Problem with key in ggplot, using double axis y

Question

I would like a help to solve this problem. I would like the caption to present dotted lines (linetype), dotted (shape) and colors (colour) equal to the graph, but this graph has a double axis for y, so I had to add the separate lines, because there is a transformation, to be on the same plotting scale.

In the figure it is possible to observe that the legend displays only 1 type of shape.

library(ggplot2)
library(lubridate)

clima<-structure(list(data=structure(c(15489,15490,15491,15492,
15493,15494,15495,15496,15497),class="Date"),VV=c(1.63,
2.87,2.33,2.33,2.43,0,3.07,4.47,4.47),Tmax=c(29.63,
30.63,30.77,29.97,29.65,30.1,31.13,31.53,31.3),Tmin=c(19.6,
15.32,15.38,14.98,14.8,15.05,15.57,15.77,15.65),Tmed=c(24.62,
22.98,23.08,22.48,20.9,22.58,23.35,23.65,23.48),UR=c(79.8,
74.6,73.3,79.3,77.6,72.9,70.5,70.4,72.8),Ubu=c(54.72,
46.31,42.61,35.19,29.59,27.15,23.05,16.77,11.97),P=c(0L,
0L,0L,0L,0L,0L,0L,0L,0L)),row.names=c(NA,-9L),class="data.frame")

clima$data<-ymd(clima$data)


 coef<-10

ggplot(clima,aes(x=data))+
  geom_line(aes(y=VV,colour="V. Vento"),size=1,linetype=2)+
  geom_point(aes(y=VV,colour="V. Vento"),shape=15,size=3)+  
  geom_line(aes(y=UR/coef,colour="UR ar"),linetype=1,size=1)+ 
  geom_point(aes(y=UR/coef,colour="UR ar"),shape=16,size=3)+ 
  scale_y_continuous(name = expression("Velocidade do Vento (m s"^2*")"),sec.axis = sec_axis(~ . *coef,name=expression("Umidade relativa do ar (%)")))+
  scale_x_date("",date_labels ="%d/%m")+
  scale_colour_manual("",breaks=c("V. Vento","UR ar"),values = c("blue","black"))+
 #scale_shape_manual("",values=c(15,16))+
  #scale_linetype_manual("",values=c(2,1))+
  theme(
    panel.background = element_rect(fill=NA),
        panel.border = element_rect(fill=NA),
        axis.title.y = element_text(size = rel(1.5), angle = 90),
        legend.text = element_text(size=12,face = "bold"),
        axis.text = element_text(size=12,face="bold"),
        legend.position = "bottom")

Answer 1

This type of problem is usually easier to solve if the data is in the long format. This allows having a single variable to be used in aes() for colour, linetype, shape or any other. The code can get a little more complicated (this is a matter of opinion, I think not, it doesn’t) but it makes it easier when you plot the graph.

library(tidyverse)
library(ggplot2)
library(lubridate)

clima$data<-ymd(clima$data)

coef <- 10

clima %>%
  select(data, VV, UR) %>%
  mutate(UR = UR/coef) %>%
  rename(`V. Vento` = VV,
         `UR ar` = UR) %>%
  pivot_longer(
    cols = c("V. Vento", "UR ar"),
    names_to = "Variavel",
    values_to = "Valor"
  ) %>%
  ggplot(aes(data, Valor, colour = Variavel)) +
  geom_point(aes(shape = Variavel), size = 3) +
  geom_line(aes(linetype = Variavel), size = 1) +
  scale_y_continuous(name = expression("Velocidade do Vento (m s"^2*")"),
                     sec.axis = sec_axis(~ . *coef,name=expression("Umidade relativa do ar (%)")))+
  scale_x_date("", date_labels ="%d/%m") +
  scale_color_manual(values = c("blue", "black")) +
  scale_shape_manual(values = c(15, 16)) +
  scale_linetype_manual(values = c(2, 1)) +
  theme(
    panel.background = element_rect(fill=NA),
    panel.border = element_rect(fill=NA),
    axis.title.y = element_text(size = rel(1.5), angle = 90),
    legend.title = element_blank(),
    legend.text = element_text(size=12,face = "bold"),
    axis.text = element_text(size=12,face="bold"),
    legend.position = "bottom")