How the Turbulence Report is made

The atmosphere comes from NOAA's Global Forecast System (GFS), the US National Weather Service's global model, at 0.25° resolution (about 28 km). Its full archive is public. For every day in the period we used the six-hour forecasts of its four daily runs, valid at 00, 06, 12 and 18 UTC: 6,988 snapshots in all. Six-hour forecasts are used rather than analyses because they carry the storm fields the engine needs, and they are what a live forecast looks like.

Each snapshot was run through FlyBud's turbulence engine at eight heights, from about 5,000 ft to 39,000 ft. At cruise height the engine combines four clear-air diagnostics of the kind used in the aviation industry's own turbulence guidance, fitted against tens of thousands of pilot reports. A storm term from instability and radar reflectivity, a mountain-wave term from wind over steep terrain and a daytime-thermals term near the ground sit beside them, and the strongest wins. The result is on the eddy dissipation rate (EDR) scale airlines use: light from 0.10, moderate from 0.20.

The app also blends ECMWF and DWD forecasts and corrects with live pilot reports. The report uses GFS alone, because it is the one model with a public archive long enough, so its numbers are from a single model, consistently.

The routes are every nonstop airport pair in the OpenFlights route database, both directions merged. It is an older list, so it is a starting list, not a timetable: every route that reaches a published top list was checked against current schedules.

Each route is flown through each snapshot the way the app forecasts a real flight: 24 to 64 points along the great circle, each at the height an airliner would be at that point, climbing, cruising at a height set by the distance, and descending. Each point reads the turbulence at its height. A route's score is the share of all those points, across all snapshots, that were light or worse, and the share that were moderate or worse. The main cause is whichever term was strongest at the bumpy points.

Only routes between airports with at least three routes in the list, or routes we have checked as flown today, are ranked, so a seasonal strip with one flight cannot top a press list; every route is in the download. Airports are ranked by the average of their routes, weighted by how many airlines fly each, if they have at least 15. Countries are ranked by the air at cruise height (the 250 and 300 hPa levels, about FL340 and FL300) over their land, weighted by area, if they cover at least twelve grid cells; borders are Natural Earth's.

It is a model climatology, not a record of flights. It says how often the forecast would have painted the air bumpy, not what any aircraft actually felt. Models are better at the large-scale causes (the jet stream, mountain waves) than at the small ones (a single storm cell), and at light turbulence than at severe: the report does not rank severe turbulence at all, because no model forecasts it reliably.

A snapshot is treated as one flight on every route, at the moment of the snapshot, so each route is seen at four times of day. Real flights take their own tracks and altitudes and avoid the worst air on the day; the report flies the great circle at a standard height, so it describes the air on the route, not what the crew would choose.

The engine's skill is measured against pilot reports, which come overwhelmingly from North America. Elsewhere the same physics applies, but it has been checked less. One check matters for a ranking: whether the engine favours some kinds of terrain. Against 171,000 pilot reports from the winter of 2025–26, it painted the air bumpy over the Rockies about as often, relative to what pilots reported, as over flat ground, so mountain routes are not under-ranked. Over the US Southeast and the Gulf it painted bumps less often than pilots felt them, so routes there may rank somewhat low.

Why this differs from other rankings. Other published turbulence rankings put Denver, Jackson and Salt Lake City at the top of the US list, and routes in western China near the top of the world's. This one puts the US West Coast and Alaska first, and western China lower. In the US, pilot reports side with this ranking: over the winter of 2025–26, 89% of pilot reports over California, Oregon, Washington and Nevada described light-or-worse bumps and 46% moderate-or-worse, against 75% and 27% over the Rockies. Pilots report bumpy air more readily than smooth, so every share is high; the comparison is what counts. Over Asia there are far fewer pilot reports to check against, so the Asian ranking is the least tested part of this report.

What it is not is noise. Ranking every route separately on two halves of the period, alternate years against each other, gives a rank correlation of 0.99, and 22 of the top 25 routes are the same in both. The places that are bumpy are bumpy every year.