Commercial aviation generates a vast amount of flight data, since recording and analysing this data is mandatory for the operators by law. Due to long lifecycles of aircraft and outdated recording standards, data quality however is low, which affects the quality of the analysis. In addition, the methods used to analyse the data are often very simplistic. Accident probabilities are e.g. computed as the quotient of accidents and the number of total operations. Given the small number of accidents compared to the total number of operations, these numbers are of little statistical significance. An ongoing field of research are model-based quantifications of accident probabilities using Markov Chain Monte Carlo (MCMC) approaches, where the model itself is solely deterministic.

The focus of this project is to combine widely valid, model-based approaches with flexible stochastic approaches in data-driven safety analysis. Therefore, we aim to develop a new statistical approach of characterizing temporal and special noise dependence structures via copula based state space representations. The idea is to combine this statistical model of complex noise structures with classical state space models (SSM) of aircraft motion and calibrate it, making use of flight data. Finally, we plan to expand the framework of data reconstruction and accident probability quantification (APQ), combining the physical and statistical representation of aircraft motion