Grégoire Maillard

How it works

US domestic, demand: passenger counts come from the US Department of Transportation's Domestic Airline Consumer Airfare Report (Table 6), which is built from the DB1B survey, a 10% sample of all airline tickets. It measures true origin and destination: someone flying Boston to Tucson via Dallas counts once for Boston–Tucson and not at all for Boston–Dallas. The four quarters from Q2 2025 to Q1 2026 are averaged, so seasonal markets aren't over- or under-counted. Only the 48 contiguous states are covered.

Cities, not airports: the DOT groups airports serving the same area into one city market (New York = JFK, LaGuardia, Newark, White Plains and Islip; Los Angeles includes Orange County, Burbank, Ontario and Long Beach; Miami includes Fort Lauderdale). A market counts as having a nonstop if any airport on one side has one to any airport on the other. In the Canadian views, Toronto combines Pearson and Billy Bishop, London combines Heathrow and Gatwick, and Tokyo combines Haneda and Narita.

Nonstop schedules: scheduled nonstop routes come from the open airline-route-data project, which collects published schedules from FlightsFrom. Nonstop means a route was scheduled in September 2026. Seasonal or dropped means it was scheduled at some point in March–May 2026 but not in September. No nonstop means none was found at any point in those snapshots. Schedule scrapes can miss a route, so treat a single pair with some care.

US – Europe and US – Asia: the DOT does collect international itineraries, but keeps them confidential, and foreign airlines don't report them at all, so there is no public origin–destination data for these markets. These two views use T-100 International Segment data instead: every passenger carried on a scheduled nonstop flight between a US airport and a foreign one, by any airline, from July 2025 to June 2026. That counts people who connect: a Denver passenger flying to Rome through Frankfurt shows up only on Denver–Frankfurt. So these maps show where the nonstop flights go and how full they are, not where people are ultimately travelling. Routes with fewer than 20,000 passengers a year are left out. "Asia" includes the Middle East; Turkey is counted with Europe; Guam and the Northern Mariana Islands are left out.

Canada domestic: Canada doesn't publish passengers by route. Statistics Canada's origin–destination tables stop in 2018 and never covered domestic city pairs publicly. The best public figure is Transport Canada's top 25 domestic markets, measured in scheduled seats per day in each direction for 2024 (from OAG), with each airline's share. Those 25 are drawn by size. Every other nonstop route between Canadian airports with at least 150,000 passengers a year (Statistics Canada) is drawn as a thin line from the September 2026 schedules, without a volume.

Canada – Europe: passenger counts are reported by the European end of each route: Eurostat route data for EU countries, Switzerland, Iceland and Turkey, and the UK Civil Aviation Authority for British airports. Both count every passenger on nonstop flights in 2024, the latest full year they publish, so like the US international views they show nonstop traffic rather than true origin and destination. Only the main airports report route data, and routes under 20,000 passengers are left out. Airlines are taken from the September 2026 schedules; a dashed grey line means the route carried passengers in 2024 but isn't scheduled any more.

Canada – Asia: no public source gives passengers on these routes: Canada doesn't publish them, and the Asian and Middle East countries at the other end mostly don't either. This view shows the nonstop network from the September 2026 schedules, with the airlines on each route, but no volumes.

Latin America: Mexico, Central and South America, Cuba, the Dominican Republic and Haiti. French Guiana, Martinique and Guadeloupe are part of France, so flights to Paris count as French domestic and are left out. The Europe view uses the same Eurostat and UK figures as Canada – Europe (2024 passengers, reported by the European airports). The Asia view shows the nonstop network from the September 2026 schedules, with the airlines on each route but no volumes. Schedules include a few "tag" flights that stop on the way, such as Beijing–Tijuana–Mexico City, which appear as two routes.

Canada ↔ US: Statistics Canada's air passenger origin and destination survey counts where travellers start and finish their trip between Canada and the US, including those who connect, like the US domestic view. 2018 is the last year published by city pair, so the volumes are pre-pandemic; the nonstop check uses the 2026 schedules, as in the US view.

Estimated opportunities (US – Europe, US – Asia, Canada – Europe, Latin America – Europe): for international routes only nonstop traffic is public, so a city pair with no nonstop has no number at all. To fill the gap, a gravity model is fitted on the existing routes of each view: passengers on a route are explained by the total traffic of each city and the distance between them (log-linear regression; R² between 0.5 and 0.65). The model is then applied to every pair of cities with at least two routes in the data and no nonstop in the 2026 schedules, and the 20 largest are shown as dashed orange lines. Treat them as a ranking of where demand is likely to be, not a forecast: the model knows nothing about fares, airport slots, bilateral agreements, aircraft range or competition from connecting hubs. Canada domestic and the Asia views without volumes have no estimates, because there's too little data to fit a model.

This is a personal project and isn't affiliated with any airline.