Forests in Transition: Visualizing Global Deforestation

INFO 526 - Fall 2023 - Project 1

The Plotting Pandas - Megan, Shakir, Maria, Eshaan, Bharath

Our Dataset

  • “Global Deforestation” dataset from “Our World in Data”, tidytuesday.

  • Data looks at global deforestation trends and soybean production in Brazil.

  • Our Data:

    • forest dataset
    • soybean_use data

Our Dataset

Aim: Provide critical insights into the ever-evolving dynamics of global forests and the influence of soybean consumption, contributing to a better understanding of essential environmental conservation and sustainable land management practices.

Question 1

What does the global forest area look like over past decades?

  • The primary goal of this presentation is to demonstrate changes in forest cover from 1990 to 2015.

  • Decrease in forest cover over the past few decades.

  • While certain countries show an increase in forest cover, we are still losing our battle against deforestation.

  • International efforts, such as the United Nations REDD+ program, aim to reduce deforestation.

Approach

  • Getting geographical data from maps package
  • Roots of the approach
Pre-processing Function 
# function to pre process the forest dataset
# input  : dataset - tibble
#          unique_countries - tibble
# output : filtered_data - tibble

processForest <- function (dataset, unique_countries) {
  filtered_data <- dataset |>
    # filtering only entity, year and net_forest_conversion columns
    select(entity, year, net_forest_conversion) |>
    # getting all the countires which are not present in forest dataset for a specific years
    # bind_rows() is used combine combine rows of two data frames
    bind_rows(
      # anti_join() is used to return only the rows from the first dataset that isn't having matching rows in the second dataset based on specified key columns
      anti_join(unique_countries, dataset, by = c("region" = "entity")) |>
        # adding year and net_forest_conversion for that specific year as NA
        mutate(year = dataset[1, "year"], net_forest_conversion = NA)
    ) |>
    # renaming USA and UK so that both these countries are matching in world dataset and forest dataset
    mutate(
      entity = case_when(
        entity == "United States" ~ "USA",
        entity == "United Kingdom" ~ "UK",
        TRUE ~ entity
      )
    ) |>
    # creating a categorical variable forest_converstion to group countries based on their forest conversion
    mutate(
      entity = coalesce(entity, region),
      forest_converstion = case_when(
        net_forest_conversion < -400000 ~ "<-400k",
        net_forest_conversion < -200000 ~ "-400k to -200k",
        net_forest_conversion < -100000 ~ "-200k to -100k",
        net_forest_conversion < 0 ~ "-100k to 0",
        net_forest_conversion < 100000 ~ "0 to 100k",
        net_forest_conversion < 200000 ~ "100k to 200k",
        net_forest_conversion < 400000 ~ "200k to 400k",
        is.na(net_forest_conversion) ~ NA_character_,
        TRUE ~ ">400k"
      )
    ) |>
    # ordering forest_converstion column using factors based on the created categories
    mutate(
      forest_converstion = as_factor(forest_converstion) |>
        fct_relevel(
          "<-400k",
          "-400k to -200k",
          "-200k to -100k",
          "-100k to 0",
          "0 to 100k",
          "100k to 200k",
          "200k to 400k",
          ">400k"
        )
    )
  return(filtered_data)
}
  • Leveraging geom_map() from ggplot
  • Branches of the approach
Function used to generate the plot 
# Function for creating the ggplot map plot
# Using the filtered_forests$`2000` dataset created earlier as a data source
# using entity as map_id for first layer
# using forest_convestion as fill aesthetic and word as map for second layer
# using highlight_filtered_data$`2000` as another dataset for creating another map layer
# using entity as map_id,forest_convestion as fill aesthetic and highlight_world as map for third layer

# input  : year - integer
# output : world_plot - plot object

generateForestConversionPlot <- function(year) {
  world_plot <- ggplot(filtered_forests[[as.character(year)]], aes(map_id = entity)) +
    geom_map(
      aes(fill = forest_converstion),
      map   = world,
      color = "#B2BEB5",
      linewidth = 0.25,
      linetype  = "blank"
    ) +
    geom_map(
      data = highlight_filtered_data[[as.character(year)]],
      aes(map_id = entity, fill = forest_converstion),
      map   = highlight_world,
      color = "#71797E",
      show.legend = F
    ) +
    expand_limits(x = world$long, y = world$lat) +
    scale_fill_manual(values = color_mapping, na.value = "#F2F3F4") +
    coord_fixed(ratio = 1) +
    labs(
      title = paste("Net Forest Conversion by Country in", year),
      subtitle = "Net change in forest area measures forest expansion minus deforestation",
      caption = "Data source: Our World in Data",
      fill = "Net Forest Conversion (hectares)"
    ) +
    theme_void() +
    theme(
      legend.position = "bottom",
      legend.direction = "horizontal",
      plot.title = element_text(size = 19, face = "bold", hjust = 0.5),
      plot.subtitle = element_text(size = 15, color = "azure4", hjust = 0.5),
      plot.caption = element_text(size = 12, color = "azure4", hjust = 0.95)
    ) +
    guides(
      fill = guide_legend(
        nrow = 1,
        direction = "horizontal",
        title.position = "top",
        title.hjust = 0.5,
        label.position = "bottom",
        label.hjust = 1,
        label.vjust = 1,
        label.theme = element_text(lineheight = 0.25, size = 9),
        keywidth = 1,
        keyheight = 0.5
      )
    )
  return(world_plot)
}
  • Creating an animation of the generated plots

Forest Conversion Analysis

  • Notable positive shifts occurred in the 2000s and 2010s in particular countries.
  • South America and Africa continues to bear the brunt of deforestation.

Challenges faced

  • Lack of geographical data in the dataset.

  • Handling NA data and countries with no data.

  • Getting the legend right!

  • Rendering issue due to too much ink being used on the plot.

Question 2

How has the consumption of Soybean in Brazil changed over time, and how does it impact the afforestation or deforestation rates?

  • Our central question revolves around the historical evolution of soybean consumption and its potential implications for afforestation and deforestation rates in this vital agricultural region.

  • Our visual representation of this data employs the versatility of ggplot, particularly using geom_line() and geom_point() methods to construct time series plots.

  • These plots provide a dynamic illustration of the trends in soybean production in Brazil, shedding light on the growth and fluctuations in this vital agricultural sector.

Approach

Pre-processing of soybean and forest data 
#Function to pre-process the total_forest, soybean_use and forest_area datasets
#Input : total_forest- tibble
#        soybean_use- tibble
#        forest_area- tibble
#Output: soybean_brazil- tibble
#        forest_brazil- tibble


#Cleaning total_forest table
total_forest_cleaned <- clean_names(total_forest)

#Making a new column to calculate the total soybean consumption
soybean <- soybean_use |>
  mutate(total = human_food + animal_feed + processed)
#Some countries do not have consumption, and shows as 0. 
#Removing the rows if total=0
soybean <- subset(soybean, total != 0)

# Filter data for Brazil
soybean_brazil <- soybean |>
  filter(entity == "Brazil", year>= 1990&year<=2013)

# Filter data for Brazil forest: 
forest_brazil <- forest_area |>
  filter(entity == "Brazil",year>=1990&year<=2013)

#Finding total forest coverage per year
total_forest_world <- total_forest_cleaned |>
  filter(year >= 1990, year <= 2013, entity == "World") |>
  group_by(year)

# Left join to add total world forest coverage to the forest_brazil dataset
forest_brazil <- forest_brazil |>
  left_join(total_forest_world, by = "year")
 
#Finding actual total coverage for Brazil (percentage * total)
forest_brazil <- forest_brazil|>
  mutate(forest_area_brazil = forest_area.x * forest_area.y / 100)
Plotting of soybean consumption in Brazil 
#Code for creating the time series plot
#Used the soybean_brazil dataset created earlier as a data source
#using year and total as x and y axis for first layer
#Plotting points over line to increase visibility as second layer
#Manual fill to show trend as positive
#Input: year and total- numeric
#Output: plot_soybean_brazil- plot object


# Create a line plot for Brazil soybean consumption
plot_soybean_brazil <- ggplot(soybean_brazil, aes(x = year, y = total, color = "Brazil")) +
  geom_line(linewidth = 2) +    #Plotting line plot of series
  geom_point(color = "#6E8B3D") +  #Plotting points for clarity
  labs(x = "\nYear", 
       y = "Total (in lb)\n", 
       title = "Soybean consumption in Brazil\n", 
       caption = "Jon Harmon | TidyTuesday") +
  theme_minimal() +
  theme(legend.position = "none", plot.title = element_text(size = 15)) +
  scale_y_continuous(labels = scales::label_number(scale = 1e-06, suffix = "M")) + #Cleaning long numbers
  scale_color_manual(values = c("Brazil" = "#a6d96a")) +
  scale_x_continuous(limits = c(1990, 2013), breaks = seq(1990, 2013, by = 2))    #Defining year range

#Saving plot to location, and defining custom width
ggsave(plot_soybean_brazil, 
       filename = "images/q2/plot_soybean_brazil.jpg", 
       height   = 8, 
       width    = 15, 
       unit     = "in", 
       dpi      = 120)
Plotting of forest coverage in Brazil 
#Code for creating the time series plot
#Used the forest_brazil dataset created earlier as a data source
#using year and total as x and y axis for first layer for line plot
#Plotting points over line to increase visibility as second layer
#Manual fill to show trend as negative
#Input: year and forest_area_brazil- numeric
#Output: plot_soybean_brazil- plot object


# Create a line plot for Brazil with points
plot_forest_brazil <- ggplot(forest_brazil, aes(x = year, y = forest_area_brazil, color = "Brazil")) +
  geom_line(linewidth = 2) +      #Plotting line plot of series
  geom_point(color="#fdae61") +   #Plotting points for clarity
  labs(x = "\nYear", 
       y = "Forest coverage (in hectares)\n", 
       title = "Forest coverage in Brazil\n", 
       caption= "Jon Harmon | TidyTuesday") +
  theme_minimal() +
  theme(legend.position = "none", plot.title = element_text(size = 15)) +
  scale_color_manual(values = c("Brazil" = "#fee08b"))+
  scale_x_continuous(limits = c(1990, 2013), breaks = seq(1990, 2013, by = 2))+    #Defining year range
  scale_y_continuous(labels = scales::label_number(scale = 1e-6, suffix = "M"))    #Cleaning long numbers

ggsave(plot_forest_brazil, 
       filename = "images/q2/plot_forest_brazil.jpg", 
       height   = 8, 
       width    = 15, 
       unit     = "in", 
       dpi      = 120)

Outcome

  • Animation of soybean usage

Outcome

  • Animation of forest coverage

Analysis

  • The visualizations provide a clear depiction of the steady increase in soybean consumption in Brazil.

  • The data shows a remarkable increase, from approximately 16.4 million pounds of soybeans in 1990 to a staggering 36.87 million pounds in 2013.

  • The forest coverage in Brazil dropped from 588 million hectares in 1990 to 507 million hectares in 2013, representing a significant loss of 81 million hectares of forest land during this time.

  • This reduction in forest area is indicative of the environmental impact in Brazil.

  • The correlation between rising soybean consumption and decreasing forest coverage in Brazil underscores the need for sustainable agricultural practices and conservation efforts.

Challenges faced

  • Lack of total forest coverage for the world.

  • Animation and frame rate selection.

  • Error in data type and rendering method selection of gif during animation.

I am Groot!?

Thank you!

Any Questions?

Groot Source: https://gifer.com

🐼 The Plotting Pandas