# Read large parquet files efficiently
library("arrow")
# Convert between country codes and standardized country names
library("countrycode")
# Provide world map data
library("maps")
# Convert ggplot outputs into interactive plots
library("plotly")
# Format numbers (commas, percent labels) for axes and legends
library("scales")
# Render interactive tables
library("DT")
# Load core tidyverse packages (dplyr, ggplot2, tidyr, stringr, etc.)
library("tidyverse")# Load flight schedule data and airport data
flight_df <- read_parquet("../data/2025_Q4.parquet")
airports_df <- read.csv("../data/airports.csv")
flihgt_rows <- nrow(flight_df)
airports_rows <- nrow(airports_df)# Clear out the flights without "Airline"
flight_df <-
flight_df |>
filter(Airline != "-")
# Pick out the ICAO code for later joining (first code in list)
flight_df <-
flight_df |>
mutate(
origin_code_first = Track_Origin_ApplicableAirports |>
str_replace_all("\\[|\\]|'|\\s", "") |>
str_replace("^-$", NA_character_) |>
str_extract("^[^,]+"),
dest_code_first = Track_Destination_ApplicableAirports |>
str_replace_all("\\[|\\]|'|\\s", "") |>
str_replace("^-$", NA_character_) |>
str_extract("^[^,]+")
) |>
select(
origin_code_first, dest_code_first,
Callsign, Airline,
Track_Origin_DateTime_UTC, Track_Destination_DateTime_UTC
)
# Airports: keep medium/large + valid ICAO + valid lat/lng
airports_df <-
airports_df |>
filter(icao_code != "") |>
filter(type %in% c("large_airport", "medium_airport")) |>
filter(!is.na(latitude_deg), !is.na(longitude_deg)) |>
select(
icao_code,
name,
continent,
iso_country,
latitude_deg,
longitude_deg
)
# Data Join (flight + airport)
flights_airports_df <-
flight_df |>
inner_join(
airports_df |>
rename(
origin_airport = name,
origin_continent = continent,
origin_country = iso_country,
origin_latitude = latitude_deg,
origin_longitude = longitude_deg
),
join_by(origin_code_first == icao_code)
) |>
inner_join(
airports_df |>
rename(
dest_airport = name,
dest_continent = continent,
dest_country = iso_country,
dest_latitude = latitude_deg,
dest_longitude = longitude_deg
),
join_by(dest_code_first == icao_code)
)
# International flight (exclude domestic ones)
international_flights_df <-
flights_airports_df |>
filter(origin_country != dest_country)This dashboard examines the global aviation network in the fourth quarter of 2025 using flight schedule records and airport infrastructure data. Drawing on background from the IATA (International Air Transport Association), it focuses on how air traffic is concentrated across countries, how strongly some national systems depend on major hubs, and how U.S. domestic and international patterns differ.
Figure 1 provides historical background on the structure of the global airline network before the pandemic; it is not part of the 2025 dataset analyzed in this dashboard.
Figure 1. Historical world airline route map (2009). Jpatokal, CC BY-SA 3.0 https://creativecommons.org/licenses/by-sa/3.0, via Wikimedia Commons.
1) Global Flight Concentration: How concentrated is global air traffic in late 2025, and which countries occupy the most central positions in international flight activity?
2) Airport Siphoning by Country: To what extent do a few major airports dominate a country’s total flight activity, and which countries show single-hub versus multi-hub structures?
3) U.S. Domestic vs. International Flights: How do domestic and international flight patterns differ in the United States, and which places function as its main gateways?
We use two public datasets: MrAirspace for dynamic flight activity and OurAirports for geographic airport reference data.
This analysis uses public, non-sensitive aviation data, but uneven ADS-B coverage may undercount flights in remote or lower-infrastructure regions. The dashboard measures flight activity rather than passenger demand, cargo volume, or route profitability.
- Source: https://github.com/MrAirspace/aircraft-flight-schedules
- Access and publicity: This dataset is hosted on GitHub and maintained by MrAirspace, making it publicly available scientific research data.
- Creator: The dataset was compiled and maintained by GitHub contributor MrAirspace. Its core raw signals originate from a globally distributed network of ADS-B receivers.
- Sample description: This dataset captures global civil aviation flight activity in the fourth quarter of 2025. It records the callsign, airline, and ICAO codes of scheduled departure and arrival airports for each aircraft. Because it reflects observed movement rather than only published schedules, it is useful for studying concentration and connectivity in the global network. It therefore provides the dynamic flight activity used throughout this project.
- Measurement objects and methods: The measured units are aircraft in operation. Flight activity is recorded through ADS-B broadcasts, which are collected by ground receivers and converted into structured flight records.
- Record size: In this study, the cleaned original flight schedule for Q4 2025 contains approximately 13542708 unique flight records.
- Temporal and geographic context: The data covers October to December 2025 (Q4). Theoretically, the geographical scope covers the entire globe, but due to limitations in the density of ground receiving stations, records for North America, Europe, and East Asia are the most detailed.
- Sample evaluation: Advantage: This sample reflects real air traffic flow rather than only published schedules, making it well suited for studying actual connectivity patterns. Limitation: Because it depends on physical receiver coverage, some regions such as the high seas and parts of sub-Saharan Africa may be underrepresented.
- Source: https://ourairports.com/help/data-dictionary.html
- Access and publicity: This dataset can be downloaded free of charge from the OurAirports website and is a completely open and free aviation geographic database.
- Creator: This project is operated by OurAirports.com, a community-driven organization comprising aviation enthusiasts, pilots, and Geographic Information Systems (GIS) experts from around the globe.
- Sample description: This dataset serves as a static reference for global aviation infrastructure. It includes airport identifiers, official names, geographic coordinates, and airport classifications. In this project, it provides the spatial framework needed to map routes, compare countries, and identify major hubs. It complements the flight schedule data with consistent geographic metadata.
- Measurement objects and methods: The recorded units are airport facilities around the world. The data is compiled from Aeronautical Information Publications (AIPs) and maintained through community updates and verification.
- Record size: The database contains detailed records for 84554 facilities globally, ensuring comprehensive analytical coverage.
- Temporal and geographic context: The data reflects the status of global aviation infrastructure during the 2024–2025 period, covering virtually all known civil landing sites with the exception of a few restricted military zones.
- Sample evaluation: Advantage: It offers broad geographic coverage and the airport metadata needed to connect flight movements to countries and regions. Limitation: Because it is partly crowdsourced, updates for smaller or less active facilities may lag behind current conditions.
### Build country-level flight totals (departures + arrivals)
inter_flight_by_country_df <-
international_flights_df |>
group_by(origin_country) |>
summarize(departure_counts = n()) |>
full_join(
international_flights_df |>
group_by(dest_country) |>
summarize(arrival_counts = n()),
join_by(origin_country == dest_country)
) |>
rename(country_code = origin_country) |>
complete(country_code, fill = list(departure_counts = 0L, arrival_counts = 0L)) |>
mutate(
total_flights = departure_counts + arrival_counts
)
# Convert country codes to country names so they match the world map labels
inter_flight_by_country_df <-
inter_flight_by_country_df |>
mutate(map_country_name = countrycode(country_code, "iso2c", "country.name"))
# Roll small territories into parent countries for cleaner visualization, then re-aggregate
inter_flight_by_country_df <-
inter_flight_by_country_df |>
mutate(
map_country_name = countrycode(country_code, "iso2c", "country.name"),
map_country_name = if_else(country_code %in% c("HK", "MO"), "China", map_country_name),
map_country_name = if_else(country_code %in% c("PR", "VI"), "United States", map_country_name),
map_country_name = if_else(country_code == "TC", "United Kingdom", map_country_name),
map_country_name = if_else(map_country_name == "U.S. Virgin Islands", "United States", map_country_name),
map_country_name = if_else(map_country_name == "Gibraltar", "United Kingdom", map_country_name)
) |>
group_by(map_country_name) |>
summarize(
departure_counts = sum(departure_counts),
arrival_counts = sum(arrival_counts),
total_flights = sum(total_flights)
)
# Load the world map polygons and standardize a few country labels
world_shape <- map_data("world")
world_shape <-
world_shape |>
mutate(
region = if_else(region == "USA", "United States", region),
region = if_else(region == "UK", "United Kingdom", region)
)
# Join flight totals onto the map polygons and draw the choropleth
country_shape_flight <-
world_shape |>
left_join(inter_flight_by_country_df, join_by(region == map_country_name)) |>
mutate(
# Create hover labels for the interactive world map
hover_country = region,
hover_total = coalesce(total_flights, 0),
hover_departures = coalesce(departure_counts, 0),
hover_arrivals = coalesce(arrival_counts, 0),
hover_text = paste0(
"Country: ", hover_country,
"<br>Total international flights: ", comma(hover_total),
"<br>Departures: ", comma(hover_departures),
"<br>Arrivals: ", comma(hover_arrivals)
)
)
global_flight_distribution_plot <-
ggplot(country_shape_flight) +
geom_polygon(
aes(
x = long,
y = lat,
group = group,
fill = total_flights,
text = hover_text
)
) +
scale_fill_viridis_c(
option = "plasma",
na.value = "grey90",
name = "Total Flights",
labels = comma
) +
coord_quickmap() +
theme_minimal() +
labs(
title = "Global Air Connectivity by Country (2025 Q4)",
subtitle = "Based on international flight counts"
)
### Sort countries by total flights in descending order and calculate cumulative share
top_n <- 20
top_flight_country_df <-
inter_flight_by_country_df |>
filter(!is.na(total_flights)) |>
arrange(desc(total_flights)) |>
mutate(
rank = row_number(),
share = total_flights / sum(total_flights),
cumulative_share = cumsum(share)
) |>
head(top_n)
# Total flights (bar)
global_total_bar_plot <-
ggplot(
top_flight_country_df,
aes(
x = reorder(map_country_name, -total_flights),
y = total_flights,
# Create hover labels for the interactive country ranking chart
text = paste0(
"Country: ", map_country_name,
"<br>Total flights: ", comma(total_flights),
"<br>Share of total: ", percent(share, accuracy = 0.1)
)
)
) +
geom_col(fill = "#440154") +
scale_y_continuous(labels = comma) +
theme_minimal() +
theme(
axis.text.x = element_text(angle = 45, hjust = 1)
) +
labs(
title = paste0("Top ", top_n, " Countries by Total Flights"),
subtitle = "International flight volume (2025 Q4)",
x = "Country",
y = "Total Flights"
)
# Turn the global map into an interactive plot
global_flight_distribution_interactive <- ggplotly(
global_flight_distribution_plot,
tooltip = "text"
)
# Turn the top-country bar chart into an interactive plot
global_total_bar_interactive <- ggplotly(
global_total_bar_plot,
tooltip = "text"
)
# Cumulative share (line)
global_cum_share_line_plot <-
ggplot(top_flight_country_df, aes(x = rank, y = cumulative_share)) +
geom_line(color = "#E6C619", linewidth = 1) +
geom_point(color = "#E6C619", size = 1.8) +
scale_y_continuous(labels = percent, limits = c(0, 1)) +
scale_x_continuous(
breaks = top_flight_country_df$rank,
labels = top_flight_country_df$map_country_name,
expand = expansion(mult = c(0.01, 0.01))
) +
theme_minimal() +
theme(
axis.text.x = element_text(angle = 45, hjust = 1)
) +
labs(
title = paste0("Cumulative Share (Top ", top_n, " Countries)"),
subtitle = "How quickly global traffic accumulates among top-ranked countries",
x = "Country (ranked by total flights)",
y = "Cumulative Share"
)### Global Hub Dominance
# Arrival data
global_country_airports_arr_count <-
flights_airports_df |>
group_by(dest_country, dest_airport) |>
summarise(arr_total = n())
# Departure data
global_country_airports_dep_count <-
flights_airports_df |>
group_by(origin_country, origin_airport) |>
summarise(dep_total = n())
# Arr + Dep = total
global_country_airports_count <-
global_country_airports_arr_count |>
full_join(global_country_airports_dep_count,
join_by(
dest_country == origin_country,
dest_airport == origin_airport
)) |>
rename(country_code = dest_country, airport = dest_airport) |>
mutate(
arr_total = if_else(is.na(arr_total), 0, arr_total),
dep_total = if_else(is.na(dep_total), 0, dep_total),
total = arr_total + dep_total
)
# Calculate the fight total in each country
# Calculate the share and rank for each airport in its country
global_country_airports_rank <-
global_country_airports_count |>
group_by(country_code) |>
mutate(
country_total = sum(total),
airport_share = total / country_total
) |>
arrange(desc(total)) |>
mutate(rank = row_number())
# We check the share of top 1 and top 1-3 airport for each country
global_country_hub_data <-
global_country_airports_rank |>
filter(country_total >= 10000) |>
group_by(country_code) |>
summarise(
top_1_share = airport_share[1],
top_3_share = sum(airport_share[1:3], na.rm = TRUE),
flight_total = country_total[1]
)
# Use the more formal country name
global_country_hub_data <-
global_country_hub_data |>
mutate(country_name = countrycode(country_code, "iso2c", "country.name"))
hub_datatable <-
global_country_hub_data |>
mutate(
top_1_share = round(top_1_share * 100,1),
top_3_share = round(top_3_share * 100,1)
) |>
select(
country_name,
flight_total,
top_1_share,
top_3_share
) |>
arrange(desc(flight_total))
global_country_hub_top20 <-
global_country_hub_data |>
arrange(desc(top_1_share)) |>
slice(1:20)
# Top 20 Hub
global_top_hub_country_plot <-
ggplot(global_country_hub_top20, aes(x = reorder(country_name, top_1_share), y = top_1_share)) +
geom_col(fill = "#2EC4B6") +
coord_flip() +
scale_y_continuous(
labels = percent_format(accuracy = 1),
limits = c(0, 1),
expand = expansion(mult = c(0, 0.05))
) +
theme_minimal() +
theme(panel.grid.minor = element_blank()) +
labs(
title = "Hub Dominance: Top 1 Airport Share (Top 20 Country/Region)",
subtitle = "Must have at least 10,000 flights",
x = NULL,
y = "Top 1 airport share"
)
# This time, pick the top 15 big total flight country
global_top15_country_hub <-
global_country_hub_data |>
arrange(desc(flight_total)) |>
slice(1:15)
global_top_country_hub_plot <-
ggplot(global_top15_country_hub, aes(x = reorder(country_name, top_3_share), y = top_3_share)) +
geom_col(fill = "#2EC4B6") +
coord_flip() +
scale_y_continuous(
labels = percent_format(accuracy = 1),
limits = c(0, 1),
expand = expansion(mult = c(0, 0.05))
) +
theme_minimal() +
theme(panel.grid.minor = element_blank()) +
labs(
title = "Hub Dominance: Top 3 Airports’ Share",
subtitle = "Top 15 by total flights (>= 10,000)",
x = NULL,
y = "Top 3 airport share"
)# Statistics on the number of international flights from US to various countries
outbound_df <- flights_airports_df |>
# The departure point is US, destination is not US
filter(origin_country == "US", dest_country != "US") |>
group_by(dest_country, dest_airport, dest_latitude, dest_longitude) |>
summarise(flight_count = n(), .groups = "drop")
inbound_df <- flights_airports_df |>
# The departure point is not US, destination is US
filter(origin_country != "US", dest_country == "US") |>
# Summarized by country and airport of origin
group_by(origin_country, origin_airport, origin_latitude, origin_longitude) |>
summarise(flight_count = n(), .groups = "drop")
# Statistics on the number of flights from US to US
domestic_df <-
flights_airports_df |>
filter(origin_country == "US", dest_country == "US")
domestic_map_df <- bind_rows(
domestic_df |>
transmute(airport = origin_airport, lat = origin_latitude, lon = origin_longitude),
domestic_df |>
transmute(airport = dest_airport, lat = dest_latitude, lon = dest_longitude)
) |>
count(airport, lat, lon, name = "flight_count")
# Total international exchanges(Outbound + Inbound)
intl_summary <- outbound_df |>
group_by(dest_country) |>
summarize(out_n = sum(flight_count)) |>
full_join(
inbound_df |>
group_by(origin_country) |>
summarize(in_n = sum(flight_count)),
by = c("dest_country" = "origin_country")
) |>
mutate(
out_n = coalesce(out_n, 0),
in_n = coalesce(in_n, 0),
total_intl = out_n + in_n
) |>
rename(country_code = dest_country) |>
mutate(country_name = countrycode(country_code, "iso2c", "country.name"))
# Top 10 selected
intl_top_countries <- intl_summary |>
slice_max(total_intl, n = 10) |>
# Sort by quantity
mutate(country_name = reorder(country_name, total_intl))
# Match airport coordinates to specific US states
states_map <- map_data("state")
domestic_state_data <- domestic_map_df |>
mutate(state_name = map.where("state", lon, lat)) |>
# clean data
mutate(state_name = gsub(":.*", "", state_name)) |>
filter(!is.na(state_name)) |>
group_by(state_name) |>
summarise(dom_flight_n = sum(flight_count), .groups = "drop")
# US State map Data
usa_state_map_data <- states_map |>
left_join(domestic_state_data, by = c("region" = "state_name"))
# Chart 1: Top 10 International Flights by Total Round Trip Volume
us_top_inter_plot <-
ggplot(
intl_top_countries,
aes(
x = country_name,
y = total_intl,
fill = total_intl,
# Create hover labels for the interactive U.S. international chart
text = paste0(
"Country: ", as.character(country_name),
"<br>Total international flights: ", comma(total_intl),
"<br>Outbound from U.S.: ", comma(out_n),
"<br>Inbound to U.S.: ", comma(in_n)
)
)
) +
geom_col(width = 0.7) +
scale_fill_gradient(low = "#74add1", high = "#313695", name = "total flights") +
# Turn the coordinate axes horizontally
coord_flip() +
geom_text(aes(label = comma(total_intl)), hjust = -0.1, size = 3.5) +
scale_y_continuous(labels = comma, expand = expansion(mult = c(0, 0.15))) +
theme_minimal() +
labs(
title = "Top 10 Countries by International Flight Volume to/from U.S.",
subtitle = "Combined inbound and outbound flight counts",
x = "Country Name",
y = "Total Flights"
) +
theme(
panel.grid.minor = element_blank(),
plot.title = element_text(face = "bold", size = 14),
axis.title = element_text(face = "italic")
)
# Chart 2: Domestic Flight Density Map in the United States
us_top_domes_plot <-
ggplot(usa_state_map_data) +
# Domestic flights within the United States
geom_polygon(
aes(
x = long,
y = lat,
group = group,
fill = dom_flight_n,
# Create hover labels for the interactive U.S. state map
text = paste0(
"State: ", str_to_title(region),
"<br>Domestic flights: ", comma(coalesce(dom_flight_n, 0)),
"<br>Measure: departures + arrivals aggregated from airport locations"
)
),
color = "white", linewidth = 0.2
) +
scale_fill_gradient(low = "#F3E5F5", high = "#4A148C",
name = "Domestic flights distribution", labels = label_comma(),
na.value = "grey90") +
# Settings
coord_fixed(1.3) +
theme_minimal() +
labs(
title = "Domestic Flight Activity by State in the U.S.",
subtitle = "Departures + arrivals aggregated from airport locations",
x = NULL, y = NULL
) +
theme(
legend.position = "right",
panel.grid = element_blank(),
axis.text = element_blank(),
plot.title = element_text(hjust = 0.5, face = "bold", size = 14),
plot.subtitle = element_text(hjust = 0.5)
)
# --- Chart 3: U.S. Domestic vs International Flight Volume (NOT "market") ---
us_domestic_n <- nrow(domestic_df)
us_international_n <- sum(outbound_df$flight_count) + sum(inbound_df$flight_count)
us_dom_vs_int_df <- data.frame(
flight_type = c("Domestic", "International"),
flight_count = c(us_domestic_n, us_international_n)
)
us_dom_vs_int_plot <-
ggplot(us_dom_vs_int_df, aes(x = reorder(flight_type, flight_count), y = flight_count)) +
geom_col(width = 0.7, fill = "#313695") +
coord_flip() +
geom_text(aes(label = comma(flight_count)), hjust = -0.1, size = 3.5) +
scale_y_continuous(labels = comma, expand = expansion(mult = c(0, 0.15))) +
theme_minimal() +
labs(
title = "U.S. Domestic vs International Flight Volume",
subtitle = "International = inbound + outbound flights involving the U.S.",
x = NULL,
y = "Flight Count"
) +
theme(
panel.grid.minor = element_blank(),
plot.title = element_text(face = "bold", size = 14),
axis.title = element_text(face = "italic")
)
# Display Interactive Chart
# Turn the U.S. international bar chart into an interactive plot
us_top_inter_interactive <- ggplotly(us_top_inter_plot, tooltip = "text")
# Turn the U.S. domestic state map into an interactive plot
us_top_domes_interactive <- ggplotly(us_top_domes_plot, tooltip = "text")- Global international flight activity is highly concentrated: a small set of countries accounts for a disproportionately large share of total flights.
- The concentration is visible both geographically (dense clusters in a few regions) and quantitatively (the cumulative share rises quickly among top-ranked countries).
- Because this section focuses on international flights only, domestic aviation patterns are analyzed separately (especially in the U.S. section).
- These results describe connectivity patterns in the dataset (scheduled/recorded flights), not passenger or cargo volume.
- Hub dominance varies widely: some countries are strongly hub-dominated, where one airport captures most national flight activity, while others distribute traffic across multiple hubs.
- Top 1 share highlights “single-hub” systems; Top 3 share distinguishes multi-hub systems from broadly distributed networks.
- Among high-volume countries, some remain highly concentrated (Top 3 share stays high), while others (e.g., large multi-center networks) show much lower concentration.
- The table provides exact values for flight totals and concentration metrics (Top 1 / Top 3 shares), enabling quick country-level comparisons.
- U.S. domestic flight volume is much larger than international flight volume (international = inbound + outbound involving the U.S.).
- International connections are concentrated among a small set of partner countries, with Canada and Mexico dominating the top ranks.
- Domestic flight activity is uneven across states, with higher intensity in major population and hub states (departures + arrivals aggregated from airport locations).
- This section measures flight counts/activity (not passenger demand or cargo trade); some territories may appear separately depending on the country-code definition.
This section list out the tools/method used in our project and explains their role in the analysis.
Missing Data Handling
We handled missing values by cleaning inconsistent entries and replacing them in ways that kept the dataset usable for analysis. This reduced distortion in our summaries and made our comparisons more complete and reliable.
Methods: str_replace(), coalesce(), if_else(), complete(), sum(na.rm = TRUE)
Conditional Transformation
We used conditional transformations to recode and standardize values when different labels referred to the same broader category. This made the data more consistent and allowed us to group and compare observations more meaningfully.
Methods: if_else()
Maps
We used map-based visualizations to display geographic patterns in the data and make spatial differences easier to interpret. This helped us communicate location-based trends more clearly than a non-spatial chart would.
Methods: map_data(), map.where(), geom_polygon()
Strings
We used string-processing tools to clean, extract, and standardize text values before analysis. This made the raw data easier to organize and ensured that similar entries were treated consistently.
Methods: str_replace(), str_replace_all(), str_extract(), gsub()
This section introduces additional tools that were not part of our core class toolkit but appeared in our project. We briefly explain what each one helped us do in our workflow and note where we learned or encountered it.
arrow + read_parquet()
We used arrow and read_parquet() to load the flight dataset from a Parquet file format. Source: We learned about this from the official Arrow R documentation while learning how to import Parquet files in R.
rename()
We used rename() to make variable names clearer and more consistent after joining or transforming data. This helped us keep track of similar fields and made later steps in the analysis easier to read.
Source: We learned about this from the official dplyr documentation while using it to distinguish similarly named fields after joins.
countrycode + countrycode()
We used countrycode() to convert ISO country codes into full country names so the data would be easier to interpret and match with map labels. This made the results more readable and more suitable for visualization.
Source: We learned about this from the official countrycode documentation while preparing country labels for maps and summaries.
cumsum()
We used cumsum() to calculate cumulative values across ranked observations. This helped us show how quickly the total share accumulated among the leading categories.
Source: We learned about this from the official R documentation while building the cumulative distribution for global flight concentration.
scale_x_continuous() and scale_y_continuous()
We used these functions to control how axes were displayed, including labels, limits, and spacing. This made the plots easier to read and helped present the values in a clearer format.
Source: We learned about this from the official ggplot2 documentation while refining chart axes and labels.
count()
We used count() as a quick way to summarize how often grouped observations appeared in the data. This simplified parts of the workflow that would otherwise require a separate group_by() and summarize().
Source: We learned about this from the official dplyr documentation while looking for a shorter grouped-counting workflow.
slice_max()
We used slice_max() to select the top observations based on a chosen variable. This made it easier to focus on the highest-volume categories without writing a longer filtering workflow.
Source: We learned about this from the official dplyr documentation while selecting the highest-ranked results for one of our plots.
geom_text()
We used geom_text() to place numeric labels directly on the charts. This made the visualizations easier to interpret because viewers could see exact values without estimating from the axis alone.
Source: We learned about this from the official ggplot2 documentation while adding direct labels to improve chart readability.
reorder()
We used reorder() to sort categorical labels based on numeric values rather than alphabetical order. This made the ranking patterns in the charts much easier to understand visually.
Source: We learned about this from the official R documentation while ordering categories by data values instead of alphabetically.
coord_quickmap(), coord_fixed(), and coord_flip()
We used these coordinate functions to control plot orientation and spatial proportions. Together, they helped us improve map accuracy, preserve visual balance, and make some bar charts more readable.
Source: We learned about this from the official ggplot2 documentation while adjusting maps and plot layouts.
Our analysis shows that global air connectivity in 2025 Q4 is highly concentrated rather than evenly distributed. International flight activity is led primarily by North America and Europe, with East Asia also playing an important but somewhat secondary role. By contrast, many countries in the Global South appear much less connected in this dataset, especially in Africa, where a large number of observations are missing or unavailable. This means the global pattern is not only concentrated, but also regionally uneven, with strong activity clustered in a relatively small number of countries and regions.
Hub dominance also varies substantially across countries, and the two hub-share plots highlight different kinds of systems. The countries that rank highest in top 1 airport share are almost entirely different from those that rank highest in top 3 airport share, showing that single-hub dominance and multi-hub concentration are not the same pattern. In other words, some countries depend heavily on one dominant gateway, while others distribute large shares of traffic across a small group of major airports. Looking at both measures gives a fuller picture of how national air networks are structured.
The U.S. case reinforces this broader pattern. Domestic flight volume is far larger than international flight volume, showing that the U.S. aviation system is driven primarily by internal connectivity. Its international traffic is also concentrated among a small set of partner countries, especially nearby countries such as Canada and Mexico, while domestic activity is unevenly distributed across states. At the same time, our results should be interpreted as estimates of flight activity rather than total transportation importance, since flight counts do not capture passenger volume, seat capacity, route distance, or cargo movement. Because the dataset only covers one quarter, it may also reflect seasonal patterns. Future work could extend this analysis by using full-year data and adding measures such as passenger counts, aircraft capacity, or route distance to better evaluate global and national air connectivity.
Harry Cheng (Major in Computer Science, minor in Informatics, University of Washington; harrist@uw.edu) is interested in web development and data analysis.
Maggie Zhou (The Information School, University of Washington; meiqiz2@uw.edu) is interested in AI research, exploring the algorithms behind AI and the thought processes behind AI research.