How can we use data to analyse aircraft landings?

In the study of aircraft safety the behaviour of the aircraft in landing is crucial. If we can gain a better understanding of what happened and why during both normal and abnormal landings, we can identify opportunities for improvement as well as threats to be addressed.This process starts with identifying when the aircraft touched down. Once we know this we can look at its behaviour on the runway as well as its behaviour on the approach to touchdown. However, although we can identify the moment of touchdown with reasonable accuracy, you may be surprised to know that often the data we get from the FDR (flight data recorder) is often much less accurate about where we touched down. We of course know the airport and the runway but that's about it.

Plotted here you can see the locations of the aircraft at the moment of touchdown as recorded in the FDR:

You can see that the accuracy of the latitude/longitude data is very poor. This is because it has been recorded in a short format in order to save space and is only recorded at a low rate. The latitude and longitude are logged alternately once per second. One second the longitude is recorded and the next it’s the latitude, etc.

To improve the accuracy of touchdown localisation we have been looking at the use of terrain data and the data recorded from the radar altimeter of the aircraft. This instrument is recorded with a much higher rate and accuracy and, along with the aircraft speed and direction, we can use it to build up a profile of the terrain below the flight path of the aircraft.

Using Python and, in particular, the Folium library to generate leaflet.js maps we can combine our flight data with mapping data from Ordnance Survey, data from OpenStreetMap, and from the Ports of Jersey. We were able to successfully match the calculated terrain profiles to the terrain on approach to the runway at Jersey Airport and generate the map shown for the same flights shown in the previous map:

As you can see, once we have done the terrain matching there is a much more useful form of data ready for us to use to learn more about aircraft landings and how to improve them.

Research continues to allow us to use the taxiway data as a comparison metric for the accuracy of the terrain matching fit to the aircraft trajectory.

If you would like to know more about how it was done, why not have a look at this talk from PyData London.

SAI brochure