Inaccuracy in country polygons which cross 180th meridian in world dataset

[HRodenhizer]

Thanks to some help over on SO, I discovered some slight offsets between portions of MULTIPOLYGONs that cross the 180th meridian in the world dataset.

It looks like this is an issue for Russia, Fiji, and Antarctica.

library(sf)
library(spData)
library(dplyr)
library(ggplot)

meridian_180 <- tibble(lon = c(-180,-180),
                       lat = c(-90, 90)) %>%
  st_as_sf(coords = c("lon","lat"),
           crs = "EPSG:4326") %>%
  summarise(geometry = st_combine(geometry)) %>%
  st_cast("LINESTRING") %>% 
  smoothr::densify(max_distance = 1)


world_180 <- world %>%
  slice(st_intersects(., meridian_180, sparse = FALSE) %>%
          as_vector() %>%
          which())

fiji = world_180$geom[1]
russia = world_180$geom[2]
antarctica = world_180$geom[3]
st_distance(st_cast(fiji, 'POLYGON')[c(1, 3)])

st_distance(st_cast(russia, 'POLYGON')[c(1, 6)])

antarctica <- antarctica %>%
  st_transform(crs = 'EPSG:3412')

ggplot(antarctica) +
  geom_sf()

[Nowosad]

Hi @HRodenhizer that was done on purpose (to avoid some invalid geometry problems with s2).

[Nowosad]

(Also, importantly datasets from spData are mostly suitable for teaching, benchmarking and testing; I would not suggest using them for research)

[HRodenhizer]

Got it - that makes sense. I find that this dataset is extremely useful for low-detail, large extent visualizations, but thanks for the warning!

[Robinlovelace]

Good to have this know bug/feature documented for anyone else who happens across this.

[Nowosad]

Context: r-spatial/sf#1649 (comment)