Modern maps with sf & projections

An equal-area, projected choropleth

world_data(2020, c(gdp = "NY.GDP.PCAP.KD"), geometry = "sf") |>
  world_map(gdp, style = "quantile", projection = "equal_earth",
            title = "GDP per capita (Equal Earth projection)")

world_map() auto-detects the sf backend and applies the projection through ggplot2::coord_sf(). Available projections are "equal_earth" (the default: equal-area and good-looking), "robinson", "mollweide", "natural_earth", "plate_carree", "mercator", "winkel_tripel", "eckert4", "gall_peters", "orthographic", "azimuthal_equal_area", "north_polar" and "south_polar".

projection_info() says what each one preserves, so you do not have to remember:

projection_info()[, c("projection", "property", "equal_area")]
#> # A tibble: 13 × 3
#>    projection           property    equal_area
#>    <chr>                <chr>       <lgl>     
#>  1 equal_earth          equal-area  TRUE      
#>  2 robinson             compromise  FALSE     
#>  3 mollweide            equal-area  TRUE      
#>  4 natural_earth        compromise  FALSE     
#>  5 plate_carree         equidistant FALSE     
#>  6 mercator             conformal   FALSE     
#>  7 winkel_tripel        compromise  FALSE     
#>  8 eckert4              equal-area  TRUE      
#>  9 gall_peters          equal-area  TRUE      
#> 10 orthographic         perspective FALSE     
#> 11 azimuthal_equal_area equal-area  TRUE      
#> 12 north_polar          equal-area  TRUE      
#> 13 south_polar          equal-area  TRUE

For a choropleth the equal-area ones are the honest choice, because the reader reads coloured area as quantity. projection_compare() draws your own data under several at once and tissot_map() shows the distortion directly (see the Honest maps vignette).

The world as a globe

globe_map() draws an orthographic globe centred on lon/lat. The default "sf" backend gives the cleanest limb; the "polygon" backend below needs only maps + mapproj (no sf):

globe_map(world_snapshot$countries, continent, backend = "polygon",
          style = "categorical", lon = 10, lat = 20)

# With the sf backend (smoother limb, real great circles):
world_data(2020, geometry = "sf") |>
  globe_map(gdp_per_capita, lon = 10, lat = 30)

spin_globe() turns that into a rotating animation: one globe_map() frame per central longitude, assembled into a looping GIF with gifski (or magick):

spin_globe(world_snapshot$countries, continent, backend = "polygon",
           style = "categorical", n_frames = 60)

Just the canvas

world_geometry() returns projected, region-subset, antimeridian-safe geometry without any data: country polygons, label-ready centroids, coastlines, a graticule or an ocean rectangle:

africa <- world_geometry("countries", geometry = "sf", region = "Africa",
                         projection = "equal_earth")
ggplot(africa) +
  geom_sf(fill = "grey85", colour = "grey40", linewidth = 0.1) +
  theme_world_map()

Recentring and the antimeridian

A Pacific-centred world is one argument away; the sf pipeline runs sf::st_break_antimeridian() before projecting, so nothing streaks across the frame:

world_geometry("countries", geometry = "sf", recenter = 150)

Region subsetting

region accepts a continent, a group name ("EU", "OECD", …), a vector of iso3c codes, or a bounding box c(xmin, ymin, xmax, ymax). Pair it with a projection suited to the subset (e.g. "azimuthal_equal_area" for a single continent, the polar projections for the Arctic) so the crop stays area-honest.

A box is the one form that clips the shapes themselves, which is what you want for a region that is neither a continent nor a group, the Mediterranean basin, say. Use the sf backend for it: that clip is a real sf::st_crop(), whereas the polygon backend can only drop the vertices outside the box and warns that a country crossing the edge comes back with an approximate outline.

med <- world_geometry("countries", geometry = "sf",
                      region = c(-10, 30, 40, 48), projection = "equal_earth")
ggplot(med) +
  geom_sf(fill = "grey85", colour = "grey40", linewidth = 0.1) +
  theme_world_map()

Simplifying for the web

High-resolution geometry can be thinned for fast plotting with simplify_geometry() (which uses rmapshaper when available). scale = "medium" is the 50m Natural Earth data; "large" is 10m and additionally needs the rnaturalearthhires package, which is not on CRAN.

world_geometry(geometry = "sf", scale = "medium") |>
  simplify_geometry(keep = 0.1)