Telling a story of CO₂ emissions over time
Application exercise
Clean and visualize data on CO₂ emissions to tell a compelling story about changes in emissions rankings over time.
ImportantGetting started
This application exercise is designed to be run in your web browser using the {webr} framework. Simply work through the exercises and use the provided code cells to execute live R code in your browser.
Data
The dataset comes from the World Bank and contains information on CO₂ emissions, GDP per capita, and population for 215 countries from 1995 to 2022. The data is available in the file wdi_co2.csv. We will import the data and perform some basic cleaning tasks.
# import raw data obtained using WDI API
wdi_co2_raw <- read_csv("data/wdi_co2.csv")
# clean the data
wdi_clean <- wdi_co2_raw |>
# remove observations that are not actually countries
filter(region != "Aggregates") |>
# select relevant columns and rename to make it easier
select(
iso2c,
iso3c,
country,
year,
population = SP.POP.TOTL,
co2_emissions = EN.ATM.CO2E.PC,
gdp_per_cap = NY.GDP.PCAP.KD,
region,
income
)
wdi_clean# A tibble: 6,020 × 9
iso2c iso3c country year population co2_emissions gdp_per_cap region income
<chr> <chr> <chr> <dbl> <dbl> <dbl> <dbl> <chr> <chr>
1 AF AFG Afghani… 2001 19688632 0.0553 NA South… Low i…
2 AF AFG Afghani… 1998 18493132 0.0713 NA South… Low i…
3 AF AFG Afghani… 2009 27385307 0.240 490. South… Low i…
4 AF AFG Afghani… 2000 19542982 0.0552 NA South… Low i…
5 AF AFG Afghani… 2012 30466479 0.335 571. South… Low i…
6 AF AFG Afghani… 1996 17106595 0.0823 NA South… Low i…
7 AF AFG Afghani… 1999 19262847 0.0582 NA South… Low i…
8 AF AFG Afghani… 2002 21000256 0.0668 344. South… Low i…
9 AF AFG Afghani… 2003 22645130 0.0730 347. South… Low i…
10 AF AFG Afghani… 2004 23553551 0.0549 339. South… Low i…
# ℹ 6,010 more rowsNext we rank countries based on their CO₂ emissions in 1995 and 2020, and then calculate the difference in rankings. We also create a variable that indicates if the rank changed by a lot (more than 30 positions). This requires substantial data cleaning and wrangling.
co2_rankings <- wdi_clean |>
# Get rid of smaller countries
filter(population > 200000) |>
# Only look at two years
filter(year %in% c(1995, 2020)) |>
# Get rid of all the rows that have missing values in co2_emissions
drop_na(co2_emissions) |>
# Look at each year individually and rank countries based on their emissions that year
mutate(
ranking = rank(co2_emissions),
.by = year
) |>
# Only select required columns
select(iso3c, country, year, region, income, ranking) |>
# pivot long
pivot_wider(
names_from = year,
names_prefix = "rank_",
values_from = ranking
) |>
# Find the difference in ranking between 2020 and 1995
mutate(rank_diff = rank_2020 - rank_1995) |>
# Remove all rows where there's a missing value in the rank_diff column
drop_na(rank_diff) |>
# Make an indicator variable that is true of the absolute value of the
# difference in rankings is greater than 30
mutate(big_change = if_else(abs(rank_diff) >= 30, TRUE, FALSE)) |>
# Make another indicator variable that indicates if the rank improved by a
# lot, worsened by a lot, or didn't change much.
mutate(
better_big_change = case_when(
rank_diff <= -30 ~ "Rank improved",
rank_diff >= 30 ~ "Rank worsened",
.default = "Rank changed a little"
)
) |>
# arrange rows by rank_diff for printing
arrange(rank_diff)Here is what the data looked like originally:
slice_head(wdi_clean, n = 5)# A tibble: 5 × 9
iso2c iso3c country year population co2_emissions gdp_per_cap region income
<chr> <chr> <chr> <dbl> <dbl> <dbl> <dbl> <chr> <chr>
1 AF AFG Afghanis… 2001 19688632 0.0553 NA South… Low i…
2 AF AFG Afghanis… 1998 18493132 0.0713 NA South… Low i…
3 AF AFG Afghanis… 2009 27385307 0.240 490. South… Low i…
4 AF AFG Afghanis… 2000 19542982 0.0552 NA South… Low i…
5 AF AFG Afghanis… 2012 30466479 0.335 571. South… Low i…And here is what it looks like after cleaning:
slice_head(co2_rankings, n = 5)# A tibble: 5 × 9
iso3c country region income rank_1995 rank_2020 rank_diff big_change
<chr> <chr> <chr> <chr> <dbl> <dbl> <dbl> <lgl>
1 ZWE Zimbabwe Sub-S… Lower… 75 39 -36 TRUE
2 DNK Denmark Europ… High … 160 127 -33 TRUE
3 SWE Sweden Europ… High … 132 100 -32 TRUE
4 SYR Syrian Arab Repu… Middl… Low i… 96 64 -32 TRUE
5 MLT Malta Middl… High … 128 99 -29 FALSE
# ℹ 1 more variable: better_big_change <chr>
ImportantInteractive table of cleaned data
Basic plot
Let’s create a basic plot that visualizes the changes in CO₂ emission rankings between 1995 and 2020.
Brainstorm improvements
Your turn: Brainstorm methods to improve the readability and interpretability of the chart through annotations.
TipPotential aspects to emphasize
- What is a “good” rank? What is a “bad” rank? Keep in mind that 1 is the country with the lowest carbon emissions per capita, and 170 is the country with the highest carbon emissions per capita.
- What are the countries that have significantly improved or worsened their rank?
- What other aspects do you feel should be emphasized?
TipMethods for annotation
- Text labels
- Arrows/lines
- Rectangles
- Colors/fills
Implement your annotations
Now try implementing your proposed annotations in the code cell below. The basic plot code is provided as a starting point. The {ggrepel} and {ggtext} packages are available if you want to use them.
NoteOne possible approach
library(ggrepel)
library(ggtext)
ggplot(
data = co2_rankings,
mapping = aes(x = rank_1995, y = rank_2020)
) +
# Add a reference line that goes from the bottom corner to the top corner
annotate(
geom = "segment",
x = 0,
xend = max(co2_rankings$rank_1995),
y = 0,
yend = max(co2_rankings$rank_2020)
) +
# Add points and color them by the type of change in rankings
geom_point(aes(color = better_big_change)) +
# Add repelled labels. Only use data where big_change is TRUE. Fill them by
# the type of change (so they match the color in geom_point() above) and use
# white text
geom_label_repel(
data = filter(co2_rankings, big_change == TRUE),
aes(label = country, fill = better_big_change),
color = "white",
seed = 895
) +
# Add notes about what the outliers mean in the bottom left and top right
# corners. These are italicized and light grey. The text in the bottom corner
# is justified to the right with hjust = 1, and the text in the top corner is
# justified to the left with hjust = 0
annotate(
geom = "text",
x = 167,
y = 6,
label = "Outliers improving",
fontface = "italic",
hjust = 1,
color = "grey50"
) +
annotate(
geom = "text",
x = 2,
y = 167,
label = "Outliers worsening",
fontface = "italic",
hjust = 0,
color = "grey50"
) +
# Use three different colors for the points
scale_color_manual(values = c("grey50", "#0074D9", "#FF4136")) +
# Use two different colors for the filled labels.
# There are no grey labels, so we don't need that color
scale_fill_manual(values = c("#0074D9", "#FF4136")) +
# Make the x and y axes expand all the way to the edges of the plot area and
# add breaks every 25 units from 0 to 175
scale_x_continuous(expand = c(0, 0), breaks = seq(0, 175, 25)) +
scale_y_continuous(expand = c(0, 0), breaks = seq(0, 175, 25)) +
# Add labels! There are a couple fancy things here.
# 1. In the title we wrap the 2 of CO2 in the HTML <sub></sub> tag so that the
# number gets subscripted. The only way this will actually get parsed as
# HTML is if we tell the plot.title to use element_markdown() in the
# theme() function, and element_markdown() comes from the ggtext package.
# 2. In the subtitle we bold the two words **improved** and **worsened** using
# Markdown asterisks. We also wrap these words with HTML span tags with
# inline CSS to specify the color of the text. Like the title, this will
# only be processed and parsed as HTML and Markdown if we tell the
# plot.subtitle to use element_markdown() in the theme() function.
labs(
x = "Rank in 1995",
y = "Rank in 2020",
title = "Changes in CO<sub>2</sub> emission rankings between 1995 and 2020",
subtitle = "Countries that <span style='color: #0074D9'>**improved**</span> or <span style='color: #FF4136'>**worsened**</span> more than 30 positions in the rankings highlighted",
caption = "Source: The World Bank.\nCountries with populations of less than 200,000 excluded."
) +
# Turn off the legends for color and fill, since the subtitle includes that
guides(color = "none", fill = "none") +
# Use theme_bw() with Roboto Condensed
theme_bw(base_family = "Roboto Condensed") +
# Tell the title and subtitle to be treated as Markdown/HTML, make the title
# 1.6x the size of the base font, and make the subtitle 1.3x the size of the
# base font. Also add a little larger margin on the right of the plot so that
# the 175 doesn't get cut off.
theme(
plot.title = element_markdown(face = "bold", size = rel(1.6)),
plot.subtitle = element_markdown(size = rel(1.3)),
plot.margin = unit(c(0.5, 1, 0.5, 0.5), units = "lines")
)
Acknowledgments
- Exercise drawn from Data Visualization with R by Andrew Heiss and licensed under CC BY-NC 4.0.