STCA, TCAS, Airproxes and Collision Risk

Abstract

The focus here is on the performance of and interaction between the Traffic Alert and Collision Avoidance System (TCAS) and the controller's short-term conflict alert (STCA) system. The data source used is UK Airprox Board Reports of close encounters between aircraft, and the focus is on commercial air transport aircraft using UK controlled airspace with a radar service. Do the systems work well together ? Are controllers surprised when they find out that a pilot has received a TCAS resolution advisory? What do TCAS and STCA events say about collision risk ? Generally, the systems seem to work together well. On most occasions, controllers are not surprised by TCAS advisories : either they have detected the problem themselves or STCA has alerted them to it. The statistically expected rate of future mid-air collisions is estimated by extrapolation of Airprox closest encounter distances. K E Y

Are There Good Air Traffic Management Safety Indicators For Very Safe Systems?

“…Remedial action in the ATC system is therefore diverse and in depth. Brooker (2005b) illustrates the complex combinations that have been observed in UK airspace, from an examination of Airproxes (UKAB, 2006).…”

“…The need for a ' time buffer ' so that the processing can be carried out is one the reasons that separation minima are set to comparatively large values -horizontally 3 n.m in terminal areas/5 n.m en route, and 1000 feet vertical separation [NB : the higher en route figure corresponds to a higher aircraft speed compared with terminal areas]. Moreover, STCA and ACAS make huge differences to safety (Brooker, 2005b).…”

“…The first step is to simplify incident CPA descriptions by using a method previously introduced in studying Airproxes, STCA and ACAS (Brooker, 2005b). Define H (n.m) and V (feet) to be the miss distances at the CPA.…”

“…Simple extrapolations from the observed part of the CPI distribution are likely to be very cautious -e.g. see Brooker (2005b). Knowledge of the distribution of CPIs for regions where there is a substantial degree of system control cannot then be extrapolated with great confidence to estimate the shape -more important the height -of the curve f(x) in the 'Random Distances ' region.…”

“…[Note that the vertical scale here is linear not logarithmic.] This data is taken from Brooker (2005b), and covers five years from 1999 to 2003. In this data set there were 29 incidents with CPI<1000, of which one could well have a CPI small enough to fall into the 'Random Distances ' region.…”