Combining Lagrangian simulations and Bayesian inference for source attribution of ocean plastic pollution

Ocean currents transport material like nutrients, plankton and plastic over the globe. The most natural way to study these transport pathways and the connections between ocean basins is by using trajectories, computed by simulating virtual Lagrangian particles in fine-resolution ocean models.

In this presentation, I will show how my team uses our open source parcels-code.org framework to simulate the dispersion of virtual plastic particles by the three-dimensional ocean flow. I will discuss how we develop new parameterizations for subgrid-scale transport processes of buoyant plastics; and compare these parameterizations to field measurements.

I will particularly focus on how we combine the resulting dispersion maps with estimates of plastic pollution sources and then apply Bayesian inference techniques to find the most likely sources for heavily polluted locations.

While our application is plastic pollution in the ocean, the framework could be applied in other geophysical contexts where the sources of a signal in a complex Lagrangian transport process have to be determined, from air pollution tracking to glaciological proxy reconstruction.