Understory Vegetation Response to Mountain Pine Beetle-Induced Lodgepole Pine Mortality in Rocky Mountain National Park, Colorado Introduction This study characterizes the response of understory vegetation to mountain pine beetle (MPB)-induced overstory tree mortality across lodgepole pine-dominated forests by measuring percent plant cover, richness, diversity, functional diversity, community composition, new tree seedling establishment, and growth release of surviving tree seedlings. By resampling plots that were established in 2008, I was able to carefully measure the changes in understory vegetation over a five-year period following peak bark beetle activity. This semester I attempted to utilize the program R to addresses three main questions: 1) How does the change in understory vegetation cover differ by plant life form (i.e., shrub, forb, graminoid) and how does this vary across forest type? 2) What is the relationship between changes in overstory canopy cover and changes in understory vegetation cover? 3) What is the extent, location, and composition of new seedling establishment and how is this related to overstory and understory plant cover? Methods of Data Collection I resampled 38 sites throughout lodgepole pine-dominated forests west of the Continental Divide in Rocky Mountain National Park, that were established using a spatially-balanced random sampling design. Each site contains two 20 m x 20 m square plots randomly positioned 90 m apart, for a total of 73 plots. In each plot, and along three transects running north-south at 5, 10, and 15 m, I placed five 1m 2 quadrats, for a total of 15 per plot. In each quadrat, I identified each plant species present, recorded the number and height of each tree seedling for each species, and measured percent cover of graminoids, forbs, shrubs, tree seedlings, and various other abiotic substrates. In addition to the quadrat seedling measurements, I identified and counted all new tree seedlings ( 5 years old) and re-measured heights of tree seedlings mapped in 2008 within a 2 m x 20 m belt transect. Five densiometer measurements were collected from designated locations to estimate tree canopy cover. Analyses in R (in the same order as performed in the script) Much time was spent initially using R to calculate means and changes in percent cover and densiometer measurements, and to create and merge data frames for plotting and further analysis. None of these procedures are included in the script. The code starts off by creating a data frame containing mean percent cover data and standard deviations to make a barplot with. This barplot shows the change in plant cover by life form (figure 1). Included in the script is code for a barplot with standard deviation and standard error bars. A table with Ratio of Means values for % cover between years is created and written to a table. This is the relative "total amount" of change across the landscape, but probably not valid
since the percent cover values are not normally distributed. Seedling data summaries are also written to table. Both of these tables were used to make other tables in Excel. Transformations of percent cover are performed to better satisfy assumptions. Logit and arcsine square root transformations are used since they are appropriate for percentage data (log with added constant transformation was also performed but this is not included in the script). To see if any relationships in change in cover exist between life forms, a multipanel scatterplot of the differences by lifeform (2013-2008) is created with histograms on the diagonal, (absolute) correlations on the lower panels and size proportional to the correlations (figure 2). These are also created using the transformed values. Transformations appear to have very little effect on the relationships. Bland-Altman plots are created to check that the differences in % cover between years have a distribution that is independent of the level. These distributions do not seem independent of the level (figure 3). A Bland-Altman plot is created for graminoids using the logit transformed values. Paired t-tests are performed for the different lifeforms. Difference between years is only significant for tree seedling cover (p-value = 0.01417), which increased. Also, the t-test for overstory canopy cover (decrease) is significant at p-value = 0.03101. Regression is used to explore the relationship between changes in overstory and understory plant cover, and between both overstory and understory cover and new seedling density. No significant relationships (p-value < 0.05) are found. See figure 4 for an example of the regression of shrub cover modeled as a function of tree canopy cover. ANOVA is used to test for differences across forest types in total understory cover, understory cover by life form, and new seedling establishment. Again, no significant relationships emerge, however, the ANOVA for number of new lodgepole seedlings modeled as a function of forest type has a p-value = 0.08, but also exhibits unequal variance (figure 5). Tukey's post-hoc comparison shows plots 6 and 2 and 1 and 2 are the most different with p-values around 0.1. Figures Figure 1
Figure 2 Figure 3 Figure 4
Figure 5