Key points are not available for this paper at this time.
Proliferation of microplastics in rivers, lakes, estuaries, coastal waters and oceans is a major global challenge and threat to the environment, livelihoods and human health. Reliable predictive tools can play an essential role in developing an improved understanding of microplastics behaviour, exposure and risk in water bodies, and facilitate identification of sources and accumulation hot spots, thereby enabling informed decision-making for targeted prevention and clean-up activities. This study presents a new numerical framework (CaMPSim-3D) for predicting microplastics fate and transport in different aquatic settings, which consists of a Lagrangian, three-dimensional (3D) particle-tracking model (PTM) coupled with an Eulerian-based hydrodynamic modeling system (TELEMAC). The 3D PTM has several innovative features that enable accurate simulation and efficient coupling with TELEMAC, which utilizes an unstructured computational mesh. The PTM is capable of considering spatio-temporally varying diffusivity, and uses an innovative algorithm to locate particles within the Eulerian mesh. Model accuracy associated with different advection schemes was verified by comparing numerical predictions to known analytical solutions for several test cases. The implications of choosing different advection schemes for modeling microplastics transport was then investigated by applying the PTM to simulate particle transport in the lower Saint John River Estuary in eastern Canada. The sensitivity of the PTM predictions to the advection scheme was investigated using six numerical schemes with different levels of complexity. Predicted particle distributions and residence times based on the fourth-order Runge-Kutta (RK4) scheme differed significantly (residence times by up to 100 %) from those computed using the traditional first-order (Euler) method. The Third Order Total Variation Diminishing (TVD3) Runge-Kutta method was found to be optimal, providing the closest results to RK4 with approximately 27 % lower computational cost.
Pilechi et al. (Thu,) studied this question.