Modelling the dispersion of aerial pollutants from agricultural buildings - an evaluation of computational fluid dynamics (CFD)

Agricultural buildings, especially those used for intensive livestock produ ction, are significant sources of aerial pollutants. This paper explores so me of the computational modelling methods available to predict near field c oncentrations of a pollutant emitted from a building opening. The modelling approaches used here are restricted in application to effectively weightle ss particles, such as a low concentration pollutant gas. The how patterns w ere predicted using computational fluid dynamics (CFD) and linked to this c omputed flow field were two dispersion models, a Eulerian diffusion model a nd a Lagrangian particle tracking technique, both used to predict ensemble mean gas concentration. Explicit account has been taken here of variations in mean wind direction, using a new technique based on the weighted summati on of individual wind direction results according to the probability densit y function of the wind direction. The atmospheric boundary layer is charact erised by variations in wind direction that do not occur in simulated wind tunnel flows and are not generally reproduced in current computational meth ods. For comparison, the results of the modelling approach are compared wit h mean concentrations of ammonia gas released as a tracer from an isolated low rise building. The results of this comparison indicate that at a distan ce of more than three building heights downstream the predictions from both models are satisfactory but that in the near wake the diffusion model is l ess successful. The weighted solutions, taking account of wind direction, g ive significantly improved predictions over unweighted results. Lack of plu me spread is identified as the main cause of inaccuracies in predictions an d this is linked to inadequate resolution of flow features and mixing in th e CFD model. The use of modelling based on CFD and existing dispersion mode ls is clearly limited and further work on non-steady state simulations of w ake Rows for dispersion studies is required. (C) 2001 Elsevier Science B.V. All rights reserved.