Parallel Implementation and Scalability Results of a Local-Scale Air Quality Model: Application to Bamako Urban City

Abstract

We present a computational framework for the parallel implementation of a local-scale air quality model described by an advection-diffusion-reaction partial differential equation, the so-called equation of reactive dispersion. The temporal discretization of the model is carried out using the forward Euler scheme. The spatial discretization is achieved using the finite element method. The strategy used for the parallel implementation is based on the distributed-memory approach using the message-passing library MPI. The simulations are focused on two road traffic-related air pollutants, namely particulate matters PM2.5 and PM10. The efficiency and the scalability of the parallel implementation are illustrated by numerical experiments performed using up to 128 processor cores of a cluster computing system.