Choropleth maps with tricolore

Here I demonstrate how to use the tricolore library to generate ternary choropleth maps using both ggplot2 and leaflet.

The data

library(tricolore)
#> Registered S3 methods overwritten by 'ggtern':
#>   method           from   
#>   grid.draw.ggplot ggplot2
#>   plot.ggplot      ggplot2
#>   print.ggplot     ggplot2
#> Please cite tricolore. See citation("tricolore").

The data set euro_example contains the administrative boundaries for the European NUTS-2 regions in the column geometry. This data can be used to plot a choropleth map of Europe using the sf package. Each region is represented by a single row. The name of a region is given by the variable name while the respective NUTS-2 geocode is given by the variable id. For each region some compositional statistics are available: Variables starting with ed refer to the relative share of population ages 25 to 64 by educational attainment in 2016 and variables starting with lf refer to the relative share of workers by labor-force sector in the European NUTS-2 regions 2016.

Take the first row of the data set as an example: in the Austrian region of “Burgenland” (id = AT11) 16.5% of the population aged 25–64 had attained an education of “Lower secondary or less” (ed_0to2), 55.7% attained “upper secondary” education (ed_3to4), and 27.9% attained “tertiary” education. In the very same region 4.4% of the labor-force works in the primary sector, 26.8% in the secondary and 68.2% in the tertiary sector.

The education and labor-force compositions are ternary, i.e. made up from three elements, and therefore can be color-coded as the weighted mixture of three primary colors, each primary mapped to one of the three elements. Such a color scale is called a ternary balance scheme1. This is what tricolore does.

ggplot2 for ternary choropleth maps

Here I show how to create a choropleth map of the regional distribution of education attainment in Europe 2016 using ggplot2.

1. Using the Tricolore() function, color-code each educational composition in the euro_example data set and add the resulting vector of hex-srgb colors as a new variable to the data frame. Store the color key separately.

# color-code the data set and generate a color-key
tric <- Tricolore(euro_example, p1 = 'ed_0to2', p2 = 'ed_3to4', p3 = 'ed_5to8')

tric contains both a vector of color-coded compositions (tric$rgb) and the corresponding color key (tric$key). We add the vector of colors to the map-data.

# add the vector of colors to the `euro_example` data
euro_example$rgb <- tric$rgb

2. Using ggplot2 and the joined color-coded education data and geodata, plot a ternary choropleth map of education attainment in the European regions. Add the color key to the map.

The secret ingredient is scale_fill_identity() to make sure that each region is colored according to the value in the rgb variable of euro_educ_map.

library(ggplot2)

plot_educ <-
  # using sf dataframe `euro_example`...
  ggplot(euro_example) +
  # ...draw a polygon for each region...
  geom_sf(aes(fill = rgb, geometry = geometry), size = 0.1) +
  # ...and color each region according to the color code in the variable `rgb`
  scale_fill_identity()

plot_educ

Using annotation_custom() and ggplotGrob we can add the color key produced by Tricolore() to the map. Internally, the color key is produced with the ggtern package. In order for it to render correctly we need to load ggtern after loading ggplot2. Don’t worry, the ggplot2 functions still work.

library(ggtern)
#> --
#> Remember to cite, run citation(package = 'ggtern') for further info.
#> --
#> 
#> Attaching package: 'ggtern'
#> The following objects are masked from 'package:ggplot2':
#> 
#>     aes, annotate, ggplot, ggplotGrob, ggplot_build, ggplot_gtable,
#>     ggsave, layer_data, theme_bw, theme_classic, theme_dark,
#>     theme_gray, theme_light, theme_linedraw, theme_minimal, theme_void
plot_educ +
  annotation_custom(
    ggplotGrob(tric$key),
    xmin = 55e5, xmax = 75e5, ymin = 8e5, ymax = 80e5
  )

Because the color key behaves just like a ggplot2 plot we can change it to our liking.

plot_educ <-
  plot_educ +
  annotation_custom(
    ggplotGrob(tric$key +
                 theme(plot.background = element_rect(fill = NA, color = NA)) +
                 labs(L = '0-2', T = '3-4', R = '5-8')),
    xmin = 55e5, xmax = 75e5, ymin = 8e5, ymax = 80e5
  )
plot_educ

Some final touches…

plot_educ +
  theme_void() +
  coord_sf(datum = NA) +
  labs(
   title = 'European inequalities in educational attainment',
      subtitle = 'Regional distribution of ISCED education levels for people aged 25-64 in 2016.'
  )

leaflet for ternary choropleth maps

The ggplot2 example above is easily adapted to leaflet. This time I use a continuous color scale.

# color-code the data set and generate a color-key
tric <- Tricolore(euro_example, p1 = 'ed_0to2', p2 = 'ed_3to4', p3 = 'ed_5to8',
                  breaks = Inf)

# add the vector of colors to the `euro_example` data
euro_example$rgb <- tric$rgb

leaflet requires geodata in spherical coordinates (longitude-latitude format). Therefore I reproject the data to a suitable crs using the sf package.

library(sf)
#> Linking to GEOS 3.12.1, GDAL 3.8.4, PROJ 9.4.0; sf_use_s2() is TRUE
library(leaflet)

euro_example %>%
  st_transform(crs = 4326) %>%
  leaflet() %>%
  addPolygons(smoothFactor = 0.1, weight = 0,
              fillColor = euro_example$rgb,
              fillOpacity = 1)

Adding a background map gives geographical context to the map. I also add a mouse pop-up of the actual data.

euro_example %>%
  st_transform(crs = 4326) %>%
  leaflet() %>%
  addProviderTiles(providers$Esri.WorldTerrain) %>%
  addPolygons(smoothFactor = 0.1, weight = 0,
              fillColor = euro_example$rgb,
              fillOpacity = 1,
              popup =
                paste0(
                  '<b>', euro_example$name, '</b></br>',
                  'Primary: ',
                  formatC(euro_example$ed_0to2*100,
                          digits = 1, format = 'f'), '%</br>',
                  'Secondary: ',
                  formatC(euro_example$ed_3to4*100,
                          digits = 1, format = 'f'), '%</br>',
                  'Tertiary: ',
                  formatC(euro_example$ed_5to8*100,
                          digits = 1, format = 'f'), '%</br>'
                )
  )

Adding the legend to the leaflet map requires a bit of a hack.

makePlotURI <- function(expr, width, height, ...) {
  pngFile <- shiny::plotPNG(function() { expr }, width = width, height = height, ...)
  on.exit(unlink(pngFile))

  base64 <- httpuv::rawToBase64(readBin(pngFile, raw(1), file.size(pngFile)))
  paste0("data:image/png;base64,", base64)
}

legend_symbol <- makePlotURI({
  print(tric$key +
          theme(plot.background = element_rect(fill = NA, color = NA)) +
          labs(L = '0-2', T = '3-4', R = '5-8'))
}, 200, 200, bg = "transparent")

df <- data.frame(
  lng = 30,
  lat = 70,
  plot = legend_symbol,
  stringsAsFactors = FALSE
)

euro_example %>%
  st_transform(crs = 4326) %>%
  leaflet() %>%
  addProviderTiles(providers$Esri.WorldGrayCanvas) %>%
  addPolygons(smoothFactor = 0.1, weight = 0,
              fillColor = euro_example$rgb,
              fillOpacity = 1,
              popup =
                paste0(
                  '<b>', euro_example$name, '</b></br>',
                  'Primary: ',
                  formatC(euro_example$ed_0to2*100,
                          digits = 1, format = 'f'), '%</br>',
                  'Secondary: ',
                  formatC(euro_example$ed_3to4*100,
                          digits = 1, format = 'f'), '%</br>',
                  'Tertiary: ',
                  formatC(euro_example$ed_5to8*100,
                          digits = 1, format = 'f'), '%</br>'
                )
  ) %>%
  addMarkers(data = df, icon = ~icons(plot))
#> Assuming "lng" and "lat" are longitude and latitude, respectively

Literature

Brewer, C. A. (1994). Color Use Guidelines for Mapping and Visualization. In A. M. MacEachren & D. R. F. Taylor (Eds.), Visualization in Modern Cartography (pp. 123–147). Oxford, UK: Pergamon.

Dorling, D. (2012). The Visualization of Spatial Social Structure. Chichester, UK: Wiley.

Schöley, J. (2021). The centered ternary balance scheme: A technique to visualize surfaces of unbalanced three-part compositions. Demographic Research (44).

  1. See for example Dorling (2012) and Brewer (1994).↩︎