Measured, not guessed
Every noise abatement answer is measured from real departures, or clearly marked as an estimate. Every engine-out departure is built from terrain and real navdata. Here is how, and what the limits are.
By the numbers
NADP1 and NADP2
Acceleration at about 3,000 ft above the airport. Quieter for the towns right next to the runway.
Acceleration at about 1,000 ft above the airport. Less fuel, noise spread further out.
In a flight track the difference shows as when the speed starts rising, how steeply the aircraft climbs through 2,000 ft, and how fast it is at 3,200 ft (for a 737, NADP1 median 169 kt, NADP2 217 kt). The altitude to set is the aerodrome elevation plus that height.
- CollectEvery day of the window the global ADS-B archive (about 4 GB a day) is streamed, keeping jet airliner departures from about 4,000 airports.
- Find each departureFrom lift-off up to 5,500 ft, with height above the airport taken from GPS altitude referenced to the aircraft's own ground position.
- Read three signals per flightWhere the speed starts rising, the climb rate through 1,800–2,600 ft, and how much the climb drops above 3,000 ft; airspeed too where it is broadcast.
- Let the data set the boundaryAcceleration heights form two clear peaks (about 1,300 and 3,100 ft); the valley between them separates NADP1 from NADP2.
- Judge per airline, aircraft and runwayAn airline's answer is its typical flight, with the NADP1 share shown so mixed practice is visible. Runways and night windows are split out when they genuinely differ.
- Rescue poorly covered airportsWhere receivers only see aircraft from ~2,700 ft up, the call is read from the speed at 3,200 ft. The brief says when this applies.
- Fill the gaps honestlyAirports with no usable flights get an estimate from similar airports; an airline never seen at an airport gets a prediction from its habits elsewhere. Both are labelled.
How the EOSIDs are made
- Start at the departure endFor every runway end, the one-engine-out climb starts 35 ft over the end of the runway, at the runway's own surveyed elevation.
- Look at the terrain aheadAlong the extended centreline, a corridor that widens with distance is checked against a 30 m terrain model and public obstacle databases (towers, masts, wind turbines).
- Fly a generic aircraftThree classes: jet (A320/737 size), turboprop and light twin. Each climbs at the net engine-out gradient of its class, in the real segments: climb, level acceleration, final climb.
- Standard or SpecialIf the straight climb clears the terrain all the way, the EOSID is Standard: straight ahead to a hold. If not, it is Special: a turn is placed before the first conflict, on the side with less terrain, flown at 15° bank with a realistic radius, sometimes with a speed limit.
- Hold at a real fixThe hold is always a real waypoint or navaid from the navigation data, never an invented point: 5 NM legs, about 2 NM turns, at least 3,000 ft. The brief shows how you would enter it (direct, teardrop or parallel) from your track.
- Re-fly and auditThe whole route, including the hold, is flown again against the terrain. Routes that fail are flagged (weight limited, no valid route) instead of hidden.
- Check for terrain warningsA terrain-awareness model predicts whether a real aircraft would get a terrain caution or warning on the route, and the brief says so where it is expected.
- Say how sure it isEach EOSID carries a confidence (high, medium, low) and the reasons, visible on the details page.
How accurate is it
| Test | Result |
|---|
"Held out" means the item being scored was left out while the method was built: it is what to expect on airports and airlines it has never seen.
Findings
What did not work
Ideas are tried on real data before they are used; these did not earn their place.
Sources and licences
| Data | Source | Licence |
|---|
The measured table is published under the ODbL: download it (CSV). Not for operational use: always use your operator's and the aerodrome's official procedures.