The management of collision risk posed by recreational unmanned aerial systems (UAS) intruding into controlled airspace is becoming more critical with the surge in accessibility and popularity of these UAS. Risk mitigation actions that could be taken by the airport operators currently are limited by the lack of reliable UAS detection equipment, which limits their ability to track UAS positions over time and predict the collision risks posed by the UAS. While recent developments in airborne collision prevention of manned aircraft could utilize Markov Decision Process with state probabilities based on historical flight track records and processed using Bayesian Network, this method is not suitable for the off-nominal case of UAS intrusion into controlled airspace. Instead, the prediction of collision risk posed by non-cooperative recreational UAS have to rely on the assumption of worst-case intention, where the UAS aims for the aircraft operating within the aerodrome, and the Reich collision risk model to generate the probable distribution of future UAS positions. This paper documents a series of flight test to simulate such scenario with a UAS operating under visual line of sight condition while aiming for an imaginary three dimensional target in the air. The data was analyzed for the deviation in UAS positions at fixed time interval in the (horizontal) longitudinal and lateral direction, as well as the deviation in altitude. A comparison between the observed deviation and a Monte-Carlo based UAS path prediction following the UAS flight dynamic model were also conducted.