Air Traffic Control issues instructions to aircraft in order to prevent collisions by ensuring adequate separation between aircraft, as well as enabling the expeditious and orderly flow of air traffic. Other than a recent dip due to Covid-19, the number of flights flown annually has increased steadily each year, a trend which is expected to continue into the future. Thus has lead to a requirement to increase capacity and reduce climate emissions, without compromising on safety, that has driven large modernisation initiatives within Air Traffic Management Systems. Meeting these requirements requires the development of tools capable of providing decision support, improving workload forecasting and planning, enabling more fuel efficient procedures and systems, optimising route networks and improving network traffic prediction [1]. Digital Twinning provides a route through which some of these goals could be accomplished. A key component of a Digital Twin of airspace is the Trajectory Prediction (TP) method used to evolve the state of simulated aircraft. TP is uncertain due to the presence of significant epistemic uncertainties which can arise due to unknown aircraft mass, wind conditions, and aircraft performance settings. This talk introduces a hybrid method for generating probabilistic trajectories within a Digital Twin of UK airspace. A probabilistic model generates functional thrust and drag terms that parameterize a PDE describing aircraft flight mechanics, thereby ensuring that the generated trajectories are plausible [2]. The model presented in this talk is trained on a dataset of 1.2 million trajectories, harvested from 9 months of UK air traffic data and includes specific models for 120 aircraft types. Finally, this talk also discusses methodologies for validating the TP model against real-world data, to ensure that the distribution of real-world aircraft trajectories is captured.