Study Group report
This is the final report on the problem of distribution-independent safety analysis that was brought to ESGI49 by NATS.

Report authors:
Chris Budd (University of Bath)
Andrew Hill (University of Bath)
André Léeger (University of Bath)

Introduction
NATS brought to the Study Group two problems involving aircraft horizontal and vertical overlap safety margins. In the first problem, azimuthal error data is currently modelled by a double exponential distribution, where the central density is captured by one exponential and the tails by the other. The horizontal overlap probability (the probability that two aircraft are on the same bearing) is determined from tail-tail convolution. The Study Group report discovered the azimuthal outliers arise through random errors but also from systematic errors introduced by the way the data is processed. Gibbs phenomena are seen as a result of interpolation error in plotting aircraft paths and we have shown they can be removed by post filtering. In the second problem, vertical error data is currently modelled by a mixed Gaussian exponential distribution, and here the report has shown there is insufficient outlier data to fit any distribution, making any estimate of the vertical overlap probability invalid.

Download 'NATS.pdf' (7 Mb).