Impact of Summer Weather on Beetle Activity: Degree-Days & Activity Index in Kenai Peninsu, Papers of Geography

Download Impact of Summer Weather on Beetle Activity: Degree-Days & Activity Index in Kenai Peninsu and more Papers Geography in PDF only on Docsity! The Relationship Between Prior Summer Warmth and Spruce Bark Beetle (Dendroctonus rufipennis) Activity During a Large Epidemic on the Kenai Peninsula, Alaska Joshua Mumm. May 24th, 2006 Abstract This study substantiates the widely accepted theory that summer warmth is an important determinant of future spruce bark beetle activity. The relationship between the first difference of accumulated degree-days above 15.6 oC from January to the end of June during year t - 2 and the first difference of a beetle activity index (BAI) during year t, was analyzed with simple linear regression for the years 1991 to 2004 during a large epidemic on the Kenai Peninsula. BAI was defined as the percentage of potential habitat that was actually exploited by the beetles. Potential habitat for year t was defined as conifer or mixed forest within a 300m dispersal distance of beetle damage in either of the previous two years. BAI was calculated with a GIS from a landcover classification and annual aerial forest damage surveys. The first difference of BAI during year t was significantly determined by the first difference of degree-days through June during year t-2, with large increases in BAI occurring two years after large increases in degree- days. The strength of the correlation was greater during the peak of the epidemic (r2 =.878, n=7), than for the entire time span (r2 = .341, n= 13). Removing only the highly anomalous 1999 data from the entire time span produced an intermediate coefficient of determination (r2 = .664, n=11). The regression is attributed to the surge in emergent adults associated with a shortening of the beetles maturation time. Introduction Spruce bark beetles, whether D. rufipennis in North America or Ips Typographus in northern Europe, serve important ecological roles and are endemic throughout the world’s boreal forests. Given a plentiful supply of mature forest and warm dry summers, however the insects erupt as massive epidemics. Such ideal conditions occurred on Alaska’s Kenai Peninsula in the late 20th century, where annual average temperatures increased by 1.24 oC from 1971 to 2000 and precipitation dropped to record lows (Reynolds and Holsten 1994, Harding et al. 2002). With host and climate requirements satisfied, spruce bark beetles swept across the peninsula in one of North America’s most devastating forest insect outbreaks. More than 4 million acres of forest were infested from 1989 to 2005, and 80% of trees in infested areas were killed (Figure 1) (Wittwer et al 2002, Berg et al. 2000). 1 The importance of warm dry summer to both the initiation of an epidemic and inter-year variation in activity levels is widely accepted and supported by anecdotal evidence (Wemelinger and Seifert 1999, Berg 2000, Wittwer et al. 2002). Research has revealed five mechanistic explanations of how warm dry summers benefits beetle populations: 1) Warm summer weather shortens the beetles lifecycle from semivoltism to univoltism (figure 2) (Hansen et al 2001); 2) Low summertime rain stresses spruce and compromises their ability to defend themselves by excreting the beetles in large amounts of sap (Reynolds and Holsten 1994); 3) Warm springs allow the beetles to emerge earlier and attack trees when their roots are still frozen and their defenses down (Reynolds and Holsten 1994); 4) Warm dry springs increase the proportion of females that take flight(Franklin and Gregoire 1999); and 5) Warm springs increase the rate of oviposition (Wermilinger and Sefreit 1999). Despite the theorized importance of warm weather on beetle activity, and the preponderance of mechanistic explanations, no landscape level empirical evidence has been published to show that warm dry weather does in fact increase beetle activity. The present study addresses this dearth of empirical evidence by investigating the relationship between degree-days through June during year t-2 and beetle activity during year t, during the Kenai Peninsula spruce beetle epidemic. The first listed benefit of warm weather on beetle activity, the shift from semivoltinism to monovoltinism actually bolsters beetle activity through two independent mechanisms. The short term effect of a shift from semivoltism to univolism is an immediate surge in the number of emergent adults. The longer term effect is a doubling of the reproductive rate, caused by a 50% reduction in maturation time. I expected the first mechanism, the immediate surge in emergent adults to be the most important determinant of year-to-year variations in beetle activity. With this in mind, I designed the study to focus on detecting this phenomenon. Specifically, this 2 resolution. This raster was projected to the same projection as the forest damage data and clipped to the extent of the study area. Climate records containing daily maximum and minimum temperatures observed at the Homer Airport spanning 1985 to 2005 were downloaded from the National NOAA Data Center as ASCII files for individual years. Analysis Three components comprised the analysis. First, a metric of summer temperature was defined and calculated for each of the 15 years. Second, a metric of beetle activity was defined and calculated for each year. Third, the relationship between the metric of prior summer warmth and the metric of beetle activity was analyzed. Cumulative degree-days above 15.6 oC (60oF) from January through the end of June two years prior to the subject year was the metric for prior summer warmth. Degree-days in the prior June was suggested by a group of spruce bark beetle experts as an important risk factor for spruce beetle damage (Reynolds and Holsten 1994). A lag of two years between weather and observed beetle activity was used because two years is the reported span between initial attack and observable red tip damage (Schmid and Frye 1977, Berg 2002). 15.6 o C was chosen as the lower threshold for calculating degree-days, because this is a good predictor of the ratio of semivoltine to univoltine individuals and is the temperature threshold required for spring flight (Hansen et al. 2001, Holsten and Hard 2001). Degree-days above 15.6 oC through the end of June was calculated for each year by reformatting the ASCII climate records and feeding them into the UC Davis online degree-day calculator using the double sine method. A Beetle Activity Index (BAI) was defined and used as my metric of beetle activity. The obvious metric for beetle activity is the total area of new red tip damage for each year. Simple 5 sum of new red tip damage however, is an unacceptable metric of beetle activity, because it does not account for the changing availability of host trees within the dispersal distance of active infestations as the epidemic progressed. During the early years of the epidemic, availability of host material increased, because the perimeter of active infestations increased, while in the latter stages of the epidemic, availability of host material decreased, because most of the conifer forests on the peninsula had previously been killed (figure 1). This changing availability of host material creates a pronounced hump in red tip damage plotted over the span of the epidemic (figure 5). Changing availability of host material is the dominant determinant of total red tip damage. This renders total red tip damage an inappropriate measure of year-to-year variation s in beetle activity caused by factors other than availability of host material. Instead of total new red tip damage, I used BAI to quantify beetle activity, because unlike a simple sum of red tip damage, BAI standardizes beetle activity to the account for the for the changing availability of host material. BAI is the percentage of potential habitat area actually exploited by the beetles in a given year. Potential habitat area is total area of conifer of mixed forest within the dispersal distance of beetle activity in either of the two previous years. Exploited habitat is the area within the potential habitat that actually shows new red tip damage. Potential and exploited area were calculated for each of the 15 years with a GIS. First polygons containing the areas within the dispersal distance of prior infestations and polygons containing the areas within the dispersal distance of prior infestations that were actually exploited were created for each year with a data model (figure 3). The model first buffers the union of the damage for individual years t-2 and t-1 to create the dispersal distance polygon for year t. A buffer distance of 300m was used to represent the dispersal distance, as suggested by 6 mark recapture experiments (Werner and Holsten 1997). The exploited polygon is the spatial intersection of the dispersal distance polygon and new observed damage. After the potential and exploited polygons were created for each of the 15 years, the area of conifer or mixed forest within each of these polygons was calculated. The area of each cover class was found by inputting the landcover raster and, first the dispersal distance, then the exploited polygons into the Tabulate Area tool of the Spatial Analyst extension. The Tabulate Area tool outputs a .dbf file containing the number of pixels of each cover type within the input polygon. These 30 .dbf files -15 for pixels within the dispersal distance, and 15 for pixels within the exploited polygon – were compiled into a single spreadsheet. Finally, to calculate BAI for each year, the sum of the mixed or conifer forest area within the dispersal distance polygon for each year was simply divided by the sum of mixed or conifer forest area within the exploited polygon for that year and multiplied by 100. The relationship between cumulative degree-days through June during year t-2 and BAI during year t was subjectively investigated by plotting both variables with respect to time. The significance of a regression between the first differences of both variables was objectively tested with simple linear regression using ANOVA. Alpha was set at .05 prior to analysis. Additionally, simple linear regression was performed on the absolute values of both variables. Results The first difference of degree-days through June during year t-2 had a significant effect on the first difference of BAI during year t (table 1). This regression can be seen by visually comparing the two variables plotted with respect to time (figure 6). The degree-day and BAI lines are similar, with local maxima and minima in degree-days during year t-2 generally 7 1999 bolstered the trees defenses. The Kenai Peninsula experienced a drought from 1989 to 1997, but in 1998 and 1999 precipitation returned to or exceeded average levels (Berg 2000). The termination of the drought in 1998 may also explain why the correlation between first difference of degree-days and beetle activity is stronger from 1991 to 1998, than from 1999 to 2004. Perhaps the return of precipitation complicated the relationship between summer warmth and beetle activity seen during the drought. The weakening correlation between difference of degree-days and BAI after 1998 might also have been related to the shift from epidemic to endemic levels of total red tip damage in 1998. The total area of red tip damage was much lower from 1998 to 2004, than from 1991 to 1998 (Figure 5). Perhaps the pattern seen during the peak of the epidemic was less apparent during the latter endemic period, because beetle activity was not sufficiently widespread during the latter period to exhibit the patterns. Also, during the latter time span, beetle activity was limited mostly to the more complicated and less homogenous environments of isolated mountain valleys, and coastal areas (Figure 1). BAI successfully standardized beetle activity to account for the changing availability of suitable habitat. This is seen by comparing the pronounced hump in the total red tip area (figure 5) to less trended beetle activity index (figure6). Calculating BAI would not be possible without GIS, further illustrating the power of the technology to reveal and analyze patterns in large amounts of spatial information. One potential non-GIS alternative to BAI as a means of detrending the total red tip damage might be the residuals of actual total red tip damage from a polynomial model representing the general rise and fall in total red tip damage. Pundits may criticize my analysis of regression in the two partial time spans as looking too hard for a regression. However, the regression is significant even when the entire dataset is considered. I included an analysis of regression in the partial data sets only to show how 10 anomalous 1999 was and how much stronger the regression during the peak of the epidemic, than during the latter endemic period. Future research should investigate the importance of other variables on beetle activity, especially precipitation and the other risk factors suggested by Reynolds and Holsten (1994). A multimodel selection approach based on Akaike’s Information Criteria would be an ideal method of evaluating the relative importance of the many potential predictive variables to predicting future beetle activity. An exciting and tremendously useful extension would be to incorporate as many significant variables future research reveals with climate and past red tip distributions into a GIS model that would predict future red tip distributions. A model capable of forecasting beetle activity predictor model would be valuable to foresters and property owners planning responses to future epidemics. Although the Kenai epidemic has largely run its course, we should apply the information it revealed about the progression of beetle infestations towards planning effective responses to the developing epidemics on western Cook Inlet and the Copper River Valley. 11 12 Figure 3. A generic version of the ArcInfo data model used to create the cumulative and new damage shapefiles. The algorithm was repeated for each of the 15 years. Figure 4. A generic version of the ArcInfo data model used to create the dispersal distance and exploited polygons. The algorithm was repeated for each of the 15 years. The dispersal distance and exploited polygons were used to calculate BAI. 15 0 100000 200000 300000 400000 500000 600000 1988 1990 1992 1994 1996 1998 2000 2002 2004 2006 year to ta l re d t ip d am ag e (a cr es ) Figure 5. Total spruce bark beetle red tip damage in study area by year. The period from 1990 to 1998 represents epidemic activity while the span from 1998 shows endemic activity. The general rise and fall is attributed to changing availability of suitable habitat within the dispersal distance of beetle populations as the epidemic progressed. The goal of the beetle activity index was to remove this general trend so that variations in beetle activity due to factors other than availability of host material could be investigated. Data was compiled from annual forest damage surveys. 0.000 5.000 10.000 15.000 20.000 25.000 30.000 35.000 1990 1992 1994 1996 1998 2000 2002 2004 2006 year B A I O R D eg re e- d ay s BAI during year t degree-days during year t-2 Figure 6. Beetle Activity Index (BAI) and degree-days through June two years ago. BAI is the percentage of coniferous or mixed forest within 300m of red tip damage in either of the two previous years that is red tip damage in the given year. Degree-days are the cumulative degree-days above 15.6 oC through June two years ago. Local maxima and minima in beetle activity generally line up with local maxima and minima in degree-days. 1999 is a major exception with record degree-days and only moderate beetle activity. 16 y = 0.4377x - 1.9161 R2 = 0.3408 -25.000 -20.000 -15.000 -10.000 -5.000 0.000 5.000 10.000 15.000 20.000 25.000 -25.000 -20.000 -15.000 -10.000 -5.000 0.000 5.000 10.000 15.000 20.000 25.000 First difference of degree days through June (t-2) F ir st d if fe re n ce o f B ee tl e A ct iv it y Figure 7. First difference of degree-days through June two years ago vs. first difference of beetle activity for the entire time series 1991 to 2004. First difference of degree-days above 15.6 oC from January through June during year t-2 significantly determined the first difference of beetle activity during year t. The outlier second from the right represents 1999 and the outlier furthest to the left represents 2000, both resulting from record degree-days in 1999, but only moderate beetle activity. y = 0.9245x - 3.0484 R2 = 0.6642 -25.000 -20.000 -15.000 -10.000 -5.000 0.000 5.000 10.000 15.000 20.000 25.000 -25.000 -20.000 -15.000 -10.000 -5.000 0.000 5.000 10.000 15.000 20.000 25.000 First difference of degree days through June (t-2) F ir st d if fe re n ce o f B ee tl e A ct iv it y Figure 8. First difference of degree-days through June two years ago vs. first difference of beetle activity with the anomalies from 1999 omitted. 1999 affected the first differences for both 1999 and 2000. Removing the 1999 and 2000 outliers greatly increased the strength of the correlation. 17