H3 Grid Support

hexify v0.5.0 adds H3 as a first-class grid type alongside ISEA. Every core function (hexify(), grid_rect(), grid_clip(), get_parent(), get_children()) works with both grid systems through the same interface.

Installation

# Install from CRAN
install.packages("hexify")

# Or install development version from GitHub
# install.packages("pak")
pak::pak("gcol33/hexify")

What is H3?

H3 is a hierarchical hexagonal grid system developed by Uber. It partitions Earth’s surface into hexagonal cells at 16 resolutions (0–15), each roughly 7\(\times\) finer than the last. Cell IDs are 64-bit integers encoded as hexadecimal strings (e.g., "8528342bfffffff").

H3 is widely adopted: the FCC uses it for broadband mapping, Foursquare for POI analytics, and DuckDB/BigQuery both have native H3 extensions.

Key difference from ISEA: H3 cells are not equal-area. Hexagon area varies by about a factor of 2 between the largest and smallest cell at any given resolution, and by about 2.4 once the twelve pentagons are counted. The variation follows position on the icosahedron rather than latitude: at resolution 3 the hexagons within 5 degrees of a face centre average 9308 km\(^2\), against 12406 km\(^2\) for those more than 30 degrees away. For rigorous equal-area analysis, use ISEA. For interoperability with H3 ecosystems, use type = "h3".

Getting Started

Create an H3 grid by passing type = "h3" to hex_grid():

library(sf)
#> Linking to GEOS 3.14.1, GDAL 3.12.2, PROJ 9.7.1; sf_use_s2() is TRUE
library(ggplot2)

# Create an H3 grid specification
grid_h3 <- hex_grid(resolution = 5, type = "h3")
#> H3 cells are not exactly equal-area; hexagon area varies ~2x within a resolution.
#> This message is displayed once per session.
grid_h3
#> <HexGridInfo> H3 resolution 5: 2,016,842 cells, mean area 252.9 km^2 (not equal-area)

Then use hexify() with the grid object, just like ISEA:

# Sample cities
cities <- data.frame(
  name = c("Vienna", "Paris", "Madrid", "Berlin", "Rome",
           "London", "Prague", "Warsaw", "Budapest", "Amsterdam"),
  lon = c(16.37, 2.35, -3.70, 13.40, 12.50,
          -0.12, 14.42, 21.01, 19.04, 4.90),
  lat = c(48.21, 48.86, 40.42, 52.52, 41.90,
          51.51, 50.08, 52.23, 47.50, 52.37)
)

result <- hexify(cities, lon = "lon", lat = "lat", grid = grid_h3)
result
#> <HexData> 10 rows, 3 columns, 10 cells (data.frame)
#>   grid: H3 resolution 5: 2,016,842 cells, mean area 252.9 km^2 (not equal-area)

H3 cell IDs are character strings, unlike ISEA’s numeric IDs:

# Cell IDs are hexadecimal strings
result$cell_id
#>  [1] "851e15b7fffffff" "851fb467fffffff" "85390ca3fffffff" "851f1d4bfffffff"
#>  [5] "851e8053fffffff" "85194ad3fffffff" "851e3543fffffff" "851f53cbfffffff"
#>  [9] "851e037bfffffff" "85196953fffffff"

# All standard accessors work
cells(result)
#>  [1] "851e15b7fffffff" "851fb467fffffff" "85390ca3fffffff" "851f1d4bfffffff"
#>  [5] "851e8053fffffff" "85194ad3fffffff" "851e3543fffffff" "851f53cbfffffff"
#>  [9] "851e037bfffffff" "85196953fffffff"
n_cells(result)
#> [1] 10

Choosing Resolution by Area

If you think in terms of cell area rather than resolution numbers, pass area_km2 instead of resolution. hexify picks the closest H3 resolution:

grid_area <- hex_grid(area_km2 = 500, type = "h3")
grid_area
#> <HexGridInfo> H3 resolution 5: 2,016,842 cells, mean area 252.9 km^2 (not equal-area)

Grid Generation

All grid generation functions work with H3 grids.

Rectangular Region

# Generate H3 hexagons over Western Europe
grid_h3 <- hex_grid(resolution = 3, type = "h3")
europe_h3 <- grid_rect(c(-10, 35, 25, 60), grid_h3)

# Basemap
europe <- hexify_world[hexify_world$continent == "Europe", ]

ggplot() +
  geom_sf(data = europe, fill = "gray95", color = "gray60") +
  geom_sf(data = europe_h3, fill = NA, color = "#E6550D", linewidth = 0.4) +
  coord_sf(xlim = c(-10, 25), ylim = c(35, 60)) +
  labs(title = sprintf("H3 Resolution %d Grid", summary(grid_h3)$resolution),
       subtitle = sprintf("~%.0f km² average cell", summary(grid_h3)$area_km2)) +
  theme_minimal(base_size = FIG_BASE_SIZE)

Clipping to a Boundary

# Clip H3 grid to France
france <- hexify_world[hexify_world$name == "France", ]
grid_h3 <- hex_grid(resolution = 4, type = "h3")
france_h3 <- grid_clip(france, grid_h3)

ggplot() +
  geom_sf(data = france, fill = "gray95", color = "gray40", linewidth = 0.5) +
  geom_sf(data = france_h3, fill = alpha("#E6550D", 0.3),
          color = "#E6550D", linewidth = 0.3) +
  coord_sf(xlim = c(-5, 10), ylim = c(41, 52)) +
  labs(title = sprintf("H3 Grid Clipped to France (res %d)", summary(grid_h3)$resolution)) +
  theme_minimal(base_size = FIG_BASE_SIZE)

Hierarchical Navigation

H3’s killer feature is its clean hierarchical structure: every cell has exactly one parent and seven children. hexify exposes this with get_parent() and get_children().

Parents

# Get parent cells (one resolution coarser)
grid_h3 <- hex_grid(resolution = 5, type = "h3")
child_ids <- lonlat_to_cell(
  lon = c(16.37, 2.35, 13.40),
  lat = c(48.21, 48.86, 52.52),
  grid = grid_h3
)

parent_ids <- get_parent(child_ids, grid_h3, levels = 1)
data.frame(child = child_ids, parent = parent_ids)
#>             child          parent
#> 1 851e15b7fffffff 841e15bffffffff
#> 2 851fb467fffffff 841fb47ffffffff
#> 3 851f1d4bfffffff 841f1d5ffffffff

Children

# Get children of a single cell (one resolution finer)
grid_coarse <- hex_grid(resolution = 3, type = "h3")
coarse_id <- lonlat_to_cell(16.37, 48.21, grid_coarse)

children <- get_children(coarse_id, grid_coarse, levels = 1)
cat(length(children[[1]]), "children at resolution", summary(grid_coarse)$resolution + 1, "\n")
#> 7 children at resolution 4
head(children[[1]])
#> [1] "841e151ffffffff" "841e153ffffffff" "841e155ffffffff" "841e157ffffffff"
#> [5] "841e159ffffffff" "841e15bffffffff"

Visualizing the Hierarchy

# Parent cell polygon
parent_poly <- cell_to_sf(coarse_id, grid_coarse)

# Children cell polygons
grid_fine <- hex_grid(resolution = 4, type = "h3")
children_poly <- cell_to_sf(children[[1]], grid_fine)

ggplot() +
  geom_sf(data = children_poly, fill = alpha("#E6550D", 0.3),
          color = "#E6550D", linewidth = 0.5) +
  geom_sf(data = parent_poly, fill = NA, color = "black", linewidth = 1.2) +
  labs(title = sprintf("H3 Hierarchy: 1 parent (res %d)", summary(grid_coarse)$resolution),
       subtitle = sprintf("%d children (res %d)",
                          length(children[[1]]), summary(grid_fine)$resolution)) +
  theme_minimal(base_size = FIG_BASE_SIZE)

Working with H3 Data

The same hexify workflow applies. Species observations across Europe, aggregated into H3 cells:

set.seed(42)

# Simulate observations across Europe
obs <- data.frame(
  lon = c(rnorm(200, 10, 12), rnorm(100, 25, 8)),
  lat = c(rnorm(200, 48, 6), rnorm(100, 55, 4)),
  species = sample(c("Sp. A", "Sp. B", "Sp. C"), 300, replace = TRUE)
)
obs$lon <- pmax(-10, pmin(40, obs$lon))
obs$lat <- pmax(35, pmin(65, obs$lat))

# Hexify with H3
grid_h3 <- hex_grid(resolution = 3, type = "h3")
obs_hex <- hexify(obs, lon = "lon", lat = "lat", grid = grid_h3)

# Aggregate: species richness per cell
obs_df <- as.data.frame(obs_hex)

richness <- aggregate(species ~ cell_id, data = obs_df,
                      FUN = function(x) length(unique(x)))
names(richness)[2] <- "n_species"

# Map it
polys <- cell_to_sf(richness$cell_id, grid_h3)
polys <- merge(polys, richness, by = "cell_id")

europe <- hexify_world[hexify_world$continent == "Europe", ]

ggplot() +
  geom_sf(data = europe, fill = "gray95", color = "gray70", linewidth = 0.2) +
  geom_sf(data = polys, aes(fill = n_species), color = "white", linewidth = 0.3) +
  scale_fill_viridis_c(option = "plasma", name = "Species\nRichness") +
  coord_sf(xlim = c(-10, 40), ylim = c(35, 65)) +
  labs(title = "Species Richness on H3 Grid",
       subtitle = sprintf("H3 resolution %d (~%.0f km² avg cells)",
                          summary(grid_h3)$resolution, summary(grid_h3)$area_km2)) +
  theme_minimal(base_size = FIG_BASE_SIZE) +
  theme(axis.text = element_blank(), axis.ticks = element_blank())

ISEA–H3 Crosswalk

h3_crosswalk() (v0.6.0) maps between ISEA and H3 cell IDs in both directions. If you run your analysis on ISEA but collaborators expect H3 cell IDs, this bridges the two.

# Start with an ISEA grid and some cells
grid_isea <- hex_grid(resolution = 9, aperture = 3)
isea_ids <- lonlat_to_cell(
  lon = c(16.37, 2.35, 13.40, -3.70, 12.50),
  lat = c(48.21, 48.86, 52.52, 40.42, 41.90),
  grid = grid_isea
)

# Map ISEA cells to their closest H3 equivalents
xw <- h3_crosswalk(isea_ids, grid_isea)
xw[, c("isea_cell_id", "h3_cell_id", "isea_area_km2", "h3_area_km2")]
#>   isea_cell_id      h3_cell_id isea_area_km2 h3_area_km2
#> 1        42280 841e15bffffffff      2591.402    1753.173
#> 2        39597 841fb43ffffffff      2591.402    1569.204
#> 3        40823 841f1d5ffffffff      2591.402    1570.136
#> 4        39585 84390cbffffffff      2591.402    1819.493
#> 5        42516 841e80dffffffff      2591.402    1885.191

The area_ratio column shows how ISEA and H3 cell sizes compare; values close to 1 mean the resolutions are well-matched.

ISEA vs H3: When to Use Which

	ISEA	H3
Cell area	Exactly equal	~2\(\times\) variation (~2.4\(\times\) with pentagons)
Cell IDs	Numeric (integer)	Character (hex string)
Apertures	3, 4, 7, 4/3	Fixed (7)
Resolutions	0–30	0–15
Hierarchy	Approximate (aperture-dependent)	Exact (7 children per parent)
Dependencies	None (built-in C++)	None (vendored H3 C library)
Industry adoption	Scientific / government	Tech industry / commercial

Use ISEA when:

Equal-area cells are required (biodiversity surveys, density estimation, statistical sampling)
You need fine control over aperture and resolution
No external dependencies are acceptable

Use H3 when:

Interoperability with H3 ecosystems (Uber, Foursquare, DuckDB, BigQuery)
Clean hierarchical operations (parent/child traversal) are a priority
Slight area variation across latitudes is acceptable for your analysis

Resolution Reference

h3_res <- hexify_compare_resolutions(type = "h3", res_range = 0:15)
h3_res$n_cells_fmt <- ifelse(
  h3_res$n_cells > 1e9,
  sprintf("%.1fB", h3_res$n_cells / 1e9),
  ifelse(h3_res$n_cells > 1e6,
         sprintf("%.1fM", h3_res$n_cells / 1e6),
         ifelse(h3_res$n_cells > 1e3,
                sprintf("%.1fK", h3_res$n_cells / 1e3),
                as.character(h3_res$n_cells)))
)
# Use scientific notation for tiny areas, fixed for larger ones
h3_res$area_fmt <- ifelse(
  h3_res$cell_area_km2 < 0.1,
  formatC(h3_res$cell_area_km2, format = "e", digits = 1),
  formatC(h3_res$cell_area_km2, format = "f", digits = 1, big.mark = ",")
)
h3_res$spacing_fmt <- ifelse(
  h3_res$cell_spacing_km < 0.1,
  formatC(h3_res$cell_spacing_km, format = "e", digits = 1),
  formatC(h3_res$cell_spacing_km, format = "f", digits = 1)
)
knitr::kable(
  h3_res[, c("resolution", "n_cells_fmt", "area_fmt", "spacing_fmt")],
  col.names = c("Resolution", "# Cells", "Avg Area (km\u00b2)", "Spacing (km)"),
  align = c("r", "r", "r", "r")
)

Resolution	# Cells	Avg Area (km²)	Spacing (km)
0	122	4,357,449.4	2243.1
1	842	609,788.4	839.1
2	5.9K	86,801.8	316.6
3	41.2K	12,393.4	119.6
4	288.1K	1,770.3	45.2
5	2.0M	252.9	17.1
6	14.1M	36.1	6.5
7	98.8M	5.2	2.4
8	691.8M	0.7	0.9
9	4.8B	0.1	0.3
10	33.9B	1.5e-02	0.1
11	237.3B	2.2e-03	5.0e-02
12	1661.0B	3.1e-04	1.9e-02
13	11626.7B	4.4e-05	7.1e-03
14	81386.8B	6.3e-06	2.7e-03
15	569707.4B	9.0e-07	1.0e-03