Could convert a square raster to a hexagon when plotting with geom_spatraster

[zzzqiii]

Hi, I am using geom_spatraster to reproduce a figure from a paper and basically finished it.
This is my result.

And this is my code:

ggplot() + 
  geom_spatvector(data = global_land, fill = "grey88", size = 0.1, color = NA) +
  geom_spatraster(data=ra, aes(fill=Intertidal_marine_forests),maxcell = 3 * 10^6) + scale_fill_hypso_c() +
  theme_void()+
  theme(legend.position = c(0.1, 0.07))+
  theme(axis.title.x = element_blank(), 
        axis.title.y = element_blank(), 
        panel.background = element_blank())

However, I found the raster is hexagon in the original figure and it looks much clearer.

I wonder if I could convert a square raster to a hexagon when plotting with geom_spatraster.
Any hint would be appreciated. And please let me know if more information is needed.

[dieghernan]

Hi,

This is more advanced but it can be done with tidyterra. The trick is, instead of using geom_spatraster(), use ggplot() + stat_summary_hex(). In fact you would be fortifying the SpatRaster and applying a new stat, see some examples here:

https://dieghernan.github.io/tidyterra/reference/fortify.Spat.html

A working example

A. Mock the data

library(geodata)
library(tidyterra)
library(terra)
library(ggplot2)
library(hexbin)

# Get global land
global_land <- geodata::gadm(country = "ITA", path = tempdir(), level = 0) %>%
  aggregate() %>%
  project("+proj=robin")


# Mock ra
ra <- geodata::worldclim_country(
  country = "ITA", res = 10, var = "prec",
  path = tempdir()
) %>%
  select(Intertidal_marine_forests = ITA_wc2.1_30s_prec_6) %>%
  project("+proj=robin") %>%
  # Downsample for demo
  spatSample(size = 1000, method = "regular", as.raster = TRUE) %>%
  # Get ra touching boundaries only
  terra::mask(as.lines(global_land))

First, ensure that global_land and ra are in the same projection, i.e global_land <- terra::project(global_land, terra::crs(ra))

global_land <- terra::project(global_land, terra::crs(ra))

B. Create the plot

# Original plot
ggplot() +
  geom_spatvector(data = global_land, fill = "grey88", size = 0.1, color = NA) +
  geom_spatraster(
    data = ra, aes(fill = Intertidal_marine_forests), maxcell = 3 * 10^6,
    show.legend = FALSE
  ) +
  scale_fill_hypso_c() +
  theme_void() +
  theme(legend.position = c(0.1, 0.07)) +
  theme(
    axis.title.x = element_blank(),
    axis.title.y = element_blank(),
    panel.background = element_blank()
  ) +
  labs(title = "Original")

The alternative. Please note that you may need to fiddle with bins argument on stat_summary_hex():

# Alt: fortifying and using geom_hex
ggplot(ra, maxcell = 3 * 10^6) + # See here
  geom_spatvector(data = global_land, fill = "grey88", size = 0.1, color = NA) +
  # And here
  stat_summary_hex(aes(x = x, y = y, z = Intertidal_marine_forests),
    show.legend = FALSE,
    # IMPORTANT: Play with this parameter
    bins = 40
  ) +
  scale_fill_hypso_c(na.value = "transparent") +
  theme_void() +
  theme(legend.position = c(0.1, 0.07)) +
  theme(
    axis.title.x = element_blank(),
    axis.title.y = element_blank(),
    panel.background = element_blank()
  ) +
  labs(title = "Using fortify and stat_summary_hex")
#> Warning: Removed 910 rows containing non-finite outside the scale range
#> (`stat_summary_hex()`).

^{Created on 2024-04-02 with reprex v2.1.0}

[dieghernan]

Another approach is to create an hexagonal grid of SpatVector via sf and rasterize to those polygons (see more in https://dieghernan.github.io/202312_bertin_dots/#alternative-hexagonal-grid):

# Hex grid with sf ----

library(sf)
library(units)

# Avg size of the cells on the initial grid
target_area <- cellSize(ra, unit = "km") %>%
  pull() %>%
  mean() %>%
  as_units("km^2")

diam_hex <- sqrt(2 * target_area / sqrt(3))

# Create hexagonal grid
ra_grid <- sf::st_make_grid(global_land, cellsize = diam_hex, square = FALSE) %>%
  terra::vect()

ra_grid$Intertidal_marine_forests <- terra::extract(ra, ra_grid, fun = "mean") %>%
  pull(Intertidal_marine_forests)

# Clean hex grid
ra_grid <- drop_na(ra_grid)

# Alt2
ggplot() +
  geom_spatvector(data = global_land, fill = "grey88", size = 0.1, color = NA) +
  geom_spatvector(
    data = ra_grid, aes(fill = Intertidal_marine_forests), color = NA,
    show.legend = FALSE
  ) +
  scale_fill_hypso_c(na.value = "transparent") +
  theme_void() +
  theme(legend.position = c(0.1, 0.07)) +
  theme(
    axis.title.x = element_blank(),
    axis.title.y = element_blank(),
    panel.background = element_blank()
  ) +
  labs(title = "With sf hex conversion")

^{Created on 2024-04-02 with reprex v2.1.0}