In valentinitnelav/plotbiomes: Plot Whittaker biomes with ggplot2

This vignette presents some examples of Whittaker biomes plots using the plotbiomes package.

Minimal example

We will build on top of this minimal plot.

library(plotbiomes)
library(ggplot2)

plot_1 <- ggplot() +
  # add biome polygons
  geom_polygon(data = Whittaker_biomes,
               aes(x    = temp_c,
                   y    = precp_cm,
                   fill = biome),
               # adjust polygon borders
               colour = "gray98",
               size   = 1) +
  theme_bw()
plot_1

Add color

Suggested colors

Ricklefs_colors is a predefined vector of colors that comes with plotbiomes package. These are the exact colors from Figure 5.5 in Ricklefs, R. E. (2008), The economy of nature. W. H. Freeman and Company. (Chapter 5, Biological Communities, The biome concept)

Ricklefs_colors

plot_2 <- plot_1 +
  # fill the polygons with predefined colors
  scale_fill_manual(name   = "Whittaker biomes",
                    breaks = names(Ricklefs_colors),
                    labels = names(Ricklefs_colors),
                    values = Ricklefs_colors)
plot_2

The same can be achieved with a simple function call (author: Sam Levin):

whittaker_base_plot() + theme_bw()

Define colors

Of course, if you feel creative, a color palette can be defined with RColorBrewer or grDevices packages. Try also the colorspace package. Or simply pick your favorite 9 colors.

library(RColorBrewer)

# the main rule - create 9 colors for the 9 biomes

# failed trial with RColorBrewer :)
my_palette_1 <- rev(brewer.pal(n = 9, name = "BrBG"))
whittaker_base_plot(color_palette = my_palette_1)

# this seems a better approach - interpolate 9 colors from given main 3  
my_palette_2 <- colorRampPalette(colors = c("#F5F5F5",  "#01665E", "#8C510A"))(9)
whittaker_base_plot(color_palette = my_palette_2)

The above is just a shorter way of this:

names(my_palette_2) <- names(Ricklefs_colors)

ggplot() +
  # add biome polygons
  geom_polygon(data = Whittaker_biomes,
               aes(x    = temp_c,
                   y    = precp_cm,
                   fill = biome),
               # adjust polygon border
               colour = "gray98",
               size   = 1) +
  # fill the polygons with desired colors
  scale_fill_manual(name   = "Whittaker biomes",
                    breaks = names(Ricklefs_colors),
                    labels = names(Ricklefs_colors),
                    values = my_palette_2)

Add data points

Example with random locations

- Generate data

Generate random locations and extract temperature and precipitation from WorldClim data. You will have already a table of coordinates, so better check directly ?raster::extract().

library(raster)
library(maptools)

# ===== Prepare raster stack
# Read temperature and precipitation as raster stack.
# Low resolution raster datasets come with 'plotbiomes' package.
path <- system.file("extdata", "temp_pp.tif", package = "plotbiomes")
temp_pp <- raster::stack(path)
names(temp_pp) <- c("temperature", "precipitation")

# ===== Generate random locations
data(wrld_simpl) # load world polygons from maptools
# Eliminate Antarctica - it is not represented in the raster datasets.
wrld_simpl <- wrld_simpl[wrld_simpl$NAME != "Antarctica", ]
# Create random locations within world's polygons.
set.seed(66) # random number generator
points <- sp::spsample(x = wrld_simpl, n = 50, type = "random")

# ===== Extract from raster stack
# Extract temperature and precipitation values from the raster datasets
extractions <- raster::extract(temp_pp, points, df = TRUE)
# Adjust temperature values to normal scale because WorldClim temperature data
# has a scale factor of 10 (integer storage for saving space).
extractions$temperature <- extractions$temperature/10
# Convert precipitation from mm to cm
extractions$precipitation <- extractions$precipitation/10

- Plot data

Plot the random locations. Note that any points outside of rasters coverage will receive NA-s. Also they will be dropped by ggplot.

plot(temp_pp[[1]]/10); points(points)
plot(temp_pp[[2]]); points(points)

Example of plot incorporating extraction data corresponding to the locations.

whittaker_base_plot() +
  # add the temperature - precipitation data points
  geom_point(data = extractions, 
             aes(x = temperature, 
                 y = precipitation), 
             size   = 3,
             shape  = 21,
             colour = "gray95", 
             fill   = "black",
             stroke = 1,
             alpha  = 0.5) +
  theme_bw()

Color points

Minimal example

When coloring the points by a certain factor (categorical variable), then a simple thing to do is to map (assign) the variable to the color aesthetic in geom_point:

# Color the outliers - use category names like "out" & "in"
my_outliers <- get_outliers(tp = extractions[, 2:3])
extractions$status <- ifelse(extractions$ID %in% my_outliers$row_idx, "out", "in")

whittaker_base_plot() +
  geom_point(data = extractions,
             aes(x = temperature,
                 y = precipitation,
                 colour = status), # map `status` variable to colour
             shape = 16,
             size  = 3) +
  theme_bw()

Color & fill

To have both border color and background fill aesthetics for the points is a bit tricky/creative since fill is already in use for the biome polygones and ggplot2 currently allows only one scale per aesthetic. So, we cannot use two different fill mappings. Nevertheless, here are some approaches:

- two geom_point calls

A solution is to have two calls to geom_point (see 6.4.1 Layers and Legends in ggplot2: Elegant Graphics for Data Analysis). With one call setting the points' boundary line and with the second coloring the points.

plot_3 <- whittaker_base_plot() +
  geom_point(data = extractions,
             aes(x = temperature,
                 y = precipitation),
             shape  = 21,
             stroke = 1, # acts as the thickness of the boundary line
             colour = "gray95", # acts as the color of the boundary line
             size   = 3.5) +
  geom_point(data = extractions,
             aes(x = temperature,
                 y = precipitation,
                 color = status),
             shape = 16,
             size  = 3,
             alpha = 0.5) +
  scale_color_manual(name   = "Status",
                     breaks = c("in", "out"),
                     values = c("in"  = "black",
                                "out" = "red"),
                     labels = c("in"  = "inside",
                                "out" = "outlier"))
plot_3 + theme_bw()

- fill vector outside aes()

Another approach is to assign a vector of colors to fill (or bg) argument for points of shape 21-25 (only these shapes are affected by the fill color). The fill argument needs to be outside the aes declaration. This is done in single geom_point call, but loses the legend for point fill aesthetic. Anyhow, not always a legend is needed and we can always give details in the figure caption.

# Color the outliers - use color names
extractions$status <- ifelse(extractions$ID %in% my_outliers$row_idx, "red", "black")

whittaker_base_plot() +
  geom_point(data = extractions,
             aes(x = temperature,
                 y = precipitation),
             size   = 3,
             shape  = 21,
             colour = "gray95", # the color of the boundary line
             # the fill color, use a vector of colors (names or hexadecimal codes):
             fill   = extractions$status, 
             stroke = 1,
             alpha  = 0.5) +
  theme_bw()

A more complex example

One can play with the ggplot settings like scale_ and theme to enhance the graph. Here is one trial (builds on top of plot_3 - see above):

my_plot <- plot_3 +
  # Optional - Overwrite axis ranges (the scale warning is expected):
  # - set range on OY axes and adjust the distance (gap) from OX axes
  scale_y_continuous(name = 'Precipitation (cm)',
                     limits = c(min = -5, max = ceiling(max(460, extractions$precipitation)/10)*10) ,
                     expand = c(0, 0)) +
  # - set range on OX axes and adjust the distance (gap) from OY axes
  scale_x_continuous(name = expression("Temperature " ( degree*C)),
                     limits = c(min = floor(min(-20, extractions$temperature)/5)*5, max = 30.5),
                     expand = c(0, 0)) +
  coord_fixed(ratio = 1/10) + # aspect ratio, expressed as y / x
  theme_bw() +
  theme(
    legend.justification = c(0, 1), # pick the upper left corner of the legend box and
    legend.position = c(0, 1), # adjust the position of the corner as relative to axis
    legend.background = element_rect(fill = NA), # transparent legend background
    legend.box = "horizontal", # horizontal arrangement of multiple legends
    legend.spacing.x = unit(0.5, units = "cm"), # horizontal spacing between legends
    panel.grid = element_blank() # eliminate grids
  )

my_plot

Interactive plot

The plotly package can take a ggplot graph and make it interactive. Note that one cannot use expression in labels or titles with plotly, so our whittaker_base_plot() function will not work smoothly here. Check this SO link for further details. Perhaps in the future this issue will be fixed, or check alternatives with MathJaX.

library(plotly)

plot_3 <- plot_2 + # check plot_2 above
  # add the temperature - precipitation points
  geom_point(data = extractions,
             aes(
               x = temperature,
               y = precipitation,
               # Text aesthetic will be ignored in `ggplot`,
               # but will be used for mouse hovering in `ggplotly`.
               text = paste0('row index: ', 1:nrow(extractions))
             ),
             size   = 3,
             shape  = 21,
             colour = "gray95",
             fill   = "black",
             stroke = 1,
             alpha  = 0.5)

ggplotly(plot_3)

Use the hover property to identify points.

Warning: Hovering will not catch overlapping points if they are not jittered and when they are jittered, it displays the jittered coordinates. This is misleading and is better to check the displayed point IDs/indices.

valentinitnelav/plotbiomes documentation built on Oct. 22, 2022, 6:22 a.m.