Practical considerations to speed up Lagrangian stochastic particle models

Abstract

The most serious drawback of Lagrangian stochastic particle models is their excessive demand of computing time. Even with today's powerful computers this makes them unattractive for many practical applications, such as for the assessment of air pollution in a long-term (e.g., yearly) scenario. This contribution therefore focuses on two issues: firstly, on which specific tasks do these models ‘waste’ computing time? Secondly, methods are presented in order specifically to reduce the computing time of these tasks. It is shown that for ‘simple’ models (e.g. those suited for Gaussian homogeneous turbulence) a speed-up factor of 10–100 can be reached when following these reduction strategies. The overall speed-up factor is about two for a more sophisticated model, which takes into account skewed and inhomogeneous turbulence. The results from the ‘speeded-up simulations’ are compared with the original model's results and it is shown that the speed-up procedures do not significantly alter the concentration patterns.