Ecotopic 3: Bison for restoration of an oak savanna ecosystem

README

          This report examines the potential of bison, in addition to controlled burns, to assist in restoration of an oak-savanna ecosystem. Data on biomass of various plant species inside and outside a bison enclosure area, as well as data on oak sapling height was collected by researchers at the University of Minnesota’s Cedar Creek Ecosystem Science Reserve, who clipped strips from each plot, sorted them by species, dried, and weighed them.
grazed: can the plot be grazed by bison
species: the plant species or group
mass: dried mass of the sample, in grams
height: height of oak sapling, in centimeters
burn_unit: which burn unit the plot is in
type: category of plant type as either gram, forb, wood, or litter
gram: short for gramminoid— a grassy species
forb: an herbaceous flowering plant other than a grass
wood: plants with woody stems, like trees
litter: dead plant material collected from the ground

Introduction

          Oak-savanna, an ecological landscape that has components of both forest and prairie ecosystems, is one of the most biodiverse and also one of the rarest ecosystems globally. Oak-savanna once prevailed over the American landscape thanks to regular burning from Indigenous people to maintain areas of openness; bison grazing may also have played a role in bolstering biodiversity. Since European settlement, many of these areas have become cleared for cropland or pasture, or were forested when fires ceased to burn as control of land was seized from Indigenous people. Bison have been shown to increase biodiversity in grassland ecosystems by preferentially grazing on species that would otherwise dominate the landscape. It has been shown that bison prefer to graze on gramminoid, or grassy species, allowing the forbs and woody species to flourish; studies have also demonstrated that grazed patches left by bison sprang back more quickly after the land was burned, since their grazing limited fuel available for the fire. Additionally, the Sandhill Wildlife Area in Wisconsin has included bison as a strategy for oak savanna restoration since 1946, and data suggests the restoration has been successful in preserving the federally endangered Karner Blue butterfly. At Cedar Creek, controlled burns happening for the last fifty years have helped restore the oak savanna ecosystem, but oaks are not surviving to maturity due to intensity of burns. If bison behaved similarly in an oak-savanna environment, oak saplings may have a better chance of survival due to reduced competition from dominant grasses and less intense burns.

Research Questions

Is bison grazing an effective strategy for restoring oak-savannah ecosystems?

          Do bison grazing help the growth of oak saplings?
          Does grazing have an effect on the growth of other plants?
          Do burned areas with grazing bison have more plant biodiversity than unburned areas with grazing bison?

Hypothesis

Bison would likely eat plants that are more dominant within the landscape and therefore would encourage oak growth when the landscape is being burned. Grazing species are typically beneficial to the ecosystem. However, it depends on the amount of bison because overgrazing could lead to desertification.

library(ggplot2)
library(dplyr)
library(lubridate)
library(grid)
 setwd("~/Documents/Fall 2021/bigdata")
percent.species.cover <- read.csv("percent.species.cover.csv")
oak.sampling.growth <- read.csv("Oak Sampling Growth.csv")
biomass.with.type <- read.csv("speciesorder.3.csv")

Data Visualization and Statistical Analysis

Plot 1

Plot 1

        Boxplots show height of oak saplings in groups of grazed and non-grazed and in spring and fall. The median height appears to be taller in fall than in spring, though differences between grazed an non-grazed plots are not apparent in the boxplots. Results of a 2-way ANOVA test suggest that the difference in heights between spring and fall are statistically significant, and the difference in height between grazed and non-grazed plots is not statistically significant; the test also suggests that interaction between grazing and season do not have a statistically significant effect on oak sapling height.
#2 way ANOVA is a stats test that can compare groups of groups (ie, grazed/not grazed is a group, and fall/spring is a group)
graze_and_season_interaction <- aov(height ~ grazed * season, data = oak.sampling.growth)
summary(graze_and_season_interaction)
##                 Df Sum Sq Mean Sq F value   Pr(>F)    
## grazed           1      8       8   0.049    0.825    
## season           1   5061    5061  29.728 5.94e-08 ***
## grazed:season    1      4       4   0.025    0.875    
## Residuals     1307 222497     170                     
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 9 observations deleted due to missingness

Plot 2

biomass.with.type%>%
  filter(mass < 50) %>% 
  ggplot(aes(x=type, y=mass, color=grazed)) +
  geom_boxplot() +
  labs(x="Biomass Categroy", y= "Mass (grams)", title = "Mass of Grazed and Ungrazed Plants by Type") +
  theme_minimal() +
  scale_color_brewer(palette="Accent")

grazed_and_type <- aov(mass ~ grazed * type, data = biomass.with.type)
summary(grazed_and_type)
##               Df  Sum Sq Mean Sq F value Pr(>F)    
## grazed         1    1585    1585   2.424  0.120    
## type           3  629866  209955 321.019 <2e-16 ***
## grazed:type    3    1505     502   0.767  0.513    
## Residuals   1778 1162861     654                   
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
TukeyHSD(x=grazed_and_type,'type')
##   Tukey multiple comparisons of means
##     95% family-wise confidence level
## 
## Fit: aov(formula = mass ~ grazed * type, data = biomass.with.type)
## 
## $type
##                   diff        lwr       upr     p adj
## gram-forb     2.027898  -1.505444   5.56124 0.4523714
## litter-forb  73.721334  67.366864  80.07580 0.0000000
## wood-forb    13.587169   8.819623  18.35471 0.0000000
## litter-gram  71.693436  65.394331  77.99254 0.0000000
## wood-gram    11.559271   6.865773  16.25277 0.0000000
## wood-litter -60.134165 -67.199836 -53.06849 0.0000000

Plot 2:

        Boxplots show the mass of several categories of plants— forbs, grams, woody, and litter— in grazed and non-grazed plots. The median mass of all categories except litter is around 1-2 grams, with litter having a mean mass closer to 50 grams. In all three non-litter categories there are many outliers above the median. Results of an ANOVA test show a statistically significant difference between the masses of different plant categories, but not between grazed and non-grazed plots, or the interaction between grazing and plant type. Results of a Tukey test suggest there is a statistically significant difference between all plant categories except forb and gram.
#testing if we're all set on the burn plots
not_burned_litter.1 <- biomass.with.type %>%  
  filter(burn_unit %in% c("309")) %>%
  ggplot(aes(x=type, y=mass, color=grazed)) +
  geom_boxplot() +
  labs(x="Biomass Type", y="Mass (grams)", title="Unburned Plots") +
  theme_minimal() +
  scale_color_brewer(palette="Accent")
burned_litter.1 <- biomass.with.type %>% 
  filter(burn_unit %in% c("102", "103", "104", "105", "108", "309", "409")) %>%
  ggplot(aes(x=type, y=mass, color=grazed)) +
  geom_boxplot() +
  labs(x="Biomass Type", y="Mass (grams)", title="Burned Plots") +
  theme_minimal() +
    scale_color_brewer(palette="Accent")

Plots 3 and 4

grid.newpage()
grid.draw(rbind(ggplotGrob(not_burned_litter.1), ggplotGrob(burned_litter.1) , size = "last")) 

grazed_and_type_burned <- biomass.with.type%>%
  filter(burn_unit %in% c("102", "103", "104", "105", "108", "309", "409"))
grazed_burned_aov <- aov(mass ~ grazed * type, data = grazed_and_type_burned)
summary(grazed_burned_aov)
##               Df  Sum Sq Mean Sq F value Pr(>F)    
## grazed         1    1585    1585   2.424  0.120    
## type           3  629866  209955 321.019 <2e-16 ***
## grazed:type    3    1505     502   0.767  0.513    
## Residuals   1778 1162861     654                   
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1

Note that the echo = FALSE parameter was added to the code chunk to prevent printing of the R code that generated the plot.

grazed_and_type_unburned <- biomass.with.type%>%
  filter(burn_unit %in% c("309")) 
grazed_unburned_aov <- aov(mass ~ grazed * type, data = grazed_and_type_unburned)
summary(grazed_unburned_aov)
##              Df Sum Sq Mean Sq F value   Pr(>F)    
## grazed        1  23547   23547  13.099 0.000408 ***
## type          3 798719  266240 148.114  < 2e-16 ***
## grazed:type   3  11062    3687   2.051 0.109360    
## Residuals   144 258844    1798                     
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
TukeyHSD(x=grazed_unburned_aov,'type')
##   Tukey multiple comparisons of means
##     95% family-wise confidence level
## 
## Fit: aov(formula = mass ~ grazed * type, data = grazed_and_type_unburned)
## 
## $type
##                     diff         lwr        upr     p adj
## gram-forb     -0.8470974  -30.571453   28.87726 0.9998536
## litter-forb  231.3979643  198.104756  264.69117 0.0000000
## wood-forb     15.1682048   -9.069491   39.40590 0.3669580
## litter-gram  232.2450617  198.711466  265.77866 0.0000000
## wood-gram     16.0153022   -8.551550   40.58215 0.3303422
## wood-litter -216.2297595 -245.012230 -187.44729 0.0000000

Plots 3 and 4

        These two plots, like plot 2, show mass of different plant categories in grazed and non-grazed plots. In plot 3, we see specifically these variables in regularly burned plots, and in plot 4 we see these variables in unburned plots. Again, litter has by far the highest mass, so much so that other plant categories can not easily be distinguished by eye. In general, burned plots have a higher mass of every plant category except litter compared to unburned plots— there appears to be more litter in unburned plots. Any difference caused by grazing is not visible in these boxplots. In the unburned burn units, an ANOVA test suggested there were statistically significant differences between grazed and ungrazed plots.

Discussion

        The boxplots and analyses of the data are inconclusive in determining the effects of bison grazing on restoring a savanna ecosystem. We were interested in examining whether grazing would increase ability of oak saplings to survive. However, oak height, one measure of survivorship of oaks, was not significantly impacted by whether or not the plot had been grazed. Additionally, we were interested in seeing if bison would preferentially graze on gramminoid species over forbs and woody species. However, there were not statistically significant differences in biomass of these plant types between grazed and ungrazed plots. We were also interested in determining if burning would change outcomes of grazing plant biomass, since bison are known to graze preferentially on newly burned areas. Results were difficult to visualize.
          Although we were not able to determine much about the effect of grazing, the data was

not totally meaningless. It showed oak saplings being taller in fall than in spring, suggesting they had grown throughout the summer. In the plots comparing burned to unburned areas, every plant type appeared to have a lower biomass in the unburned plots except for plant litter. This suggests that burning clears litter and promotes growth of living plants, as has been shown in many previous studies. In the future, more years of data could be explored— perhaps over time it will become more conclusive. If possible, it would also be ideal to conduct a similar experiment in a larger area, and therefore be able to better mimic the true historical behaviors of bison, who grazed and roamed across long distances, allowing areas of land to rest and not be overgrazed. It would also be interesting to examine the use of other grazing animals, like cattle, in an oak savanna ecosystem, as their effectiveness is being studied in grasslands restorations.

Works Cited

Knapp, A. et al, The Keystone Role of Bison in North American Tallgrass Prairie: Bison increase habitat heterogeneity and alter a broad array of plant, community, and ecosystem processes, BioScience, Volume 49, Issue 1, January 1999, Pages 39–50, https://doi.org/10.1525/bisi.1999.49.1.39

Hess, A. American bison influences on lepidopteran and wild blue lupine distribution in an oak savanna landscape, J Insect Conserv (2014) 18:327–338 DOI 10.1007/s10841-014-9640-x

Petersen, D. Reich, P. Ecological Applications, Volume 11, Issue 3, June 2001 https://doi.org/10.1890/1051-0761(2001)0110914:PFIOSF2.0.CO;2

Moura, L. et al., The legacy of colonial fire management policies on traditional livelihoods and ecological sustainability in savannas: Impacts, consequences, new directions, Journal of Environmental Management Volume 232, 15 February 2019, Pages 600-606 https://doi.org/10.1016/j.jenvman.2018.11.057