PDF Archive

Easily share your PDF documents with your contacts, on the Web and Social Networks.

Share a file Manage my documents Convert Recover PDF Search Help Contact



Climate, weather, and recent .pdf


Original filename: Climate, weather, and recent.pdf
Title: Climate, weather, and recent mountain pine beetle outbreaks in the western United States
Author: Eric P. Creeden

This PDF 1.7 document has been generated by Elsevier / Acrobat Distiller 8.0.0 (Windows), and has been sent on pdf-archive.com on 14/03/2017 at 19:47, from IP address 100.15.x.x. The current document download page has been viewed 406 times.
File size: 1.6 MB (13 pages).
Privacy: public file




Download original PDF file









Document preview


Forest Ecology and Management 312 (2014) 239–251

Contents lists available at ScienceDirect

Forest Ecology and Management
journal homepage: www.elsevier.com/locate/foreco

Climate, weather, and recent mountain pine beetle outbreaks in the
western United States
Eric P. Creeden, Jeffrey A. Hicke ⇑, Polly C. Buotte
University of Idaho, Department of Geography, 810 W 7th Street, McClure Hall 203, Moscow, ID 83844-3021, USA

a r t i c l e

i n f o

Article history:
Received 29 April 2013
Received in revised form 5 September 2013
Accepted 26 September 2013
Available online 31 October 2013
Keywords:
Adaptive seasonality
Cold-induced mortality
Climate suitability modeling
Drought stress
Lodgepole pine (Pinus contorta)
Mountain pine beetle (Dendroctonus
ponderosae)

a b s t r a c t
Recent outbreaks of mountain pine beetle (Dendroctonus ponderosae) have impacted large areas of western North America. Climate and weather conditions influence beetle population dynamics, and managers
and policymakers are concerned about the potential effects of climate change on outbreaks. Here we
studied five locations with extensive outbreaks in lodgepole pine (Pinus contorta) forests across the western United States. Using observations and modeling, we quantified means and changes relative to prior
years of three climate or weather factors associated with outbreaks: (1) year-round temperatures that
affect adaptive seasonality; (2) low temperatures that induce mortality of overwintering beetles; and
(3) drought stress of host trees. Climate variable means varied among locations, indicating the beetle’s
tolerance to different climate during outbreaks. Analyses of climate or weather factors as outbreaks progressed revealed that year-round temperatures during outbreaks were typically higher than in prior
years, and outbreak years lacked very low winter temperatures that often occurred in prior years.
Drought was present at each location during some time of an outbreak, and increases in beetle-caused
tree mortality at lower beetle population levels (as indicated by killed trees) were usually coincident with
drought. Furthermore, drought was not required to maintain large outbreaks; in several locations, relief
from drought during periods of high tree mortality did not cause subsequent declines in tree mortality.
We did not find strong evidence that maladaptive seasonality, cold-induced mortality, or drought stress
was responsible for decreases in tree mortality, suggesting the role of host depletion. Large variability in
the relationships between climate or weather variables and outbreaks suggests that different climate and
weather factors may have been limiting outbreaks at different times and that these factors did not influence beetle-caused tree mortality similarly among locations. Our results increase understanding of the
climate and weather factors that influence beetle outbreaks and their variability in space and time and
will lead to more accurate predictions of future patterns of outbreaks that consider future climate.
Ó 2013 Elsevier B.V. All rights reserved.

1. Introduction
Native bark beetle (Coleoptera: Curculionidae, Scolytinae) outbreaks are natural processes of forested ecosystems of western
North America. However, recent outbreaks of some bark beetles
are the largest in recorded history and have exceeded previously
documented ranges, impacts, and frequencies (Carroll et al.,
2004b; Raffa et al., 2008; Logan et al., 2010; Safranyik et al.,
2010). Notable among these species is the mountain pine beetle
(Dendroctonus ponderosae Hopkins), whose outbreaks have impacted millions of hectares of lodgepole (Pinus contorta) and other
pine forests across western North America (Meddens et al., 2012).
A challenge for current forest management is to predict beetle outbreaks given probable climate change in the coming decades.

⇑ Corresponding author. Address: 875 Perimeter Drive Stop 3021, Moscow, ID
83844-3021, USA. Tel.: +1 208 885 6240; fax: +1 208 885 2855.
E-mail address: jhicke@uidaho.edu (J.A. Hicke).
0378-1127/$ - see front matter Ó 2013 Elsevier B.V. All rights reserved.
http://dx.doi.org/10.1016/j.foreco.2013.09.051

Climate and weather influence outbreaks of mountain pine beetle through three main factors: (1) adaptive seasonality; (2) coldinduced mortality; and (3) drought stress on host tree species.
These factors manifest themselves through changes in the physiology and life history traits of the beetle as well as through changes
in host tree physiology (Safranyik and Carroll, 2006; Bentz et al.,
2010). The first factor, adaptive seasonality, results from the conditions of univoltinism (one-year life cycle) as well as synchronous
adult emergence and life cycle timing (Bentz et al., 1991; Logan
and Bentz, 1999; Logan and Powell, 2001). Temperature directly
controls life stage development rates of mountain pine beetle,
including progression through egg, larvae, pupae, and adult life
stages (Logan and Bentz, 1999; Logan and Powell, 2001; Safranyik
and Carroll, 2006). Maladaptive seasonality, such as fractional voltinism (two- or three-year life cycles) found in colder environments, can limit the outbreak potential of mountain pine beetles
(Amman, 1973; Logan and Powell, 2009). Genetic variability in
sensitivity to temperatures has been demonstrated for different
mountain pine beetle populations (Bentz et al., 2001, 2011a).

240

E.P. Creeden et al. / Forest Ecology and Management 312 (2014) 239–251

Low summer temperatures impacting synchronous adult emergence and life cycle timing hindered the building of an incipient
beetle population in central Idaho in 1993, highlighting the potential of this factor to influence beetle populations (Logan and Powell,
2009).
The second factor is beetle larvae mortality from very low temperatures during the cold season (including late fall, winter, and
early spring). Low temperatures during this time have contributed
to population declines of mountain pine beetle across its range
(Evenden and Gibson, 1940; Lessard et al., 1987; Safranyik and
Linton, 1991; Kipfmueller et al., 2002; Bentz et al., 2011b), can
determine thermal boundaries of available habitat (Safranyik,
1978; Logan and Powell, 2009), and can reduce outbreak probability (Sambaraju et al., 2012). During large outbreaks, however,
cold-induced mortality events may act to slow down outbreaks,
not completely stop them (Evenden and Gibson, 1940; Hopping
and Mathers, 1945; Safranyik, 1978). Overwintering larvae progressively gain cold tolerance with decreasing temperatures (Bentz
and Mullins, 1999), and mortality thresholds have been developed
in laboratory studies (Wygant, 1940; Bentz and Mullins, 1999).
Using a model of winter survival developed by Régnière and Bentz
(2007), Safranyik et al. (2010) proposed that population growth occurs when the modeled winter survival probability is greater than
0.2; probabilities below that threshold lead to declines in
population.
The third factor is drought stress on host trees. Drought affects
the capability of a host tree to defend itself against bark beetle attack (Raffa and Berryman, 1983; McDowell et al., 2011). Warmer
summers and drought conditions were associated with previous
bark beetle outbreaks in North America, including spruce beetle
(Dendroctonus rufipennis) (Berg et al., 2006; Hebertson and Jenkins,
2008), pinyon ips beetle (Ips confusus) (Breshears et al., 2005), and
mountain pine beetle (Safranyik et al., 1975; Thomson and
Shrimpton, 1984; Raffa et al., 2008; Thomson, 2009; Chapman
et al., 2012). Increased host tree vigor increases resistance to attack
by mountain pine beetle (Mitchell et al., 1983), and growth rate
responses to climatic water deficit and drought influenced patterns
of pine mortality in California (Millar et al., 2007, 2012). Not all
documented outbreaks, however, correspond with drought conditions (Beal, 1943; Powell, 1969; Crookston, 1977). Furthermore,
drought may negatively influence beetle populations because
decreased survival and brood production occurs in depleted or
dried phloem tissue (reduced phloem thickness) of host trees
(Amman, 1972; Safranyik and Carroll, 2006).
Models have been developed to simulate the role that climate
or weather plays in mountain pine beetle outbreaks. Process models based on insect physiology and climate or weather data have
been developed for predicting adaptive seasonality (Bentz et al.,
1991; Logan and Powell, 2001) and cold-induced mortality
(Régnière and Bentz, 2007). These models have been used to predict current and future climate suitability for the beetle (Bentz
et al., 2010; Safranyik et al., 2010), but lack evaluation across much
of the beetle’s range. Statistical models have identified relationships between beetle metrics such as area affected, trees killed,
or population growth rate and climate or weather metrics using
correlation, regression, or other statistical modeling (Kipfmueller
et al., 2002; Aukema et al., 2008; Evangelista et al., 2011; Jewett
et al., 2011; Preisler et al., 2012).
Descriptive and graphically focused methods showing climate
or weather patterns related to mountain pine beetle outbreaks
have also been conducted (Powell, 1969; Thomson and Shrimpton,
1984), and may be better at establishing the nature of relationships
than explicitly quantifying them (Thomson and Shrimpton, 1984).
Powell (1969) averaged conditions five years before, during, and
five years after outbreaks in western Canada in the 1900s and
found mixed results regarding the influence of climate or weather

on outbreaks. Such an analysis may highlight general differences in
climate among time periods, but may also mask interannual variations in weather important to beetle and host tree ecology.
Although much is understood about the role of climate and
weather in influencing beetle outbreaks, gaps remain that limit
our understanding and therefore ability to predict future changes
in outbreak regimes. Mountain pine beetle populations exhibit variability in traits related to climate and weather, such as temperature sensitivity of development rates (Bentz et al., 2001, 2011a),
yet more information is needed to ascertain differences among
populations across the beetle’s range. Analyses documenting patterns of climate and weather before, during, and after outbreaks
of mountain pine beetle at multiple locations in the United States
have not been conducted. Although the causes of population declines of some previous beetle outbreaks have been reported in
the literature (Safranyik and Linton, 1991; Stahl et al., 2006),
causes of decline are understudied and likely involve multiple factors, including host tree availability, cold-induced mortality, and
maladaptive seasonality.
To address these gaps, we conducted a study of the patterns of
climate and weather factors at five locations of recent (since 1980)
mountain pine beetle outbreaks across the western US. We analyzed patterns of climate and weather variables and tree mortality
during the progression of outbreaks. Our objectives were to: (1)
compare climate means during outbreaks among locations; (2)
document the temporal patterns of weather factors for decades before, during, and after outbreaks; and (3) examine the relationship
between interannual changes in weather factors and changes in
beetle-caused tree mortality during outbreaks.
2. Materials and methods
2.1. Study locations
We studied five major mountain pine beetle outbreaks in lodgepole pine forests across the western US in Utah, Colorado, Idaho,
Montana, and Oregon (Fig. 1). Locations were chosen by examining
the cumulative beetle outbreak extent within lodgepole pine from
all years of acquired aerial survey maps, USFS Forest Insect and
Disease Condition Reports (e.g., USDA, 1977; Man, 2010), lodgepole
pine host range (Little, 1971b), outbreak initiation and collapse
locations, and the number of trees killed within a location. Study
locations varied in size and elevation (Table S1).
2.2. Data used in analysis
We analyzed the number of lodgepole pine trees killed by
mountain pine beetles. Aerial survey data collected by US and
Canadian agencies have been used to document and investigate
large-scale patterns of forest insect outbreaks (Peltonen et al.,
2002; Aukema et al., 2008; Powell and Bentz, 2009). Observers in
aircraft map damage polygons, attribute tree mortality to disturbance agents, and estimate attack severity (number of trees killed).
Spatial location errors, errors in discerning host tree type, and errors discerning host tree mortality are inherent using aerial detection survey techniques and therefore convey uncertainties, but
these databases provide high spatial coverage of forest disturbance
(Aukema et al., 2008; Powell and Bentz, 2009). We used 1-km gridded data sets of tree mortality in Washington and Oregon from
1980 to 2008 (Preisler et al., 2012) and throughout western North
America from 1997 to 2010 (lower estimate of Meddens et al.,
2012). To extend the period of tree mortality prior to 1997, we obtained polygonal aerial survey maps from USFS Regions 1 and 4
(UT, ID, MT) from 1991 to 1996 and USFS Region 2 (CO, WY) from
1995 to 1996. Aerial surveys were not always flown in all areas

E.P. Creeden et al. / Forest Ecology and Management 312 (2014) 239–251

241

Fig. 1. Mountain pine beetle outbreaks (gray) in lodgepole pine range (hatched; Little, 1971a) from 1996 to 2009 in the western United States and 1979–2009 in Washington
and Oregon from aerial surveys. Corresponding time series of the number of lodgepole pine trees killed at five study locations are shown at left.

every year, and the western half of the Montana location was not
flown in 2007, potentially omitting trees killed and making the increase in detection of tree mortality from 2007 to 2008 steeper
than reality. We subtracted one year from the aerial survey data
to account for the one-year lag between the year of mountain pine
beetle attack and the year of detection by surveyors. We spatially
aggregated the 1-km gridded tree mortality data set to 4-km to
match the spatial resolution of the Parameter-elevations Regressions on Independent Slopes Model (PRISM) data set (Daly et al.,
1994), used for climate and described below.
We selected a variety of climate or weather variables documented to be ecologically important for mountain pine beetle outbreaks and representing adaptive seasonality, cold-induced
mortality, and drought stress (Table 1). Different variables representing one factor were considered because of uncertainties in
which variable best represents each factor. To ensure a temporal
match of beetle attack with weather data and process model output, variables were calculated for the yearlong beetle developmental period (we assumed univoltinism) or during the summer of
beetle attack. Within each outbreak location, we identified grid
cells with cumulative lodgepole pine mortality >4000 trees per
16-km2 grid cell ( 1 killed lodgepole pine tree/acre) over the period of available aerial survey data for subsequent analysis. These
grid cells thus contained sufficient number of susceptible trees
based on both stand (e.g., stem density, size, age) and weather conditions to support large beetle populations. We then averaged
weather variables for these grid cells to produce one time series
per location per weather variable. For each factor, we show a subset of representative variables; other variables exhibited similar
patterns.

One set of variables consisted of temperature and precipitation
from monthly and daily gridded data sets. Monthly gridded weather data sets from 1895 to 2009 from PRISM were used at each
study location. In addition, we obtained gridded, daily data from
1915 to 2009 from the 1/8 degree ( 12-km) variable infiltration
capacity (VIC) macro-scale hydrologic model (Liang et al., 1994;
Hamlet and Lettenmaier, 2005) to analyze mean temperatures at
a finer temporal resolution at the outbreak locations.
A second set of variables was computed from temperature and
precipitation (or seasonal or yearly averages) or from models. Summer (JJA) self-calibrated Palmer Drought Severity Index (PDSI), a
location-specific metric, was derived from the PRISM dataset
(Wells et al., 2004; Kangas and Brown, 2007). Climatic water deficit, calculated as potential evapotranspiration minus actual evapotranspiration (AET) and based on the water year, October 1 through
September 30, was computed using time series of aggregated,
monthly PRISM data and the program AET Calculator (Gavin and
Hu, 2006) (http://geography.uoregon.edu/envchange/pbl/software.
html). AET Calculator uses a modified Thornthwaite method
(Willmott et al., 1985) to calculate climatic water balance, and
we assumed a field capacity of 100 mm and a nonlinear declining
availability function for plant-available water as used by Littell
et al. (2008) for Douglas-fir growth–climate relationships. Mean
latitudes of the outbreak locations (Table S1) were used in the
AET Calculator for calculating day length. Climatic water deficit is
a more biologically meaningful measure of drought than other
metrics (Stephenson, 1998).
A third set of variables consisted of outputs of process models
that simulated weather suitability for mountain pine beetle outbreaks. We ran these models with the BioSIM modeling framework

242

E.P. Creeden et al. / Forest Ecology and Management 312 (2014) 239–251

Table 1
Climate and weather variables representing effects on mountain pine beetle outbreaks. Sources: Parameter-elevation Regressions on Independent Slopes Model gridded data set
(PRISM); National Weather Service Cooperative Observer Program (NWS COOP); variable infiltration capacity model (VIC); AET Calculator; BioSIM.
Factor

Description

Variable

Climate
database

Model

Temporal resolution of climate
database

Adaptive
seasonality

(1) Univoltinism
(2) Population synchrony
(3) Correct emergence
timing
Mountain pine beetle
mortality
caused by low
temperatures

Water year (October–September) mean
temperature (°C)
March–August mean temperature (°C)
Adaptive seasonality model (Logan2b)
Minimum temperature of coldest winter
month (DJF) (°C)
Water year (October–September) minimum
daily mean temperature
Cold tolerance (ColdT, probability of survival)
Summer (JJA) Palmer Drought Severity Index (PDSI)
Climatic water deficit (October–September) (mm)

PRISM



Monthly

PRISM
NWS COOP
PRISM


BioSIM


Monthly
Daily
Monthly

NWS COOP

VIC

Daily

NWS COOP
PRISM
PRISM

BioSIM

AET
Calculator



Daily
Monthly
Monthly

Cold-induced
mortality

Drought stress

Decreased host vigor and
resistance capability

March-August precipitation (cm)
PRISM
Water year (October–September) precipitation (cm) PRISM

(Régnière et al., 1996) for grid cells with high mortality (as defined
above) within each outbreak location. We used a database of National Weather Service Co-op stations of daily weather data from
1981 to 2010 as input to BioSIM, which were then interpolated
to grid cells via inverse distance weighting and vertical lapse rates
to compute daily weather variables (Régnière et al., 1996). BioSIM
then ran suitability models using these weather variables as inputs. We calculated adaptive seasonality (Logan2b; Bentz et al.,
1991; Logan et al., 2010) and probability of winter survival associated with cold tolerance (ColdT; Régnière and Bentz, 2007). The
adaptive seasonality model predicts a binary response of Logan2b
when using annual weather (not climatological) inputs. Outputs
of both models near zero indicate low suitability and near one indicate high suitability for beetles. We computed spatially averaged
yearly output of these models from 1982 to 2010 for each outbreak
location. See Supplementary Data and Safranyik et al. (2010) for
more details about both models.
2.3. Data analysis
We sought to compare weather in outbreak years among locations. To identify outbreak years, we defined the initial year of outbreaks at each location as the year when growth rate (rt) of
lodgepole pine mortality exceeded two and was followed by at
least three additional years of positive growth (Table S2). Safranyik
and Carroll (2006) noted that incipient-epidemic populations generally increase slowly, and growth rates may not exceed two for a
number of years. Using Mt as the observed number of trees killed in
year t, growth rate is:

Mt
rt ¼
:
M t 1

ð1Þ

Lodgepole pine mortality data were available beginning in 1979
for the Oregon location, and we classified the first year of outbreaks as 1976 based on information from forest condition reports
(USDA, 1977). Because aerial survey data do not capture the transition to epidemic populations, which has been identified as 20 attacked (not killed) trees per hectare (Boone et al., 2011), we tested
how these comparisons changed when we redefined outbreak
years as those with large numbers of killed trees (2000 in Utah,
Colorado, and Idaho and 2003 in Montana). Little change occurred
in our results, illustrating the lack of sensitivity to this definition.
Large decreases of tree mortality were designated as years of outbreak decline (Table S2).
We performed several analyses to address our objectives. We
first compared mean weather conditions during outbreak years

Monthly
Monthly

among locations for each variable listed in Table 1. Second, we assessed time series of weather variables together with the number
of lodgepole pine trees killed to document the temporal patterns
of weather factors before, during, and after outbreak years at each
location (where possible). Although aerial survey data cannot capture beetle population dynamics at low populations (Boone et al.,
2011), we analyzed weather factors for several decades before
the first available aerial surveys to evaluate weather variability
during periods of low levels of populations.
Finally, we assessed changes in outbreaks relative to changes in
weather by plotting the number of trees killed as a function of each
weather variable. These plots allowed us to analyze the magnitude
and year-to-year change in tree mortality relative to changes in
weather variables. Importantly, this technique allowed us to identify changes in the relationship as a function of outbreak size. Furthermore, this method was useful because actual weather values,
as opposed to anomalies, are critical when considering influences
on beetle populations. For example, cold-hardened beetles experience a higher probability of mortality when temperatures approach 40 °C (Wygant, 1940), not when winter temperatures
are anomalously lower by some amount compared to long-term
means that are well above 40 °C. We show both the actual and
log10 number of lodgepole pine trees killed versus weather variables to illustrate patterns of variables both at high and low outbreak levels, respectively.
3. Results
The aerial survey databases quantified the progression from
lower to higher levels of beetle activity at most outbreak locations
(Fig. 1). The exception was Oregon, where beetle attacks were documented in 1976 but available aerial survey data began in 1979
when tree mortality was relatively high. All study locations experienced some degree of pine mortality in every year that aerial survey data was acquired. However, the number of trees killed
increased rapidly within 1–3 years at each location. These
increases were not related to the area surveyed by the Aerial
Detection Survey program, but instead appeared to be related to
increases in beetle-caused tree mortality.
3.1. Comparison of mean weather during outbreaks among locations
Outbreak locations showed similar annual cycles of temperature and differing annual cycles of precipitation during outbreak
years (Fig. 2). Oregon and Montana had consistently higher
monthly temperatures than other locations, whereas Idaho

243

E.P. Creeden et al. / Forest Ecology and Management 312 (2014) 239–251

experienced the lowest mean temperatures in nearly every month.
Oregon and Idaho experienced high precipitation during winter
and very low precipitation during summer, whereas locations further inland had peak monthly precipitation in late spring and early
summer. Monthly precipitation was lowest in either July or August
at all outbreak locations.
Temporal distributions of variables representing adaptive seasonality indicated that Utah and Idaho typically had the lowest
mean water year and March–August temperatures, and Montana
and Oregon had the highest mean temperatures (Fig. 3). Water
year mean temperatures (coinciding with the time of beetle development) ranged from 1.9 °C in Idaho to 5.7 °C in Oregon. Patterns
of Logan2b (adaptive seasonality model results) among sites were
similar to the temperature variables, although Idaho values were
higher and more similar to the other sites compared with the temperature variables.
Water year minimum daily mean temperature and ColdT (cold
tolerance model results) exhibited similar intersite variability,
and differed from minimum temperature of the coldest winter
month (Fig. 3). Montana experienced the lowest minimum daily
mean temperatures ( 25 °C) and Oregon had a mean of 11 °C
(Fig. 3). Means of ColdT varied from 0.35 at Montana to 0.75 at Oregon. In contrast to the other variables, values of the minimum temperatures of the coldest winter month in Montana were relatively
higher compared with other sites, illustrating the importance of
cold snaps that occur outside of December through February that
are captured by the other cold season metrics.
Variables representing drought stress were more similar
among locations than adaptive seasonality and cold-induced mortality variables (Fig. 3). Locations were typically in drought during
outbreaks (negative PDSI, larger water deficits). The influence of
summer drought was evident in Oregon and Idaho despite average to higher water year precipitation in these locations. Mean
summer PDSI computed for outbreak periods was negative for
all locations.
3.2. Weather factors leading up to and during each outbreak
In Table 2, we summarize relative changes in weather factors
during outbreaks compared with prior decades as derived from
Figs. 4–7 and Figs. S1-S5. Warmer conditions typically occurred

during outbreaks than in the decades before. Both winter and
year-round temperatures increased, affecting winter survival and
adaptive seasonality. At most locations, years prior to outbreaks often were associated with low ColdT and therefore low probability
of winter survival, whereas such values were typically absent during peak outbreak periods. Conditions during outbreaks were often
more favorable for adaptive seasonality, although sometimes conditions were similar to those in years prior to outbreaks (i.e., no
change).
In addition to warming, drought also appeared to be an important factor. In Utah, Colorado, Idaho, and Montana, wetter conditions in the late 1990s were followed by a severe, multiyear
drought in the early 2000s that occurred prior to or early in outbreaks. The timing of tree mortality increases at these four locations relative to drought was not the same, however. Idaho tree
mortality peaked in 2001, a year after the drought initiation. In
contrast, large increases in the number of trees killed occurred later in Colorado and Montana, with mortality peaking at both locations in 2007. Oregon also experienced drought conditions early in
the outbreak, but had more interannual variability (more wet
years) than the other locations.
3.3. Changes in weather factors during outbreaks
March–August mean temperature and Logan2b during outbreaks showed high spatial and temporal variability (Fig. 5). A
range of temperatures and Logan2b values were associated with
both high and low tree mortality across and within outbreak locations. Year-to-year increases and decreases in temperature corresponded to increasing and decreasing beetle activity,
respectively, but only in some years during 1990–1995 in Utah
and Idaho; this positive relationship occurred during low tree
mortality. In Idaho in 1993 and in Utah in 1991, very low
March–August temperatures and near zero values of Logan2b occurred together with a substantial decrease in number of trees
killed; low values also occurred in Montana in 1993 when beetle
activity was already low.
Winter minimum temperatures and ColdT were variable
among years and across outbreak locations (Fig. 6). Winter temperatures during outbreaks varied by 5 °C among locations, and
ColdT varied from close to 0 to 0.8. ColdT was nearly always

(a)

150

(b)

Utah: 1997−2009
Colorado: 1996−2007
Idaho: 1998−2002
Montana: 2000−2007
Oregon: 1976−1987

15

Precipitation (mm)

Mean Temperature (C)

10

5

0

100

50

−5

−10

0
J

F

M

A

M

J

J

Month

A

S

O

N

D

J

F

M

A

M

J

J

A

S

O

N

D

Month

Fig. 2. (a) Mean monthly temperature (°C) and (b) precipitation (mm) during outbreaks using 4-km, monthly weather data from Parameter-elevation Regressions on
Independent Slopes Model (PRISM, Daly et al., 1994).

244

E.P. Creeden et al. / Forest Ecology and Management 312 (2014) 239–251

(g)

(a)

(d)
(h)

(e)

(b)

(i)

(c)

(f)

(j)

Fig. 3. Distributions of means of weather variables during years of mountain pine beetle outbreaks. Adaptive seasonality: (a) water year (October–September) mean
temperature (°C); (b) March–August mean temperature (°C); and (c) Logan2b model output. Cold-induced mortality: (d) minimum temperature of coldest winter (DJF) month
(°C); (e) lowest daily mean temperature of water year (°C); and (f) cold tolerance (ColdT, probability of survival). Drought stress: (g) water year precipitation (cm); (h) March–
August precipitation (cm); (i) climatic water deficit (mm); and (j) summer (JJA) Palmer Drought Severity Index.

Table 2
Weather factors relative to preoutbreak years for each location.
Early in outbreak

During peak of outbreak

During decline

Winter
T

Year-round T

Drought

Winter T

Year-round T

Drought

Winter T

Year-round
T

Drought

Utah

Warmer

Warmer

Warmer

Warmer

Drought

N/A

N/A

N/A

Colorado

Warmer

Average to high
Logan2b; higher T

Wetter then multiyear drought
Wetter then multiyear drought

Warmer

Some years
with drought

Warmer

Warmer

Average to
wet

Idaho

Warmer

Average to warmer

Warmer

Drought

Montana

Warmer

Average to warmer

Wetter then multiyear drought
Multi-year drought

Average

Average

Average to dry

Warmer,
then average
Average to
warmer
Average to
cold

Average to
warmer
Average to
colder
Average to
cold

Average and
dry years
Average

Oregon

Average
Logan2b; higher
T
Average to
warmer
Average to
warmer
Average to
warmer

Average to
warmer
Average

above the 0.2 threshold for population growth proposed by Safranyik et al. (2010). ColdT values close to or below 0.2 were
experienced in some years in Utah, Idaho, Montana, and Oregon.
Most of these cases occurred when tree mortality was low and

Average to dry
Wet

Average to
dry

corresponded to decreases in the number of lodgepole pines
killed.
Drought occurred during each of the outbreaks, and for Utah,
Colorado, Idaho, and Montana, occurred during the early 2000s

245

E.P. Creeden et al. / Forest Ecology and Management 312 (2014) 239–251

and corresponded to increases in tree mortality (Fig. 7). The first
year of available beetle information in Oregon (1979) was also
dry. Extreme drought occurred early in the outbreaks: Oregon
experienced its second lowest value of summer PDSI (-4.1) and
lowest water year precipitation (51.1 cm) on record in 1977,
and summer PDSI in 2002 was the lowest on record in Colorado
and second lowest on record in Utah. At high levels of tree mortality, relief from drought seemed to have little to no negative effect on outbreaks. Tree mortality was high in Utah, Colorado,
Montana, and Oregon during average or high PDSI conditions,
and increases in PDSI at times corresponded to increases in tree
mortality.

the last years prior to outbreak decline (1986–1987) and three
out of four years during decline (1988–1989, 1991). ColdT values
were also low in 1989 and 1991, potentially impacting beetle
activity. Lower-than-average values of Logan2b at the Oregon location were recorded in the last years prior to outbreak decline
(1986–1987) and three out of four years during decline (1988–
1989, 1991). At other locations, years of large declines did not
co-occur with weather conditions unfavorable for the beetle. For
example, probability of survival from cold-induced mortality was
high in 2008 and 2009 in Colorado and Montana, and conditions
appeared to be adaptive for the beetle in these years. Drought also
did not appear to play a role in these large declines.

3.4. Changes in weather factors during declines in tree mortality

4. Discussion

Large declines in tree mortality occurred at the Idaho location in
2003–2005, temporarily in Utah from 2004 to 2005 and 2007–
2008, and in Oregon from 1988 to 1991. Outbreaks were declining
in the last years of the study period in Colorado and Montana
(Fig. 1, Table S2). Low values of Logan2b occurred in Oregon in

Our results revealed differences of mean weather factors during
outbreaks among locations. Proximity to oceans and latitude cause
differences in climate among regions dominated by lodgepole pine
in which similar climate might be expected. Oneil (2006) and
Crookston (1977) postulated that mountain pine beetle outbreaks

Normalized num. trees killed

1980

1985

1990

1995

2000

2005

2010

1995

2000

2005

2010

100

UTAH
COLORADO
IDAHO
MONTANA

80
60
40
20

(a)
0
0.5

Logan 2b

0.4
0.3
0.2
0.1

(b)
0.0
0.8

ColdT

0.6

0.4

0.2

(c)
0.0
4

Summer PDSI

2
0
-2
-4

(d)
1980

1985

1990

Fig. 4. Time series of number of trees killed and weather variables for four outbreak locations. (a) Number of trees killed normalized to maximum at each location. (b)
Logan2b, representing adaptive seasonality. (c) Cold tolerance (probability of survival during cold season). (d) Summer Palmer Drought Severity Index, representing drought
stress on hosts.

E.P. Creeden et al. / Forest Ecology and Management 312 (2014) 239–251

in Washington and other Pacific Northwest locations have different
limiting factors related to climate and weather than other areas,
and our analysis also suggests this. An additional explanation for
these differences may be related to genetic variability among beetle populations (locations) in sensitivity to weather factors (Bentz
et al., 2001, 2011a).
Conditions favorable for adaptive seasonality occurred in most
years at the study locations, and conditions during outbreak periods
were often more favorable for outbreaks than in prior years. Maladaptive seasonality resulting from low summer temperatures in
Utah and Idaho in 1993 likely negatively affected outbreaks, agreeing with findings by Logan and Powell (2009) in central Idaho. Temperature variables did not consistently reflect Logan2b values, most
likely because the Logan2b model incorporates daily temperature
data to represent complex, nonlinear life stage development rates
important to beetle development timing and adult emergence.
Other studies have also found that higher temperatures were
associated with mountain pine beetle outbreak probability in
British Columbia (Aukema et al., 2008), the Greater Yellowstone
Ecosystem (Jewett et al., 2011), and Washington/Oregon (Preisler
et al., 2012).
Large differences in mean values of cold-induced mortality variables existed among outbreak locations. Montana and Idaho had
lower values of minimum daily mean temperature and ColdT compared to the other locations, suggesting that cold-induced mortality potentially plays a more dominant role in outbreak regulation

at these locations. However, low temperatures able to induce very
high mortality (i.e., ColdT < 0.2) rarely occurred during periods of
high tree mortality. Years with ColdT values 6 0.2 corresponded
to decreased tree mortality, especially at low outbreak levels. During years of unavailable aerial survey data, we simulated low ColdT
values that matched reported outbreak declines in October 1984 in
Colorado (ColdT = 0.22) (Lessard et al., 1987) and in 1983–1984 in
Montana (ColdT = 0.035) (Kipfmueller et al., 2002). Higher values
of ColdT were at times associated with increases in tree mortality
(e.g., Utah, Colorado, Montana, and Idaho in 2000). Preisler et al.
(2012) and Aukema et al. (2008) also found increased odds of
mountain pine beetle outbreak with higher winter temperatures.
Safranyik (1978) suggested that winter temperatures in much of
the distributional range of the mountain pine beetle are rarely severe enough to reduce populations to endemic levels. However, we
found that cold season mortality likely was a factor in keeping tree
mortality low in several of our locations in the United States (prior
to outbreaks).
Drought was present at some point during all outbreaks, and increases in tree mortality occurred under drought conditions at all
study locations. Dry conditions have previously been associated
with beetle outbreaks (Raffa et al., 2008; Jewett et al., 2011;
Chapman et al., 2012; Preisler et al., 2012). Impacts of drought
were more pronounced at lower outbreak levels than higher levels.
However, severe drought did not occur at higher levels to the extent it did in the early 2000s and so we could not evaluate the

(b)

0805
00
09

03

5

07

06
0805
00
09 9492
99
90
91989796
9395

07

6

01

5

09

0

84
82

6

83

4

00
9190
9798
94
95
92
9396
99

06 03
05 04
02
01
0094
95
99
9397
96991
890
92

84
82
83

85

85

8687

5.5

6

5.0

89

0
4

6

8

10

90
91

4

6

Mar-Aug Tmean (deg C)

8

10

4.5

05 08

03

5

99

04

97
95

06 07

05 08 00
99 969890
95
97
9109

94 92

06 07

00
09

90
98
91
96

4

92

07

7

06 05
04
02

6

08
03
01

5

09

0

03 06
05
04
91 98
97 01
9290
94
9596
9902

93

97
93

00

858482

6

90
91
9596 98
9294
99

6.5
6.0

6

88

5.5

6

5.0

89
88
91

0
0.0

0.1

89

0.2

90

90

0.3

0.4

4
3

858482 83

86

87

00

83

86

87

log10 (NTK)

3

93
09

6

5

94

08

7

6

03

04

07

6.5
6.0

88

01
02

01
02

93

4

94 95

5

3

8687

6

88
91 89 90

03 02
04
01 00
99
97
96
94
9598

5

0

5

9796

7

06 03
05 04
02

01 00
99 98

09 05
06

5

3

08
09

08

6

4

93
07

7

5

90
92
98
94
91 97
95
96

04

0

07

06

99

5

6

6

02

6

0
6

7
03

04

6

94

02 01
03
04

01

08

4.5

91

0.5

0.0

0.1

0.2

log10 (NTK)

02

95

0
07
08
09 05
06

6

4

NTK

03
0402
00
01
9996 94
97
95 98

2

91

07

7

4

0.3

0.4

log10 (NTK)

9796

5

NTK

5

99
98

0905 06

6

NTK

6

02
05
01 9400
990
995
98
97
92
96
91

93

6

99
98 95 94
92
97 90
96

93

04

7

NTK

6

5

MONTANA

0905 06

6

07

7

0403
02
00
01

6

NTK

07

NTK

08

08

6

NTK

0

91

03
06
07
08

5

NTK

NTK

COLORADO

01
00
94
07

0

OREGON

2

02

7
7

94

log10 (NTK)

05
92
919593 9899
909796

4

99
9593 98909792
96

04

5

07

IDAHO

07

08

090603
07
04 08
02
05
01 00

6

log10 (NTK)

6

6

log10 (NTK)

06

03

09

log10 (NTK)

NTK

UTAH

6

06 03 07
04
05 02
01
00

log10 (NTK)

08 09

log10 (NTK)

(a)
09

log10 (NTK)

246

0.5

Logan 2b

Fig. 5. Temporal trajectories of the number of lodgepole pine trees killed (NTK) and log10(NTK) versus weather variables representing adaptive seasonality. (a) Mean March–
August temperature (°C). (b) Logan2b. Higher values indicate conditions more favorable to beetles.

247

E.P. Creeden et al. / Forest Ecology and Management 312 (2014) 239–251

(b)

0708

08

03
04
02
00
01
98 99

6

06 09
05

6

03
04
020198
0099
97
96 95
94

NTK

5

05
06
0807
00
09
99
90
92
98
94
97 95
96

03
04

5

91

91

05
06
0807
00
09
99
92 96
9897
9495
9390

NTK

MONTANA

0

91 93

06 04
03 05
02
01
9000
97
96
94 98
99
95
92

5

93

8687

4

00
90
9794 98
96 9995 92

88

5.5

6

5.0

89
89
91

0
-20

-15

-10

90

4.5

91

-5 -20

-15

-10

02
0706

03
04

91
96

0706
08 05 00
90
92
98
9196 9909 95 9394
97

-5

Tmin coldest DJF (deg C)

6

05

08

5

00

90
92
98
94
97

95

4
3

93

7
6
5

09
0306 05
0292
01
91969904
97 939490
95
98
84

NTK

6

00

84

8285
8687

4

00

90
91
97
98
9699 93949592

3
6.5

8285 83
8687

83

6.0

6

88

5.5

6

5.0

89
88
89
91

0
0.0

0.2

0.4

90

90

0.6

log10 (NTK)

6

0708
09
0306 05
04
02 01

08

7

4

01
03
04

09
99

07

6.0

6

01

02

5

0

5

95
94

5

6.5

83

88
90

03 04
02
01
99 98
9697
95
94 00

5

3

858284 83
8687

858284

6

NTK

6

6

96 97

7

09
91

06 09 05

5

0

7

03 04
02
01 00
99 98

6

3

0807
09
06 03 05
04
01 02

08

6

4

93

07
7

5

0

08

0
6

2

07
08
06 09 05

6
6

4
98
94 9395
9792
90
96

91

07

6

4

01
02
03
04

5

91

7

94 95

01
02

5

OREGON

5

97
96

6

0

6

02
05
99 90
94
9800
9792
9395
9601

7

NTK

6

04

07

7

06 09
05

NTK

NTK

COLORADO

7

0

IDAHO

0

91

07

7

99

08

5

NTK

91

7

log10 (NTK)

0

05
01
00
99
94
939798
9596
9092

6

6

4.5

91

0.8

1.0

0.0

0.2

0.4

log10 (NTK)

2

03
06
07

log10 (NTK)

96

02

NTK

04

5

4

99

log10 (NTK)

94
939798
909295

log10 (NTK)

6

09
0307
06
04 08
0205
01 00

6

log10 (NTK)

03
06
07
08

log10 (NTK)

NTK

UTAH

6

09

6

0.6

log10 (NTK)

03
0609
080704
02
05
01 00

log10 (NTK)

(a)
09

0.8

1.0

ColdT

Fig. 6. Temporal trajectories of the number of lodgepole pine trees killed (NTK) and log10(NTK) versus weather variables representing winter mortality. (a) Lowest monthly
minimum temperature December–February (°C) and (b) cold tolerance (ColdT, probability of survival); dashed vertical line indicates threshold below which ColdT has an
effect on beetle population growth (Safranyik et al., 2010). Higher values indicate conditions more favorable to beetles.

impact of severe drought at peak outbreak levels. Because mountain pine beetles can kill larger and more vigorous trees with more
defensive capacity at higher populations (Boone et al., 2011), the
role of host stress from various factors, including drought, may
be diminished during that population phase.
At high levels of tree mortality, continued drought was not
necessary to maintain outbreaks. A return to average precipitation levels and presumably lower host stress showed no apparent
negative effects on increasing tree mortality in Colorado and
Montana in the mid to late 2000s, a finding similar to that found
by Chapman et al. (2012) in Colorado and Wyoming. Indeed, the
role of drought stress may switch at high levels of tree mortality.
The number of emerging brood has been found to increase with
phloem thickness (Amman, 1972), potentially facilitating increased reproduction and brood survival in unstressed trees.
Thus, higher precipitation could provide a greater supply of trees
with thicker phloem able to support more brood and, and during
larger outbreaks, these trees would not need to be stressed by
drought to be successfully attacked (Amman and Cole, 1983; Safranyik and Carroll, 2006).
The warming and drought in the early 2000s likely played roles
in increasing tree mortality in Utah, Colorado, Idaho, and Montana.
However, the difference of several years in the timing of large increases among these locations suggests that other factors or

conditions also influenced outbreaks. Nonlinear dynamics, different timing of threshold breaching, and other factors such as populations dynamics during endemic phases or interactions with
natural enemies, each undetectable with our data sets, may have
played roles (Safranyik and Carroll, 2006; Raffa et al., 2008).
We did not find strong evidence of large outbreak declines to be
related to weather factors. Slightly lower values of Logan2b occurred in 2004 and 2005 in Idaho and 2009 in Montana, and the
decline in Oregon was also potentially influenced by low values
of adaptive seasonality (Logan2b) and ColdT. These results suggest
other factors play a larger role in decreased tree mortality at these
locations, including suitable host availability (food resources), as
has been suggested for the Colorado and Montana locations
(Man, 2010; Chapman et al., 2012).
Climate change in recent decades has been implicated as
responsible for several bark beetle outbreaks in western North
America (Breshears et al., 2005; Sherriff et al., 2011), including
the northern (Carroll et al., 2004a; Sambaraju et al., 2012) and
upper (Logan et al., 2010) range limits of mountain pine beetle.
Here we found that warming and drought in recent decades were
associated with increases in beetle-caused tree mortality within its
historical range. Temperatures in recent years have increased by
1–2 °C at outbreak locations compared with 1900–1929 means,
and the number of years with cold winters has decreased


Related documents


climate weather and recent
41i16 ijaet0916831 v6 iss4 1803to1812
ips typopgraphus
2011 colinet et al jtb
2004 legrand colinet et al aab
nature14283 1


Related keywords