Use two colour scales possible (with work around)?

Question

I'm trying to plots insect counts of 2 species in 18 experimental plots onto a single graph. Since the second species population peaks later, it is visually doable (see picture below). I would like the 18 population lines from species 1 to be green (using "Greens" from RColorBrewer) and the 18 of species 2 to be red (using "Reds"). I do realize this may be problematic for a colourblind audience, but that is irrelevant here.

I've read here that it is not possible with standard ggplot2 options: R ggplot two color palette on the same plot but this post is more than two years old.

There is a short of "cheat" for points: Using two scale colour gradients ggplot2 but since I prefer lines to show the population through time, I can't use it.

Are there any new "cheats" available for this? Or does anyone have another idea to visualize my data in a way that shows population trends through time in all plots and shows the difference in timing of the peak? I've included a picture at the bottom that shows my real data, all in the same colour scale though.

Sample code

# example data frame
plot <- as.factor(rep(c("A","B","C"),each=5))
time <- as.numeric(rep(c(1:5),times=3))
S1 <- c(1,4,7,5,2, 2,8,9,3,1, 1,6,6,3,1)
S2 <- c(0,0,2,3,2, 1,2,1,5,3, 0,1,1,6,7)
df <- data.frame(time, plot, S1, S2)

# example colour scales
S1Colours <- colorRampPalette(brewer.pal(9,"Greens"))(3)
S2Colours <- colorRampPalette(brewer.pal(9,"Reds"))(3)
names(S1Colours) <- levels(df$plot)
names(S2Colours) <- levels(df$plot)

# example plot
ggplot(data=df) +
  geom_line(aes(x=time, y=S1, colour=plot)) +
  geom_line(aes(x=time, y=S2, colour=plot)) +
  scale_colour_manual(name = "plot", values = S1Colours) +
  scale_colour_manual(name = "plot", values = S2Colours)
  # this gives the note "Scale for 'colour' is already present. Adding another scale for 'colour', which will replace the existing scale."

Plot real data

Maybe you could use different line types for S1 and S2 of the same group? — mt1022, Dec 18 '19 at 11:53
That's indeed something I've tried before, but because I have 36 lines with quite some overlap, it is not very visible and that's why I prefer different colours. — Tingolfin, Dec 18 '19 at 11:55
Have you tried a manual colour scale? Creating your required theme by concatenating 18 values picked from a light green to dargkreen colour ramp and another 18 from a light red to darkred ramp? (I don't have at hand the instructions on how to create continuous colour palettes from two colours and then sampling n colours from them, but it's only a couple lines of code). — cymon, Dec 18 '19 at 12:09
but then I would create a legend of 36 plots and will probably have to rename 18 of them to "plotS2" or someting. Though that might actually be a solution if I just cut the legend and assume the reader is not necessarily interested in which line is which exact plot... I'll certainly give this a try! — Tingolfin, Dec 18 '19 at 13:25

TobiO · Accepted Answer · 2019-12-18T13:50:55.810

2

I also would go by creating a manual color scale for all the combinations.

library(tidyverse)
library(RColorBrewer)
df_long=pivot_longer(df,cols=c(S1,S2),names_to = "Species",values_to = "counts") %>% # create long format and
  mutate(plot_Species=paste(plot,Species,sep="_")) # make identifiers for combined plot and Species 


#make color palette
mycolors=c(colorRampPalette(brewer.pal(9,"Greens"))(sum(grepl("S1",unique(df_long$plot_Species)))),
         colorRampPalette(brewer.pal(9,"Reds"))(sum(grepl("S2",unique(df_long$plot_Species)))))
names(mycolors)=c(grep("S1",unique(df_long$plot_Species),value = T),
                grep("S2",unique(df_long$plot_Species),value = T))


# example plot
ggplot(data=df_long) +
  geom_line(aes(x=time, y=counts, colour=plot_Species)) +
  scale_colour_manual(name = "Species by plot", values = mycolors)

edited Dec 18 '19 at 13:50

answered Dec 18 '19 at 13:46

TobiO

1,335
1
9
24

1

[You should use `library` instead of `require`.](https://stackoverflow.com/a/51263513/1968) – Konrad Rudolph Dec 18 '19 at 13:49
Thanks! I learned something :-) and edited my answer accordingly – TobiO Dec 18 '19 at 13:51
1

@Tingolfin If you want to have the legend ordered by species, just switch the order of variables in the `plot_Species=paste()` statement – TobiO Dec 18 '19 at 13:58

score 2 · Answer 2 · answered Dec 19 '19 at 17:36

You can do this easily with the ggnewscale package (disclaimer: I'm the author).

This is how you would do it:

library(RColorBrewer)
library(ggplot2)
library(ggnewscale)


plot <- as.factor(rep(c("A","B","C"),each=5))
time <- as.numeric(rep(c(1:5),times=3))
S1 <- c(1,4,7,5,2, 2,8,9,3,1, 1,6,6,3,1)
S2 <- c(0,0,2,3,2, 1,2,1,5,3, 0,1,1,6,7)
df <- data.frame(time, plot, S1, S2)

# example colour scales
S1Colours <- colorRampPalette(brewer.pal(9,"Greens"))(3)
S2Colours <- colorRampPalette(brewer.pal(9,"Reds"))(3)
names(S1Colours) <- levels(df$plot)
names(S2Colours) <- levels(df$plot)


ggplot(data=df) +
  geom_line(aes(x=time, y=S1, colour=plot)) +
  scale_colour_manual(name = "plot 1", values = S1Colours) +

  new_scale_color() +

  geom_line(aes(x=time, y=S2, colour=plot)) +
  scale_colour_manual(name = "plot 2", values = S2Colours)

^{Created on 2019-12-19 by the reprex package (v0.3.0)}

your package is super nice, congrats! – Bruno Conte Leite Jan 19 '20 at 18:50 — Bruno Conte Leite, Jan 19 '20 at 18:50

Use two colour scales possible (with work around)?

2 Answers2