River National Forests, Colorado, 1952. RX-INF, USDA Agricultural
Research Administration, Bureau of Entomology and Plant Quarantine, Division of Forest
Insect Investigations, Forest Insect Laboratory, Fort Collins, CO.
Kurz, W. A., C. C. Dymond, G. Stinson, G. J. Rampley, E. T. Neilson, A. L. Carroll, T. Ebata,
and L. Safranyik. 2008. Mountain pine beetle and forest carbon feedback to climate change.
Nature 452:987–990.
Liaw, A., and M. Wiener. 2002. Classification and Regression by randomForest. R News 2:18–
22.
Logan, J. A., and B. J. Bentz. 1999. Model analysis of mountain pine beetle (Coleoptera:
Scolytidae) seasonality. Environmental Entomology 28:924–934.
Sarah J. Hart et al.
21
MacDonald, G. M., and R. A. Case. 2005. Variations in the Pacific Decadal Oscillation over the
past millennium.
Massey, C., and N. Wygant. 1954. Biology and control of the Engelmann spruce beetle in
Colorado. USDA Forest Service, Washington, D.C.
Mattson, W. J., and R. A. Haack. 1987. The role of drought in outbreaks of plant-eating insects.
BioScience 37:110–118.
McCabe, G. J., J. L. Betancourt, S. T. Gray, M. A. Palecki, and H. G. Hidalgo. 2008.
Associations of multi-decadal sea-surface temperature variability with US drought. Quaternary
International 188:31–40.
McCambridge, W. F., and F. B. Knight. 1972. Factors affecting spruce beetles during a small
outbreak. Ecology 53:830–839.
Newman, M., G. P. Compo, and M. A. Alexander. 2003. ENSO-Forced Variability of the Pacific
Decadal Oscillation. Journal of Climate 16:3853–3857.
Pallardy, S. G., and T. T. Kozlowski. 2008. Physiology of woody plants. Academic Press.
Peet, R. K. 2000. Forests and meadows of the Rocky Mountains. Pages 75–123 in M. G. Barbour
and D. W. Billings, editors. North American Terrestrial Vegetation. 2nd Edition. Cambridge
University Press, New York, NY.
R Development Core Team. 2010. R: A language and environment for statistical computing. R
Foundation for Statistical Computing, Vienna, Austria.
Raffa, K. F., B. H. Aukema, B. J. Bentz, A. L. Carroll, J.A. Hicke, M. G. Turner, and W. H.
Romme. 2008. Cross-scale drivers of natural disturbances prone to anthropogenic
amplification: The dynamics of bark beetle eruptions. BioScience 58:501-517.
Sarah J. Hart et al.
22
Ray, A. J., J. J. Barsugli, K. B. Averyt, K. Wolter, M. Hoerling, N. Doesken, B. Udall, and R. S.
Webb. 2008. Climate change in Colorado: A synthesis to support water resource management
and adaptation. Denver, Colorado.
Rodionov, S. N. 2004. A sequential algorithm for testing climate regime shifts. Geophysical
Research Letters 31:L09204.
Rodionov, S. N. 2006. Sequential Regime Shift Detection. Pacific Marine Environmental
Laboratory, Seattle, WA.
Safranyik, L., and A. L. Carroll. 2007. The biology and epidemiology of the mountain pine
beetle in lodgepole pine forests. Pages 3–66. The mountain pine beetle: a synthesis of biology,
management and impacts on lodgepole pine.
Salzer, M. W., and K. F. Kipfmueller. 2005. Reconstructed temperature and precipitation on a
millennial timescale from tree rings in the Southern Colorado Plateau, USA. Climatic Change
70:465–487.
Schmid, J. M., and R. H. Frye. 1977. Spruce beetle in the Rockies. USDA Forest Service, Rocky
Mountain Forest and Range Experiment Station, Fort Collins, CO.
Schmid, J. M. 1981. Spruce beetles in blowdown. Resource Note, USDA Forest Service, Rocky
Mountain Forest and Range Experiment Station, Fort Collins, CO.
Schoennagel, T., T. T. Veblen, D. Kulakowski, and A. Holz. 2007. Multidecadal climate
variability and climate interactions affect subalpine fire occurrence, western Colorado (USA).
Ecology 88:2891–2902.
Sherriff, R. L., E. E. Berg, and A. E. Miller. 2011. Climate variability and spruce beetle
(Dendroctonus rufipennis) outbreaks in south-central and southwest Alaska. Ecology
92:1459–1470.
Sarah J. Hart et al.
23
Stokes, M. A. and T. A. Smiley. 1968. An Introduction to Tree Ring Dating. University of
Arizona Press.
Swetnam, T. W., and A. M. Lynch. 1989. A tree-ring reconstruction of western spruce budworm
history in the southern Rocky Mountains. Forest Science 35:962–986.
Veblen, T. T., K. S. Hadley, E. M. Nel, T. Kitzberger, M. Reid, and R. Villalba. 1994.
Disturbance regime and disturbance interactions in a Rocky Mountain subalpine forest.
Journal of Ecology 82:125–135.
Veblen, T. T., K. S. Hadley, M. S. Reid, and A. J. Rebertus. 1991. Methods of detecting past
spruce beetle outbreaks in Rocky Mountain subalpine forests. Canadian Journal of Forest
Research 21:242–254.
Veblen, T. T., D. Kulakowski, K. S. Eisenhart, and W. L. Baker. 2001. Subalpine forest damage
from a severe windstorm in northern Colorado. Canadian Journal of Forest Research 31:2089–
2097.
Ware, D. 1995. A century and a half of change in the climate of the NE Pacific. Fisheries
Oceanography 4:267–277.
Werner, R. A., and E. H. Holsten. 1983. Mortality of white spruce during a spruce beetle
outbreak on the Kenai Peninsula in Alaska. Canadian Journal of Forest Research 13:96–101.
Wygant, N. D., and A. L. Nelson. 1949. Four billion feet of beetle-killed spruce. USDA
Yearbook Agric:417–422.
Supplementary Material
Appendix A: Sampling information for each tree-ring site.
Appendix B: Results from Mann-Whitney U tests comparing climate conditions during
outbreaks and non-outbreak years.
Sarah J. Hart et al.
24
TABLE 1: The initiation of spruce beetle outbreaks across NW Colorado (modified from
Hebertson and Jenkins 2008).
Initiation year End year National Foresta Source
1850 1859 GMNF Knight and McCambridge 1952
1853 1889 PNF Hopkins 1909
1882 1887 WRNF Hopkins 1909
1939 1951 WRNF Knight 1953, Massey and Wygant 1954
1942 1948 ARNF Wygant and Nelson 1949
1944 1951 GMNF Knight 1953, Hinds et al. 1965, Schmid
and Frye 1977
1957 1960 RNF McCambridge and Knight 1972
1971 1979 GNF Schmid and Frye 1977
1997 2007 RNF Colorado State Forest Service 2008,
Ciesla 2011
2004 2010 GMNF Colorado State Forest Service 2005,
Ciesla 2011
2004 2010 GNF Colorado State Forest Service 2005,
Ciesla 2011
2005 2010 RONF Colorado State Forest Service 2005,
Ciesla 2011
a – GMNF = Grand Mesa National Forest; WRNF = White River National Forest; RNF = Routt
National Forest; GNF = Gunnison National Forest; RONF = Roosevelt National Forest;
PNF=Pike National Forest; ARNF= Arapaho National Forest
Sarah J. Hart et al.
25
FIGURE CAPTIONS
FIGURE 1: Locations of climate-sensitive tree-ring sites and spruce beetle history sites used in
this study.
FIGURE 2: Variability of annual AMO and annual PDSI (a) and spruce beetle outbreaks (b)
through time. a) Tree-ring based reconstructions of AMO and PDSI. Values above 0 indicate
warm phases of the AMO that correspond with periods of drought. Periods of drought are
represented by negative values of the PDSI. b) Tree-ring and documentary records of spruce
beetle outbreaks in 5-year periods. Dark gray bars represent the percentage of USFS National
forest districts within NW Colorado with a documented spruce beetle outbreak. Crosshatched
bars indicate the percentage of tree-ring reconstruction sites exhibiting a statistically significant
release (n=39 events binned into 23 5-year periods) and light gray bars indicate the percentage of
sites exhibiting a tree-ring record of spruce beetle-induced death. The dashed horizontal lines
indicate the number of potential tree-ring recording sites. Periods of outbreak identified by the
intervention analysis (Rodionov 2004) are highlighted by light gray shading.
FIGURE 3: Temporal synchrony analysis between different types of tree-ring based evidence of
spruce beetle outbreak (a) and dates of spruce beetle outbreak at 18 sites across northwestern
Colorado (b) over the time period 1800-1990. (a) Backwards bivariate event analysis of temporal
synchrony between growth releases attributed to spruce beetle outbreak (n=23) and death dates
(n=86) attributed to spruce beetle outbreak across sites with mortality dates. (b) Bidirectional
multivariate event analysis of temporal synchrony in the initiation of growth releases attributed
to spruce beetle outbreak (n=28) across all sites. In both a and b, the solid black line is the L(t)
Sarah J. Hart et al.
26
function, the K function transformed, where values >0 indicate synchrony and values <0 indicate
asynchrony. The dashed lines indicate 99% confidence envelopes based on 1000 Monte Carlo
simulations. Gray shaded areas indicate years of significant synchrony (L(t) >0) or asynchrony
(L(t) <0).
FIGURE 4: Results from random forest (RF) analysis of spruce beetle outbreak years with
instrumental data (1900 to 2011). a) Variable importance plots for the top ten predictor variables
from RF used for predicting the occurrence of outbreak. The mean decrease in accuracy variable
is the normalized difference of the classification accuracy when the data for that variable are
included and when they have been randomly permutated. Higher values indicate variables that
are more important to classification. Palmer Drought Severity Index (PDSI) and vapor pressure
deficit (VPD) variables are defined seasonally. A lower case p in front of the season indicates
values from the previous year. The right two panels show partial dependence plots of b) annual
AMO and c) summer VPD. Partial dependence is the dependence of the probability of outbreak
on one predictor after averaging out the effects of other predictor variables. d) Classification tree
for determining spruce beetle outbreak years from non-outbreak years (1900-2011). On the tree,
if condition is satisfied, proceed to the left of the tree. Tree nodes describe the predicted
condition, the probability of outbreak, and percent of observations. The tree was constructed
using the top five predictor variables identified in RF.
Sarah J. Hart et al.
27
FIGURE 1
Sarah J. Hart et al.
28
FIGURE 2
Time
index
-2012
--PDSIAMO
a)
1650 1680 1710 1740 1770 1800 1830 1860 1890 1920 1950 1980 2010
Death eventsRelease eventsDocumentary outbreaks
% o
f site
s
010
2030
4050
510
15
Num
ber o
f tre
e-rin
g si
tes
b)
Sarah J. Hart et al.
29
FIGURE 3
t(years)
L(t)
func
tion
-50 -40 -30 -20 -10 0
-50
5
a)
t(years)
L(t)
func
tion
0 10 20 30 40 50
-50
510
b)
Sarah J. Hart et al.
30
FIGURE 4
PDSI.p.summer
TEMP.p.fall
PDSI.p.fall
VPD.p.summer
PDSI.winter
VPD.p.fall
PDSI.spring
PDSI.summer
VPD.summer
AMO
5 10 15
mean decrease in accuracy
a)
-4 -2 0 2 4
-0.9
-0.6
-0.3
annual AMO index
b)
6 7 8 9 10
-0.8
-0.6
summer VPD index
c)(lo
git o
f pro
babi
lity
of o
utbr
eak)
/2
AMO < 1.6
AMO < -0.35
VPD.summer < 7.3no
outbreak0.02 47% no
outbreak0.19 15%
outbreak0.56 15%
outbreak0.81 24%
d)