A review of dispersion modelling and its application to the dispersion of particles: An overview of different dispersion models available

Abstract

This paper provides the first review of the application of atmospheric models for particle dispersion. The different types of dispersion models available, from simple box type models to complex fluid dynamics models are outlined and the suitability of the different approaches to dispersion modelling within different environments, in regards to scale, complexity of the environment and concentration parameters is assessed. Finally, several major commercial and non-commercial particle dispersion packages are reviewed, detailing which processes are included and advantages and limitations of their use to modelling particle dispersion. The models reviewed included: Box models (AURORA, CPB and PBM), Gaussian models (CALINE4, HIWAY2, CAR-FMI, OSPM, CALPUFF, AEROPOL, AERMOD, UK-ADMS and SCREEN3), Lagrangian/Eulerian Models (GRAL, TAPM, ARIA Regional), CFD models (ARIA Local, MISKAM, MICRO-CALGRID) and models which include aerosol dynamics (GATOR, MONO32, UHMA, CIT, AERO, RPM, AEROFOR2, URM-1ATM, MADRID, CALGRID and UNI-AERO).

Introduction

Dispersion modelling uses mathematical equations, describing the atmosphere, dispersion and chemical and physical processes within the plume, to calculate concentrations at various locations. Whilst, there have been various review papers on atmospheric modelling and their approaches to dispersion in street canyons (Vardoulakis et al., 2003) and comparisons between different models using test meteorological data (Ellis et al., 2001; Sivacoumar and Thanasekaran, 2001; Hall et al., 2002; Caputo et al., 2003), these have focussed on modelling gaseous dispersion.

Unfortunately, only a few studies have simultaneously measured particle concentration with gases and the differences between the studies may be partially responsible for the differences observed. In open sites several studies have shown varying correlations between the concentrations of gases and particles. Monn et al. (1997) showed a poor correlation between the outdoor PM₁₀ concentrations and NO₂ concentrations in an urban environment with a better correlation between PM_2.5 and NO₂, although only two locations were studied in the latter case. In contrast, Clairborn et al. (1995) showed a good correlation between SF₆ and PM₁₀ although only distances up to 60 m from the motorway were measured. Roorda-Knape et al. (1998) observed that benzene, PM_2.5 and PM₁₀ showed no significant decrease in concentration up to 300 m from a major motorway. This was consistent with the small decrease in the PM_2.5 concentration observed by Hitchins et al. (2000). In that study the authors observed that particle number concentration decreased faster than NO₂ concentration from a motorway. Zhu and Hinds et al., 2002a, Zhu and Hinds et al., 2002b showed that number concentration of particles between 6 and 220 nm correlated well with CO concentration from a motorway. All of these studies were made in an open environment where the wind direction was perpendicularly away from the road. However, differences have been observed between the local dispersion of gases and particles (Morawska, 2003; Holmes et al., 2005). Simultaneous measurements of CO and particle number concentration showed that CO concentration was not significantly correlated to particle number concentration around the site and examination of between-site comparisons with the two pollutants showed different spatial and temporal trends. In another study of urban sites Harrison and Jones (2005) observed that particle concentrations correlated only weakly with NO_x, with the highest correlation observed at a curbside monitoring location, where concentrations are less affected by dispersion. In addition, an examination of many urban studies (Morawska, 2003) has shown that the vertical profiles of particle number concentration around buildings differed from that of gases. These studies differ from the previous studies in that they were conducted in a more complex environment where wind flows were heavily affected by turbulence and emissions were not limited to a single line source. In general the studies show that in open environments the gas and particle concentrations correlate quite well, whilst in a more complex urban environment significant differences are observed between gas and particle dispersion. In an urban environment where traffic emissions are the dominant source of particles Van Dingenen et al. (2004) showed PM_2.5 and PM₁₀ had an R² value of 0.95 across all sites in the monitoring network. However, the PM₁₀/PM_2.5 ratio varied too much to propose a single PM₁₀/PM_2.5 ratio. In the same study they observed no correlation between annual average particle number concentration and either PM_2.5 or PM₁₀ concentrations. This is in contrast to Harrison et al. (1999) who found that in an urban measurement study hourly particle number concentration more closely correlated with PM_2.5 than PM₁₀ measurements, although both PM ranges showed good correlation with the hourly particle number concentrations during the 3 month period.

Therefore, models that are designed to model the dispersion of passive scalars, such as inert gases should be capable of modelling the PM_2.5 and PM₁₀ concentrations in certain open environments, especially for longer averaging periods and in the larger airshed where short term variations resulting from transient particle formation events are evened out.

The modelling of particle number concentration involves the incorporation of aerosol dynamics modules into dispersion models. Thus the discussion of particle dispersion modelling must involve both a discussion of the limitations of the various dispersion approaches to the treatment of particles and the aerosol dynamic packages used to evaluate particle processes occurring within the plumes. To complicate the situation further, Lohmeyer (2001) observed that concentrations calculated by the different models differed by a factor of four and even when the same model was employed results varied between groups. The agreement with predicted concentrations was seen to depend on the quality of the input data.

This review will outline the different model types, looking at specific requirements for the different spatial scales from local to regional models, and deficiencies with respect to particle dispersion and aerosol dynamics within different scales. In addition, whilst not being a comprehensive review of all models available a large number of models are included in the review and the more important model parameters and inputs for the models are listed in Table 1a, Table 1b, Table 2.

Although several models claim to be able to model particle dispersion, without specific treatment of particle dynamics the results are limited to calculation of particle mass, usually in the form of PM_2.5 and PM₁₀, and cannot calculate particle number concentration.

Furthermore particle validation studies are not available for many models. Where this is the case the authors have attempted to highlight model performance in terms of gas dispersion validation studies. Since several studies have shown a good correlation between non-reactive gases and particles within a larger airshed, validation studies involving gases should be a good indicator of the performance of the model in terms of calculating particle mass concentrations, as discussed earlier. In addition, different averaging times between average gas and particle concentrations make comparison difficult and mean that it is often impossible to determine whether changes between gas and particle concentrations predicted by the model correlate so well.

A number of local and regional models exist that include extensive treatment of aerosol dynamics. The majority of these are non-commercial packages and have been coupled to existing dispersion models in order to provide a package that is able to model changes to particle number concentration within different size groups. This means that the performance of these models depends on both the accuracy and specific processes included in the dynamics module as well as the performance of the dispersion model. It is often possible to integrate the aerosol dynamics module with different dispersion models to adapt the coupled dispersion package to better suit the planned study.