Peeling Back the Globe: Cartographic Innovation Meets Data Science with the Interrupted Goode Homolosine Projection in R

Main Facts

In the realm of cartography, representing a three-dimensional, spherical Earth on a flat two-dimensional surface has always been an exercise in compromise. Every map projection forces a choice between distorting area, direction, distance, or shape. Among the most creative solutions to this age-old geographic dilemma is the interrupted Goode homolosine projection—a pseudocylindrical, equal-area map that minimizes distortion by splitting the world into distinct, lobed sections, much like peeling an orange and flattening the rind.

Recently, data visualization and spatial analysis enthusiasts have gained a powerful, programmatic toolkit to bring this unique projection to life using modern programming tools. By leveraging the R programming language—specifically through spatial and data visualization libraries such as sf, giscoR, and ggplot2—developers and cartographers can generate stunning, customized representations of the globe. This approach not only preserves the true relative sizes of Earth’s landmasses but also dresses them in striking, thematic color palettes, transforming complex spatial mathematics into an aesthetically engaging data art piece.

The core technical achievement lies in R’s ability to manipulate spatial geometries. By defining specific coordinate boundaries ("lobes"), managing complex polygonal masks, and applying precise coordinate reference systems (CRS)—such as +proj=igh—analysts can render the interrupted Goode homolosine projection cleanly. This prevents overlapping meridians and parallels across the map’s characteristic "cuts," resulting in a polished visualization that bridges the gap between rigorous geographic science and compelling graphic design.

Chronology of Cartographic Innovation and Spatial Programming

To understand the significance of rendering the Goode homolosine projection in R, it is helpful to trace the evolution from classical cartography to modern computational spatial analysis.

• 1923: American geographer John Paul Goode introduces the Goode homolosine projection. Designed as an equal-area (homolosine) projection that combines the interrupted versions of the Molleweide and Sinusoidal projections, it is meant to accurately compare continental sizes without the massive distortions seen near the poles in standard cylindrical maps like Mercator.
• Late 20th Century: Cartographers rely heavily on manual drafting, specialized GIS workstations, and early desktop software (like early versions of ArcInfo or GIS extensions) to generate interrupted projections, a process requiring intensive computational and manual masking.
• 2010s: The open-source R ecosystem experiences a massive revolution in spatial data handling. The release and subsequent maturation of the sf (Simple Features) package by Edzer Pebesma standardizes vector data handling in R, aligning it with modern standards (ISO/OGC). Meanwhile, Hadley Wickham’s ggplot2 becomes the gold standard for grammar-of-graphics plotting.
• Recent Years: Advanced spatial visualization techniques emerge, pioneered by data scientists and cartography educators like Claus O. Wilke, who demonstrate how to handle complex masking and outlining for interrupted projections in R. Concurrently, packages like giscoR simplify the retrieval of high-resolution global boundaries and administrative geometries directly from Eurostat and other global repositories.
• Present Day: Developers combine sf, giscoR, and ggplot2 to automate the creation of publication-quality, themed interrupted Goode homolosine maps in mere seconds, merging automated scripting workflows with fine-tuned graphic design.

Supporting Data and Technical Architecture

Creating an interrupted Goode homolosine map in R requires a disciplined, multi-step pipeline that handles vector data, geometric transformations, and graphic layer masking. Below is an examination of the technical architecture required to execute this visualization.

1. Data Acquisition and Coordinate Transformation

The foundation of the map begins by fetching high-resolution coastal lines and establishing the target Coordinate Reference System (CRS).

library(sf)
library(ggplot2)
library(giscoR)

# Retrieve global coastal lines at a defined resolution
world <- gisco_get_coastal_lines(resolution = 10)

# Define the interrupted Goode homolosine projection string
goode_crs <- "+proj=igh"

The projection string +proj=igh tells the PROJ cartographic projections library to apply the Goode formula. However, because the map is interrupted, standard bounding boxes will capture unwanted spatial artifacts. To solve this, developers construct a custom spatial polygon (outline) that traces the exact borders of the globe’s lobes based on specific latitude and longitude vectors:

lat <- c(90:-90, rep(c(-90:0, 0:-90), 3), -90:90, 90:0, 0:90, 90)
lon <- c(
  rep(180, 181),
  rep(c(80.01, 79.99, -19.99, -20.01, -99.99, -100.01), each = 91),
  rep(-180, 181),
  rep(c(-40.01, -39.99), each = 91),
  180
)

outline <- st_sfc(st_polygon(list(cbind(lon, lat))), crs = 4326) |&gt;
  st_transform(goode_crs)

rectangle <- st_as_sfc(st_bbox(outline) * 1.1)
mask <- st_difference(rectangle, outline)

2. Graticule and Lobe Partitioning

In a standard un-interrupted map, meridians and parallels wrap continuously around the globe. In an interrupted projection, these lines must be segmented so they do not incorrectly bridge across the empty spaces (the "cuts" between the orange peel lobes).

The script accomplishes this by iterating through a predefined list of six geographic lobes, generating localized sequences of meridians (vertical grid lines every 30 degrees) and parallels (horizontal grid lines):

lobes <- list(
  c(-180, -40, 0, 90), c(-40, 180, 0, 90),
  c(-180, -100, -90, 0), c(-100, -20, -90, 0),
  c(-20, 80, -90, 0), c(80, 180, -90, 0)
)

lines <- lapply(lobes, function(b) 
  west <- b[1] + 0.01
  east <- b[2] - 0.01
  south <- max(b[3], -89.99)
  north <- min(b[4], 89.99)
  meridians <- seq(-180, 180, 30)
  meridians <- meridians[meridians > west & meridians < east]
  parallels <- seq(-60, 60, 30)
  parallels <- parallels[parallels >= south & parallels <= north]
  c(
    lapply(meridians, function(x) 
      st_linestring(cbind(x, seq(south, north, length.out = 181)))
    ),
    lapply(parallels, function(y) 
      st_linestring(cbind(seq(west, east, length.out = 361), y))
    )
  )
)

grat <- st_sfc(do.call(c, lines), crs = 4326) |&gt;
  st_transform(goode_crs)

3. Theming and Aesthetic Rendering via ggplot2

The final step assembles the layers inside ggplot2. By utilizing an orange-themed color palette (#FFF0D9 for backgrounds, #F9B566 for landmasses, and #E87516 for borders), the map achieves a playful yet professional visual metaphor. The mask polygon is applied in solid white to hide any stray geometric elements falling outside the true lobe boundaries, ensuring clean, sharp edges.

ggplot() +
  geom_sf(data = outline, fill = "#FFF0D9", colour = NA) +
  geom_sf(data = world, fill = "#F9B566", colour = "#B34D08", linewidth = 0.25) +
  geom_sf(data = grat, fill = NA, colour = "#E8A45B", linewidth = 0.25) +
  geom_sf(data = mask, fill = "white", colour = NA) +
  geom_sf(data = outline, fill = NA, colour = "#E87516", linewidth = 0.4) +
  coord_sf(
    crs = goode_crs,
    xlim = st_bbox(outline)[c("xmin", "xmax")],
    ylim = st_bbox(outline)[c("ymin", "ymax")],
    expand = FALSE,
    datum = NA
  ) +
  theme_void() +
  theme(
    plot.background = element_rect(fill = "white", colour = NA),
    plot.text = element_text(color = "#E87516", face = "bold")
  ) +
  labs(
    title = "Let's peel the world",
    caption = "Interrupted Goode homolosine projection rendered via R"
  )

Official Responses and Expert Perspectives

Geospatial data scientists, cartographers, and open-source software developers have widely praised programmatic approaches to complex map projections.

According to spatial data educators and contributors within the R community, moving away from closed, proprietary desktop GIS software toward script-based cartography offers unprecedented reproducibility and creative freedom. When discussing advanced projection workflows, cartographic experts emphasize that techniques like masking and custom graticule clipping—once considered tedious manual labor in legacy vector graphics editors—can now be executed dynamically.

Furthermore, community feedback highlights the educational value of projects inspired by cartographers like Claus O. Wilke. By breaking down how projections handle spatial boundaries, developers gain a deeper appreciation for the mathematical compromises inherent in cartography. The ability to seamlessly integrate packages like giscoR for administrative data with the rendering power of ggplot2 empowers researchers to communicate spatial phenomena accurately without falling into the distortions typical of default web-mapping projections (such as the Web Mercator distortion that exaggerates the size of high-latitude landmasses like Greenland and North America relative to equatorial regions).

Implications for Data Science, Education, and Cartography

The convergence of statistical programming languages and advanced cartographic projections carries profound implications across multiple disciplines:

1. De-biasing Geographic Perception

For decades, popular web mapping applications have relied on the Mercator projection, which severely distorts land area as latitude increases. By making equal-area projections like the interrupted Goode homolosine easily accessible via open-source scripts, data scientists can educate the public and decision-makers on the true scale of global regions—particularly highlighting the true landmass of continents like Africa and South America.

2. The Rise of "Data Art" and Computational Design

The intersection of programming and graphic design—often termed generative cartography or data art—demonstrates that technical reports and academic papers do not need to rely on sterile, default visualizations. By utilizing custom color scales, thematic metaphors (such as the orange peel analogy), and clean geometric masking, analysts can capture audience attention while maintaining rigorous geographic accuracy.

3. Open-Source Ecosystem Maturity

The seamless execution of this workflow highlights the maturity of the R spatial ecosystem. The interoperability between sf (handling simple features geometry), giscoR (handling web-based geographic data retrieval), and ggplot2 (handling grammar-of-graphics rendering) proves that open-source tooling can rival or exceed the capabilities of expensive commercial geographic information systems for specialized cartographic tasks.

As data visualization continues to evolve, the ability to programmatically manipulate how we view our planet ensures that cartography remains both a precise science and a vibrant art form.

References

• Wilke, Claus O. Fundamentals of Data Visualization / Interrupted Goode homolosine. Available online via Wilke Lab Practical GG.
• PROJ Cartographic Projections Library. Interrupted Goode Homolosine (igh). Documentation accessible at PROJ.org.
• Pyranhia. Projections cartographiques : le guide pour ne pas y laisser sa santé mentale. Technical cartography guide (in French). Available at Pyranhia.eu.
• Pebesma, Edzer. Simple Features for R: Standardized Support for Spatial Vector Data. The R Journal.