Calculate and Compare Carbon Footprints of Different Flights

Navigate to the Bureau of Transportation Statistics (transtats.bts.gov) and download the T-100 Domestic Segment dataset for a recent year. Select columns including: carrier, origin, destination, aircraft type, departures, seats, passengers, distance, and — critically — fuel consumed (gallons).

Save the data as a CSV file. You'll have roughly 250,000 rows covering every domestic airline route. This is the same data airlines report to the federal government every quarter.

Load the CSV into a pandas DataFrame. Filter out rows with zero passengers or zero fuel consumption (ferry flights, repositioning moves). Create new columns: fuel_per_pax (gallons per passenger) and co2_per_pax_km using the standard conversion factor of 21.1 pounds of CO2 per gallon of jet fuel (Jet-A).

Use df.describe() and df.groupby('AIRCRAFT_TYPE').mean() to get a feel for the data. Which aircraft types appear most often? What's the range of distances?

Use matplotlib to create a horizontal bar chart of average CO2 per passenger-km for the 15 most common aircraft types. Color-code by aircraft category (regional, narrow-body, wide-body). You'll see a clear pattern: larger aircraft on longer routes are significantly more fuel-efficient per passenger.

Create a second plot: a scatter chart of route distance vs. CO2 per passenger-km, with each dot colored by aircraft size category. This reveals the "short-haul penalty" — short flights burn disproportionately more fuel per km because takeoff and climb dominate.

Pick 3–4 popular city pairs (e.g., LAX–SFO, JFK–ORD, ATL–DFW) and compare the carbon footprint across airlines operating those routes. Do airlines using newer aircraft (A320neo vs. older 737s) show measurably lower emissions? Create a grouped bar chart showing the comparison.

Calculate the total annual CO2 for each route by multiplying per-flight emissions by departure frequency. Which routes contribute the most total emissions? Are they long-haul high-frequency routes or short-haul shuttles?

Use scikit-learn's LinearRegression to predict CO2 per passenger-km from two features: route distance and aircraft seat capacity. Split your data 80/20 into training and test sets. Fit the model and check the R-squared score — how well do just two variables explain emission intensity?

Plot the predicted vs. actual values. Where does the model fail? Flights with very low load factors (few passengers relative to capacity) are hard to predict because the model doesn't know how full the plane is.

Add more features to your model: load factor (passengers / seats), number of stops, and carrier (as a one-hot encoded variable). Retrain and compare R-squared. Load factor should dramatically improve predictions because a half-empty plane has roughly double the per-passenger emissions.

Try a polynomial regression (degree 2) to capture the non-linear relationship between distance and efficiency. Does it improve the fit?

Compile your findings into a clear report with 4–5 key visualizations. Answer the questions: Which aircraft types are cleanest? How much does route length matter? How much does load factor matter? What's the single biggest lever for reducing aviation's carbon footprint?

Include your model's performance metrics and discuss its limitations. A strong conclusion might note that the biggest gains come from filling seats and replacing older aircraft — not from choosing different routes.