With the aspiring development towards Trajectory-based Operations, novel tools for robust trajectory prediction are necessary. For this, the impact of uncertain input variables to the trajectory prediction must be understood to permit higher automation with increasing look-ahead times. In this study, a neural network provides input probability density functions for the aircraft mass and speed intent (multiple phases with constant calibrated air speed or Mach number). With our flight performance model, 10,000 climb phases are predicted in a Monte-Carlo simulation with a look-ahead time of 600 seconds for six different aircraft types. The resulting trajectory uncertainty is analyzed to prove that the stochastic characteristics of the input can be used to predict the arising uncertainties in future positions. Since the selected uncertainties are interdependent and time-lagged, the normality of the input vanishes in the trajectory uncertainty. However, a Beta distribution provides a good fit for up to 90% of the cases. The findings are applicable to decision support tools if the expected uncertainty shall be included.