Elephants affect the spatial distribution of savannah trees

2017-12-01

I have been attempting to write some scripts that will hopefully serve me well as I progress through writing my PhD. As I’m looking at the spatial distribution of trees and their canopy properties in dry tropical woodlands/savannahs, I thought it would be good to learn how to effectively plot tree distributions in plots. So I got hold of some data which contains the lat-long coordinates of trees in two 25 ha permanent savannah plots, one of which is known to be home to elephants, and one which isn’t. As always, the code I used is in R.

The data and an example script which I used can be found here .

First load the necessary packages:

# Packages ----
library(ggplot2)  # Plotting
library(dplyr)  # Manipulating data
library(readr)  # Importing data
library(viridis)  # Colour schemes

Then load in the data:

# Set working directory ----
setwd(dirname(rstudioapi::getActiveDocumentContext()$path))

# Import data ----
# Tree locations and diameters
tree_loc_diam_1_2_summ <- read.csv("tree_loc_diam_1_2.csv")

# Plot bounding boxes
plot_bbox_1 <- read_csv("plot_bb_1.csv")
plot_bbox_2 <- read_csv("plot_bb_2.csv")

First I want to plot the trees according to their DBH (Diameter at Breast Height, ~1.3 m), to see if there is a difference between plots. I’ll make a map, and a box plot:

# Plot tree DBH as point size ----
ggplot(tree_loc_diam_1_2_summ, aes(x = dec_lon,
	y = dec_lat,
	size = dbh_cm,
	colour = species)) +
geom_point(alpha = 0.5) +
theme(aspect.ratio = 1) +  # coord_map() doesn't work with facet_wrap()
facet_wrap(~ plot, scales = "free") +
theme(legend.position = "bottom")


# DBH boxplots ----
ggplot(tree_loc_diam_1_2_summ, aes(x = plot, y = dbh_cm)) +
	geom_boxplot()
Bubble plot of trees
Boxplot of tree diameters

Now I want to create a map that shows how clustered together the trees are in any given plot. For this I use the handy stat_density2d geom from ggplot2, and fill according to ..level... Also I found I had to manually increase the size of the plot, otherwise the density shading polygons get clipped and look really messy.

# Is there an effect of elephants on spatial clustering of trees? ----
# Split data frames
tree_loc_diam_1_summ <- tree_loc_diam_1_2_summ %>%
	filter(plot == "elephants")

tree_loc_diam_2_summ <- tree_loc_diam_1_2_summ %>%
	filter(plot == "no_elephants")

# Create heatmaps based on density of trees
elephant_plot <- ggplot(tree_loc_diam_1_summ, aes(x = dec_lon, y = dec_lat)) +
	stat_density2d(aes(fill = ..level..), geom = "polygon") +
	scale_fill_viridis() +
	geom_polygon(data = plot_bbox_1, aes(x = dec_lon, y = dec_lat), fill = NA, colour = "black") +
	geom_point(size = 0.8) +
	xlim(min(plot_bbox_1$dec_lon) - 0.001, max(plot_bbox_1$dec_lon) + 0.001) +
	ylim(min(plot_bbox_1$dec_lat) - 0.001, max(plot_bbox_1$dec_lat) + 0.001) +
	xlab("Decimal Longitude") +
	ylab("Decimal Latitude") +
	theme_classic() +
	theme(legend.text = element_text(size = 10),
				axis.text = element_text(size = 10)) +
	labs(fill = "Tree Density")

no_elephant_plot <- ggplot(tree_loc_diam_2_summ, aes(x = dec_lon, y = dec_lat)) +
	stat_density2d(aes(fill = ..level..), geom = "polygon") +
	scale_fill_viridis() +
	geom_polygon(data = plot_bbox_2, aes(x = dec_lon, y = dec_lat), fill = NA, colour = "black") +
	geom_point(size = 0.8) +
	xlim(min(plot_bbox_2$dec_lon) - 0.001, max(plot_bbox_2$dec_lon) + 0.001) +
	ylim(min(plot_bbox_2$dec_lat) - 0.001, max(plot_bbox_2$dec_lat) + 0.001) +
	xlab("Decimal Longitude") +
	ylab("Decimal Latitude") +
	theme_classic() +
	theme(legend.text = element_text(size = 10),
				axis.text = element_text(size = 10)) +
	labs(fill = "Tree Density")
Heatmap of trees with elephants
Heatmap of trees without elephants

The spatial clustering of trees in a plot with elephant activity (left) and without elephant activity (right). Elephants clearly have caused spatial clustering of trees.