Climate Change Could Alter the Distributions of Mountain Pine Beetle Outbreaks in Western Canada

The mountain pine beetle Dendroctonus ponderosae is an eruptive, tree-killing bark beetle that causes large-scale pine tree mortality in the forests of North America during epidemic conditions. Beetles attack en masse and, along with vectored fungi, overwhelm pine trees’ defensive capacity. Usually, extremely cold winter temperatures limit the spread of mountain pine beetles to northern latitudes by causing widespread beetle mortality. However, climate change may transform previously unsuitable habitats to suitable ones as temperature increases lead to milder winters and warmer summers. Despite the importance of extreme cold temperatures in limiting eruptive beetle populations, few landscape level models have examined how cold temperature regimes and future climatic variability will limit or enhance beetle outbreaks. Using a spatiotemporal statistical model framework, Sambaraju et al. (2011) examined the relationship between elevation and temperature and the occurrence of mountain pine beetle outbreaks. Additionally, they investigated the spatial outbreak patterns of the beetle under a combination of four simulated climate change and two climatic variability scenarios using adapted terms from the model and data from a peak outbreak year. The authors found that timing, frequency, and duration of cold snaps had a severe negative relationship with occurrence of an outbreak in a given area. Drops in temperature and extreme winter minimum temperatures reduced outbreak probability. Increases in mean temperature by 1°C to 4°C increased the risk of outbreaks, with the effects manifesting first at higher elevations, then at increasing latitudes. Increasing the variance associated with mean temperature did not change the trend in outbreak potential. These results demonstrate how climate change could result in a higher frequency of mountain pine beetle outbreaks, ultimately causing higher rates of pine forest mortality.—Megan Smith
Sambaraju, K.R., Carroll, A. L. Zhu, J., Stahl, K., Moore, R.D., Aukema, B.H. 2011. Climate Change Could Alter the Distribution of Mountain Pine Beetle Outbreaks in Western Canada. Ecography 34: 211–213. DOI: 10.1111/j.1600-0587.2011.06847.x

              Mountain pine beetle range expands from northern Mexico through the Pacific Northwest to northwestern Alberta in Canada. Beetle communities establish at elevations of 1500–2600 m above sea level. In Canada, the lodgepole pine Pinus contorta var. latifolia, is the predominant host; however other species such as ponderosa pine P. ponderosa, western white pine P. monticola, and jack pine P. banksiana, are also attacked. The beetle is univoltine over most of its range, but its life cycle can extend to two years (semivoltinism) at elevations greater than 2600 m due to sub-optimal temperatures for insect growth. Adults emerge from brood trees in mid-July and mid-August, and then attack trees en masse. A high beetle attack rate, along with vectored fungi, causes tree mortality. There are four instars in the beetle life cycle, and the third and fourth instars are commonly exposed to the coldest weather between December and February. The beetles are capable of surviving winter temperatures in the range of –34°C to –37°C. However, frequent and/or prolonged occurrences of severe winter temperatures can cause widespread beetle mortality by killing off a large portion of the immature beetles. For example, cold temperature extremes played a key role in the collapse of a mountain pine beetle outbreak in the Chilcotin area of central British Columbia during 1984–1985.
              This study investigated the association of temperature and elevation with the occurrence of mountain pine beetle outbreaks in western Canada from 1992 to 2007. It also examined the spatial outbreak patterns of the beetle under four simulated climate change and two climatic variability scenarios.
              The authors generated a grid covering the Canadian provinces of British Columbia, Alberta, Saskatchewan, and portions of the territories of the Yukon, and Northwest territories in ArcMap. This grid was split into cells that measured 0.1° latitude X 0.2° longitude. These cells were approximately 12 X 12 km in size and 17,063 were created.
              Then, the authors obtained aerial survey data sets of red attack (trees that had been killed by mountain pine beetle in the previous year) for the province of British Columbia from Forest Insect and Disease Survey (FIDS) of the Canadian Forest Service for 1990–1996 and from British Columbia Ministry of Forestry and Range for 1999–2007. The locations and extent of tree mortality had been mapped on a 1:250,000 NTS map, and the maps were digitized into shapefiles using ArcGIS. The yearly infestation shapefiles were overlaid on the study grid for each year. Sambaraju et al. used a binary variable (presence = 1, absence = 0) of a tree-killing population of mountain pine beetle per cell, as a response variable in their study. A map displaying the study grid was constructed.
              Daily minimum and daily maximum temperatures were derived for each grid cell by interpolation from station observations from 1990 to 2006. The database of observations included data from 1974 stations assembled from the climate station network of Environment Canada and the networks of fire weather stations maintained by the Forest Service in each province. Elevation distributions for each grid cell were extracted from the higher resolution DEM Can3d30. Due to the complex topography (large elevation ranges within grid cells) temperatures were interpolated to the center of each grid cell using the 10thpercentile elevation.
              Sambaraju et al. used logistic regression to analyze the relationship between the explanatory variables and the occurrence of mountain pine beetle outbreaks in each cell using PROC LOGISTIC (SAS 2008). Cold temperature variables included cold snaps (a minimum of four continuous days of average winter temperatures at or below –20°C), sudden temperature variations, and extreme minimum temperatures during winter. Other temperature and degree-day terms were extracted from previous models. Analysis was limited to neighboring spatial structures and occurrences of infestations in previous years for a given cell. Each spatial and temporal variable was fit in combination with the others to find the best spatial-temporal dependence structure. Models were selected based on the Akaike information criterion (AIC) score, with a lower AIC value indicating a better fit. The influence of individual temperature terms on outbreak probability was then studied by including a single variable along with the best fitting combination of spatial and temporal terms. All the covariates were included in a multivariable logistic regression along with the best fitting spatial and temporal terms. A model with the best subset of variables was selected by a backward selection method. A table displaying temperature-related terms, spatial neighborhood structures, and temporal infestation terms used in the spatial-temporal logistic regression models was constructed.
              The authors selected four climate change scenarios of temperature increases of 0°C, 1°C, 2°C, and 4°C. To assess whether variability in the climate change scenarios influenced outbreak trend, they also selected two scenarios where the variance associated with a given mean temperature increase was 1°C or 2°C. These simulations were conducted with outbreak data from 2005. Climate change was simulated by adding non-overlapping streams of random numbers from a Gaussian distribution of known mean and variance to the daily maximum and minimum temperature data for years 2003 and 2004 for the entire study area. This altered temperature data set was used to redefine the temperature covariates that were then run through the multivariable spatial-temporal logistic model. The logits were back-transformed to generate probabilities of outbreak in a given cell.
              Results were analyzed using two methods. The first grouped cells into five classes from very low to very high outbreak probabilities. The mean numbers of cells in each outbreak class for one hundred simulations for each combination of mean temperature increase and variability scenario were calculated. Regression analyses were performed. The second approach calculated the median probability for each cell across one hundred simulations for the four mean temperature increase scenarios. Median probabilities were also grouped under the five outbreak probability classes. Then the authors evaluated changes in the outbreak class for each cell under 1°C, 2°C, and 4°C compared to the 0°C mean temperature increase scenario. Cells were mapped if a change in outbreak class occurred to identify their locations in terms of whether changes occurred at elevations 1500 m or at latitudes further north of the current beetle range. The authors graphed the number of cells and percentage of total cells per temperature increase scenario whose outbreak risk increased or decreased based on elevation.
              Sambaraju et al. found that the likelihood of finding an outbreaking population of mountain pine beetle killing trees in a given area increased with occurrences of patches of dead trees in the same area in the past three years, as well as in the surrounding 18–22 km in the same year. All spatial and temporal covariates had a positive association with outbreak likelihoods.
              Cold snaps or periods of four consecutive days with winter average temperatures below –20°C decreased the outbreak probability with a cell. Occurrence of early cold snaps in October and November had the largest negative impact on beetle populations, more so than cold temperature episodes occurring in the spring. However, temperatures did not have to drop below –20°C for cold temperatures to have a negative influence on beetle populations. Increasing frequencies of sudden drops in average temperatures of >10°C during consecutive days in winter negatively impacted beetle populations. Increases in temperature or day-to-day declines in temperatures of 0–5°C did not decrease the outbreak probability. Extreme minimum temperatures of –30°C or below decreased the odds of an outbreak response within a cell. A table displaying the maximum likelihood estimates of cold temperature coefficients in spatial-temporal logistic regression models was constructed.
              Outbreak probabilities increased with higher maximum, minimum and mean annual temperatures, average summer temperatures, and with sufficient accumulation of degree-days for 50% egg hatch and adult emergence. Outbreak probabilities also decreased at the lowest and highest elevations across the study area relative to the mean elevation.
              A total of 21 variables were included in the final model and most maintained the same sign as individual analyses. All cold-temperature terms selected by the model reduced the probability of mountain pine beetle outbreaks. Increases in maximum temperature, mean temperature, and annual temperature change led to an increased probability of outbreaks. Furthermore, successful accumulation of degree-days for a univoltine life cycle for mountain pine beetle strongly increases the probability of an outbreak. Graphs displaying the odds ratio estimates for cold temperature covariates were constructed.
              The outbreak trends under the different climate change and climate variability scenarios demonstrate that increase in average temperatures by 1°C or more caused a decrease in the mean number of cells where the probability of an outbreak was very low. An increase in temperatures by 2°C and 4°C increased the mean number of cells in the ‘medium’ risk class. The highest outbreak potential occurred with mean increases of 2°C, but decreased above and below this value. Increasing the variability associated with a mean temperature did not cause large differences in the mean number of outbreak cells under the different probability classes.
              Similar results were observed after examining the median outbreak probability per cell. Increased temperatures increased the risk of outbreaks, especially at higher elevations. At an increase of 1°C, a significant percentage of the higher risk cells occurred at the periphery of the existing outbreak in areas of higher elevation. Similar results were seen with a mean increase of 2°C and 4°C. However, increasing outbreak probability was manifested at more northern latitudes in addition to higher elevations at 4°C. With an increase of 4°C over a year, there was a 15% increase in outbreak potential at 120.8°W/58°N, which is 25 km north of the current range in British Columbia. Increasing mean temperatures caused increases in the numbers of higher risk cells. The percentage of reduced risk cells occurring at elevations 1500 m was higher for 1°C. Graphs displaying the mean outbreak potential in response to simulated mean increases in temperature were constructed, as were maps displaying elevational and latitudinal changes in median outbreak probabilities. Finally, graphs displaying the number of cells, and percentages of total cells per temperature increase scenario, showing increased or decreased risk of mountain pine beetle outbreak under climate change scenarios were constructed. 
              Overall, these results demonstrate that temperature effects on the tree-killing behavior of mountain pine beetle are manifested through winter weather patterns. Extreme cold temperatures and sudden temperature drops limit insect populations. However, climate change may transform previously unsuitable habitat into temperature regimes conducive to beetle population success. For example, forested regions at high elevations and latitudes further north of the current range of the mountain pine beetle in British Columbia could become outbreak-prone due to climate change. Small increases in temperature also resulted in occurrences of new areas of outbreaks first at higher elevations, and then at northern latitudes. Yet, if temperatures reach a specific maximum threshold, hotter habitats could also decrease outbreak probabilities. 

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s