Semi-Lagrangian transport algorithm for episodic atmospheric pollution

The paper describes a semi-Lagrangian transport algorithm for predicting the fate of episodically released atmospheric pollutants in multi-dimensions. The method employs a computational grid that aligns with the natural coordinate system of the atmospheric flow field. This results in a simplified advection-diffusion-reaction system that is perfectly suited to solution using an unconditionally stable semi-Lagrangian method. This uses third order interpolation for the advective fluxes with a convexity preservation algorithm to prevent grid scale oscillations: solutions are thus conservative, monotonic and highly accurate. The method is demonstrated with reference to two classic atmospheric transport problems: (1) the passive advection of a multiscale pollutant concentration distribution in a forced vortex and (2) advection-diffusion of a suddenly released pollutant mass in a uniform rectilinear flow. For the former case accuracy improves with increasing time step. Results for the latter are shown in terms of an accuracy portrait that reflects the influence of key non-dimensional parameters. Finally, the authors demonstrate that the approach is well suited to implementation on massively parallel computers.