Bayesian inference of plastic sources by back-tracking virtual plastic particles in the Black-Sea

In this thesis project, the baseline of a new method has been developed to back-track the sources of plastic marine litter by coupling Bayesian inference with Lagrangian simulations of virtual particles tracking, for the Black Sea region. It has been concluded that the use of Bayesian statistics provides convincing results that can only be upgraded by the addition of new data. However, this approach still needs more thorough validation, through the use of observational data, to confirm its accuracy. In addition, the efficiency of this approach is limited by the quality of the prior knowledge and information about the studied domain. Specifically to the Black Sea, when only considering the largest rivers of the basin as a source of marine litter, it has been found that the Danube is the main contributor of plastic pollution in most of the zones of the Black Sea. In addition, the entropy of mixing has been calculated in order to understand over which timescales the sources of plastic could be inferred. For the open sea, the sources can be back-tracked over a timescale up to five years. After this period, all the particles are beached, and hence cannot be back-tracked anymore. This is mainly due to the northerly winds and the induced Stokes drift that drives the majority of the particles towards the Southern region of the Black Sea. Thus, if the plastic particle is located on the Southern beaches or along the corresponding coastal areas, its source can be inferred over a maximum timescale of 2 years.