ggplot2 - how to assign geom_text with arrow icon to second yaxis scale

Question

I have 20 sets of the following dataset and would like to combine two plots into one with two y-axes that properly reflects the variable values.

hour=c(0:23)
conc=c(20.7, 19.4, 15.6, 11.7, 10.3, 9.1, 9.7, 10.6, 12.7, 12.6, 11, 9.8,
       9.3, 9.3, 10.3, 12.9, 17.1, 22.5, 22, 22.3, 23, 24.8, 24, 26.8)
ws=c(0.75, 0.7, 0.68, 0.64, 0.61, 0.64, 0.57, 0.58, .62, .65, .85, 1.12, 
     1.21, 1.29, 1.2, 1.18, 1.0, 0.84, .87, .75, .69, .69, .77, .74)
wd=c(295, 299, 288, 290, 292, 296, 306, 303, 300, 293, 259, 231, 94, 119, 
     95, 47, 75, 306, 302, 304, 319, 309, 290, 298)
mydf=data.frame(hour, conc, ws, wd)

I can plot them separately as a 1-by-2 graph using the following code:

conchrly <- ggplot(mydf, aes(x=hour,y=conc)) +
  geom_line() + 
  xlab("Hour") +
  ylab("Concentration") +
  theme(legend.position="bottom")     

windhrly <- ggplot(data = mydf, aes(x=hour, y=ws)) + 
  geom_text(aes(angle=-wd+90), label="←") +
  xlab("Hour") +
  ylab("Wind Speed") +
  theme(legend.position="bottom") 

conchrly + windhrly + plot_layout(ncol=1)

which gives this graph:

I'd like to combine them together as one graph with two y-axes. Most examples I found here are related to the second-axis as a duplicate or function of first y-values. My second y-variable is 'ws' which is not a function on my first y variable (ie. 'conc').

The following is my attempt to graph them together. However the y-value of second plot (geom_text which should represent my ws) are not plotted using the second y-axis.

ggplot(data=mydf, aes(x = hour, y = conc)) + 
  geom_line(color = "black") + 
  geom_text(data=mydf, aes(x=hour, y=ws,angle=-wd+90), label="←", colour = "blue") +
  scale_x_continuous(breaks = seq(0, 24, by=6), limits=c(0,24), name="Hour") + 
  scale_y_continuous( breaks = seq(0,30,10), name= "Concentration", 
     sec.axis = sec_axis(~./20, name = "Wind Speed (m/s)")) + 
              theme(axis.title.y = element_text(color = "black"),
                    axis.title.y.right = element_text(color = "black"))

How do I get those blue arrows which represent my 'ws' to be on the right scale (ie. using the second y-axis)? This is only one set and I have 19 others to do that's why I prefer to combine them into one graph per data set if possible.

Thanks in advance! :)

Answer 1

We can achieve what you're after if we scale values of ws such that they use the same range as conc, and then apply the inverse transformation inside sec_axis. The transformation that we are applying takes the form y = a + b * x, where y are the conc and x the ws measurements. The two scaling coefficients a and b are then obtained by solving a system of two linear equations:

# Calculate scaling coefficients a and b
y1 <- min(mydf$conc)
y2 <- max(mydf$conc)
x1 <- min(mydf$ws)
x2 <- max(mydf$ws)
b <- (y2 - y1) / (x2 - x1)
a <- y1 - b * x1

We then calculate ws.scaled and apply the inverse transformation x = (y - a) / b inside sec_axis.

library(tidyverse)
mydf %>%
    mutate(ws.scaled = a + b * ws) %>%
    ggplot(aes(x = hour, y = conc)) +
        geom_line(color = "black") +
        geom_text(aes(x = hour, y = ws.scaled, angle = -wd + 90), label = "←", colour = "blue") +
        scale_x_continuous(breaks = seq(0, 24, by=6), limits=c(0,24), name="Hour") +
        scale_y_continuous(
            breaks = seq(0,30,10),
            name=  "Concentration",
            sec.axis = sec_axis(~ (. - a) / b, name = "Wind Speed (m/s)")) +
            theme(
                axis.title.y = element_text(color = "black"),
                axis.title.y.right = element_text(color = "black"))

PS. Dual y axes are often not a good idea. Personally I prefer showing two plots.