Boxplot with trend line

Posted at — Jul 26, 2019

Last week my colleague showed me a drawn graph and asked can we plot this kind of graph. It is a graph with grouped boxplot. Additionally, it contains a trend line that connects mean of each data box. So this post explains how to draw a graph like that using two ways;

r ggplot2 library
r base plots (using codes from The R Graph Gallery)

This data is sample data set from my colleague. The data contains VI (Vegetation Index) time series values for three different crops extracted from Landsat 8 data.

Load data

Load libraries

library(tidyverse)
library(RColorBrewer)

Read and prepare data

# Read data
read.df <-
  read.csv("D:/Blog_data/jblog/static/data/sample_ts.csv", header = TRUE)

# Make a copy of the read table
sample.df <- read.df

# Checking the levels in the date column
levels(sample.df$date)

>> [1] "04-Jul-15" "07-Aug-15" "10-May-15" "24-Apr-15"

# Checking the levels in the crop column
levels(sample.df$crop)

>> [1] "maize"      "potato"     "sugar beet"

# Reorder levels
sample.df$date <-
  factor(
    sample.df$date,
    levels = c("24-Apr-15", "10-May-15",
               "04-Jul-15", "07-Aug-15"),
    labels = c("April", "May", "July", "August")
  )

Plot with ggplot library

# Create a color palette for three crops
my_colors = c("greenyellow", "gold3", "darksalmon")

# Create a plot uisng ggplot
ggplot(data = sample.df, aes(x = date, y = NDVI, fill = crop)) +
  geom_boxplot() + ylim(c(-0.2, 1.2)) +
  stat_summary(
    fun.y = mean,
    colour = "red",
    geom = "point",
    position = position_dodge(width = 0.75)
  ) +
  stat_summary(
    fun.y = mean,
    colour = "red",
    aes(group = 1),
    geom = "line",
    lwd = 1,
    lty = 1
  ) +
  theme_bw(base_size = 12) +
  scale_fill_manual(
    values = my_colors,
    name = "Crop Type",
    labels = c("Maize", "Potato", "Sugar Beet")
  ) +
  labs(
    x = "Month",
    y = "VI",
    title = "Phenology of the selected crops",
    subtitle = "VI extracted from Landsat 8 time series data"
  )

Plot with base library

# Make the boxplot using base
boxplot(
  sample.df$NDVI ~ sample.df$crop + sample.df$date ,
  ylim = c(-0.2, 1.2),
  xaxt = "n" ,
  xlab = "Date" ,
  col = my_colors ,
  pch = 20 ,
  cex = 0.3 ,
  ylab = "VI",
  xlab = "Month",
  main = "Phenology of the selected crops \n VI extracted from Landsat 8 time series data"
)
abline(v = seq(0, 3 * 4, 3) + 0.5 , col = "grey")
axis(1, labels = levels(sample.df$date) , at = seq(2, 3 * 4, 3))

# Add general trend
a = aggregate(sample.df$NDVI , by = list(sample.df$crop, sample.df$date) , mean)
lines(a[, 3], type = "l" , col = "red" , lwd = 2)

#Add legend for crops
legend(
  "top",
  fill = my_colors,
  legend = c("Maize", "Potato", "Sugar Beet"),
  horiz = T
)

Jayan Wijesingha

Have a great day…!