A New High Performance Version of the Lagrangian Particle Dispersion Model Spray, Some Case Studies

Abstract

Lagrangian particle models are powerful tools to simulate the atmospheric dispersion of gaseous releases. Although having a quite complex mathematical basis (Thomson, 1987, Rodean, 1994), their practical implementation is generally simple and intuitive, allowing to easily take into account complex situations such as the presence of the topography or terrain inhomogeneities, low wind speeds, spatial and temporal variations of meteorological fields. In these models the atmospheric dispersion is simulated by the motion of fictitious particles splitted in a mean part due to the mean wind, and a stochastic fluctuation related to the statistical characteristics of the turbulent flow. It is quite clear that the model accuracy is strongly dependent on the number of emitted particles and the computer time often limits the kind of simulations that can be performed. For this reason, the earlier version of these models were mainly devoted to reproduce the dispersion of a limited number of emissions at local scale. The recent wide and rapid diffusion of very fast computational tools lead to the development of more sophisticated codes, able to take into account more general situations. SPRAY (Tinarelli et al., 1992) is a Lagrangian stochastic particle model designed to perform dispersion simulations in complex terrain. The version 1 of the code, based on a three dimensional form of the Langevin equation for the random velocity with coupled non-gaussian random forcing following Thomson (1984, T84 in the following) and subsequently improved (Tinarelli et al., 1992), was able to satisfactorily reproduce local to regional scale dispersion both over flat (Brusasca et al., 1989, 1992) and complex terrain (Brusasca et al., 1995, Nanni et al., 1996) taking into account the emission from single or multiple sources. The development of a better based theory (Thomson, 1987) and the further demand of more complex regional scale simulations able to cover longer periods with a variety of emissions of different kinds (i.e. main roads, industrial or urban area) called for a new version of the code. The new version 2 of SPRAY code contains some improvements regarding the theoretical approach, turbulence parametrizations and time response characteristics. In this paper we describe these new developments, comparing model performances with those of the previous version through simulations performed both in theoretical and real cases.