geom_sf with continuous values and NA ggplot R

Question

I have an sf-object that I would like to plot. It is a map over counties in Sweden (21 counties) and for each county, I have a value. This value however can be either a positive or negative value, or even an NA. I would like to plot the map over Sweden and color all counties that have NA as white, and then color all other counties with a gradient (depending on value) for all the continuous values. But when I try I only get this error message: Error: Continuous value supplied to discrete scale

This is not my real data but I am not sure how to demonstrate sf-object. So here is just a dataframe with name of county and values (so no coordinates or anything):

data <- data.frame(Diff = c("NA", "NA", "5", "6.89", "9", "-4", "3.56"), 
                   County = c("Halland", "Gotland", "Skane", "Jonkoping", "Gotaland",
                              "Blekinge", "Dalarna"))

ggplot(data) +
  geom_sf(aes(fill = Diff), color = "black") +
  scale_fill_manual(values = c("blue", "yellow", viridis::viridis(4))) +
  coord_sf(datum = NA)

So I would like to plot the whole map and if a county have an NA, the county should be white. And then I would like to have different color gradient for the values (ranged from negative to positive values) for each county.

Have enayone experienced the same problem?

I'm afraid I get a different error for your example: Error: stat_sf requires the following missing aesthetics: geometry. I haven't used sf before but dput would help to capture the data. My initial thought is that the Diff is stored as a character. Try converting this to a numeric or an integer if that's what's expected e.g. as.integer(Diff) — Jonny Phelps
Ah ok, the Diff is stored as a character in the above example as it's wrapped in quotes. E.g. try class(data$Diff) — Jonny Phelps
Yes, that is correct. As I mentioned in my post, I am not sure how to demonstrate an sf-object, so I just demonstrated a dataframe with values and counties. But you are correct, in my dataframe, all my values are characters. But in my original sf-object they are numeric. — paula456
I guess that you get the error because with scale_fill_manual you are specifying a discrete scale with 6 colors, while your variable is continuous. Why not using scale_fill_viridis_c(na.value = "white")? Another option would be scale_fill_stepsn(colors = c("blue", viridis::viridis(4), "yellow"), na.value = "white"). — stefan

Bappa Das Bappa Das · Accepted Answer · 2021-04-09T09:48:02

The following code works for me

library(sf)
library(tidyverse)

#Downloading data from DIVA GIS website
get_sweden_map <- function(cong=113) {
  tmp_file <- tempfile()
  tmp_dir  <- tempdir()
  zp <- sprintf("http://biogeo.ucdavis.edu/data/diva/adm/SWE_adm.zip",cong)
  download.file(zp, tmp_file)
  unzip(zipfile = tmp_file, exdir = tmp_dir)
  fpath <- paste(tmp_dir)
  st_read(fpath, layer = "SWE_adm1")
}

swe <- get_sweden_map(114)

#Plot the Counties
plot(swe[5])

data <- tibble(Diff = c("NA", "NA", "5", "6.89", "9", "-4", "3.56"), 
                   County = c("Halland", "Gotland", "Skane", "Jonkoping", "Gotaland",
                              "Blekinge", "Dalarna"))

#Join the data to the swe shapefile
sweden <- swe %>% left_join(data, by=c("NAME_1"="County"))

#Plotting the data
ggplot(sweden) +
  geom_sf(aes(fill = Diff), color = "black") +
  scale_fill_manual(values = c("blue", "yellow", viridis::viridis(4))) +
  coord_sf(datum = NA)

geom_sf with continuous values and NA ggplot R

1 Answers