Dispersion models

The science dealing with air quality and its modeling is closely related to meteorology. While meteorology is an old and well established science discipline, monitoring and modeling of air quality has a relatively short history.
The atmosphere has been polluted by natural processes since ever. The reasons are phenomena like wildfires, volcanic activity or crashes of extraterrestrial objects. Anthropogenic pollution of the atmosphere leads to problems at local levels since a few centuries. The source of this pollution is wood, other biomass, coal, oil, gas, waste and chemicals combustion. In the nineteenth and the beginning of the twentieth century atmospheric pollution was caused mainly by coal and products of chemical industry combustion. In 1905 Harold Antoine Des Voeux invented the term "smog" describing a phenomenon developed by a combined influence of smoke and fog, that occurred regularly in most industrial cities of the United Kingdom in this time. During a few following decades smog was the reason of events, during which thousands of people were dying because of increased concentrations of pollutants in air. The best known example is the situation in London in December 1952, when over 4000 people died owing to the combination of high concentrations of harmful substances in the atmosphere and bad dispersion conditions lasting for a few days. Therefore, such type of smog is called the "London smog".
The boom of motoring and industrial activities in the twentieth century increased the occurrence of another type of smog, called photochemical. According to the place of it’s most frequent and intensive occurrence, this type of smog started to be called the "Los Angeles smog". A research conducted in the half of the twentieth century suggested the major components of photochemical smog to be nitrogen oxides, tropospherical ozone and some volatile organic compounds
During the last two centuries the issue of air pollution was attracting more and more attention, despite the fact that progress in industry and energetic had a largely positive impact on air quality. As an example it is possible to name the invention of the electromotor, leading to a decrease in the of use of the steam engine.
In the half of the twentieth century first computers started to be available and this supported the introduction of first, simple (box models)able to simulate chemical reactions in the atmosphere. In the 1960s and 1970s models working with two and also three dimensional data started to emerge. These models were already comprising emission parameters of pollutants, as well as their transport, reactions and deposition. As meteorological data interpolated fields of various meteorological parameters were used (e.g. air temperature, dew point, wind speed and direction). Since the 1970s, besides pollution in urban areas, attention is focused also on relatively new phenomena such as acid deposition (acid rain), stratospherical ozone depletion or global climate change. Currently, models integrating the modeling of chemical processes in the atmosphere and the use of dynamic meteorological data are used to simulate such phenomena.

Currently, modeling of the dispersion of polluting substances is defined as a mathematical simulation of a pollutant dispersion in the free atmosphere. Special computer programs are constructed to conduct the simulation. These programs solve mathematical equations and algorithms simulating the dispersion of pollutants. Dispersion models are used to estimate or to predict (forecast) concentrations of airborne pollutants emitted from sources such as industrial facilities, local heating or traffic. These models are essential for the work of state agencies, which are responsible for air quality protection. Most frequently dispersion models are used to verify if current or planned industrial facilities correspond or would correspond the law standards for air quality protection of various countries. Dispersion models also assist the planning of strategies for the reduction of pollutants emissions.

smog L.A.

Dispersion models demand input data that usually contain:

Meteorological conditions (wind speed and direction, occurrence of different classes of atmosphere stability, air temperature, eventually its gradient and inversions occurrence)
Data describing emissions (location of the source, its height and diameter, velocity and volume of emitted flue gases, concentration of pollutants…)
Terrain elevation model for the area of interest
Location, height and width of eventual obstructions (e.g. buildings), located downwind from the source.

Go up

Classification of dispersion models

In dispersion modeling, various approaches can be applied. Here, it is distinguished between dynamic and statistic modeling.

Dynamic models are suitable for studying the influence of the dynamics of the pollution source (emission changes in time), e.g. in the case of accidental pollutant leakages or their dispersion in complex conditions (dense built up area, complex terrain). Dynamic models can be used for a detailed study of areas of limited size and specified meteorological conditions. On the contrary, they are not suitable for application in large areas and cases, when meteorological data for longer periods are statistically processed.
Statistic models use diffusion equations which are simplified using the knowledge obtained by real observations. They describe the real air flow in a simplified way and are based on simplified assumptions and limiting conditions (e.g. the pollution source is considered a point with constant emission). The models ATEM, AEOLIUS and SYMOS´97 are examples of statistic/static dispersion models. The last one is a mandatory method according to the Czech government directive 597/2006 on monitoring and evaluation of air quality. Each of the models has its field of application strictly defined. The model SYMOS´97 (introduced in more detail here (link – tutorial) is used especially in the Environmental Impact Assessment (EIA) process when the possible impact of pollution sources (new constructions or changes in technologies), has to be estimated. Also, SYMOS´97 is being used for complex evaluations of pollution issues in larger areas.

Dispersion models can be further classified into two groups, however, most of them use approaches located between these two limiting categories:

The black box method – doesn’t explain the principles of the processes, rather it describes the relation between input and output using a mathematical equation. These models are simple, but their applicability is limited
Methods based on the process mechanism knowledge – use the knowledge of particular processes occurring in the system. These are mathematically described. These models are more robust, though, they are more demanding regarding their construction and the computation time.

Approaches to emission modeling

smog

Most frequently models based on the process mechanism knowledge are used. According to their emission modeling approaches these models can be classified as follows:

Gaussian models are based on the Gaussian distribution which describes the distribution of pollutant concentration in the plume both in the vertical and horizontal dimensions. They are usually user friendly and computationally simple; therefore they are still widely used especially by state authorities for Environmental Impact Assessments, despite their obvious drawbacks. For instance, Gaussian models are usually not suitable for the computation of pollutant concentrations during situations with very weak wind, calms or inversions; they also often fail in the closest surroundings of sources. A usual feature of these models is also an insufficiently described meteorology. Gaussian models are often integrated into Lagrangean or Eulerian models. A typical example of Gaussian models used in Czechia is the dispersion model SYMOS´97 developed by the Czech Hydrometeorological Institute, the corresponding software was created by the Idea–envi company. An example of a worldwide used Gaussian model is the American model AERMOD.
Lagrangian models are based on the principle of the simulation of atmospheric particles releases conducted in regular time intervals (here the term „particle“ does not refer to a dust particle, rather it refers to a certain amount of pollutant defined in advance). These particles are carried on by the wind, which is usually plotted in a grid, where for each cell the wind speed and direction, stability of the atmosphere and eventually other characteristics are defined. A computation of paths (trajectories) of these particles is essential for the model. The modeled area is covered by a grid of reference points for which final pollutant concentrations are computed. Opposite to Gaussian models, these models usually offer more parameters that can be included into the simulation. Thus Lagrangian models are computationally more demanding and therefore not so often routinely used so far. Especially because of higher computational demands it is not possible to obtain the simulation results in such a good spatial resolution as when working with Gaussian models. On the other hand, Lagrangian models can be used for modeling much larger areas.
Eulerian models are able to describe the vertical transport of pollutants. This is of advantage e.g. in built up areas. They work on similar principles as Lagrangian models, which means pollutant concentrations are computed for particular grid cells. Their main advantage however is their ability to model also several chemical processes affecting the pollutants. Similarly to Lagrangian models, Eulerian models have a quite poor spatial resolution of output data (up to tens of kilometers). This problem is usually solved using a hybrid approach, when the modeling of chemical reactions of substances and background concentrations for a larger area is provided by an Eulerian model, while highly spatially resolved concentrations are computed by a Gaussian model.
Box models work with the space divided into a regular grid of cuboids (boxes) behaving like a closed system. This system reacts to inputs. Inside the box the entering amount of pollutants is modified by a system of equations simulating chemical and physical processes. Simplicity and computational modesty is considered an advantage of box models, however, especially meteorological inputs are usually quite imperfect. Box models are described in more detail in the chapter Box models.

Go up

Other criteria for classifying dispersion models

- Area extent which the model is suitable for

Dispersion models can be classified according to the spatial extent for which they can be applied. While some Gaussian models can be applied for areas up to tens of kilometers, others, especially Eulerian models are most frequently used for areas large as a continent.

- Ability to model the dispersion of various pollutants

Most models are able to simulate the dispersion of gaseous pollutants as well as dust particles. The latter is easier, because dust is not subject to chemical reactions and degradation in the atmosphere (however, reactions take place on its surface). However, it is necessary to know the distribution of the particulate size fractions, because smaller particles are usually transported over longer distances. In the case of gaseous pollutants the situation is more complicated. Especially Gaussian models are often not able to model chemical reactions affecting pollutants in the gaseous phase. However, usually these models can be used to estimate concentrations of SO_x, NO_x, ozone and other frequently monitored gaseous pollutants. A difficult issue is the modeling of substances partly bound on particles and partly in the gas phase. The particle-bound fraction usually has a longer atmospheric lifetime, so it is supposedly transported over longer distances. There can be hardly found scientific articles describing dispersion studies of such substances, opposite to dust and routinely monitored gaseous pollutants mentioned above. Such problematic substances are e.g. polycyclic aromatic hydrocarbons. At the RECETOX centre, some modeling experiments with these substances are taking place currently. An example is presented here.

- Number and types of sources that can be involved in the simulation

Generally, the simpler the model, the more sources can be involved in the computation (due to lower computational demands) and vice versa. This criterion is then stricter for the more computationally demanding Lagrangian and Eulerian models, which can model up to hundreds or thousands of sources. In case of Gaussian models the maximum number of sources is usually higher or unlimited.
Probably all dispersion models are able to cover emissions from point sources (industrial sources and other more significant sources like district heating plants or boiler plants). Linear sources are usually converted into points with constant distance in between, which however doesn’t mean that one segment of the road (delimited by two crossroads) equals one point. Spatial sources represent a large amount of point sources, e.g. a built up area. Some models also offer the possibility to cover volume (three-dimensional) sources. An example of a volume dust source is a dump or a pile of sand or coal. A volume source of gaseous pollutants could be e.g. a building with a high amount of small sources (chimneys, windows).

- Computational time and demands

The computational time is influenced by several factors. First it is the characteristics of the model itself, e.g. the number of phenomena involved in the computation. Further the computational time is increased by a higher number of sources and receptors, a higher spatial resolution of the computational grid and data describing the terrain, as well as a better quality of meteorological data and other factors. Last, but not least the duration of computation depends on the PC used.

- Spatial and temporal resolution of outputs

As mentioned before, Lagrangian and Eulerian models usually generate outputs in the form of a regular grid with a maximum spatial resolution of 1 km (supposing the modeled area of interest is much larger). From this point of view they can hardly compete with Gaussian models. These offer the possibility to generate a relatively dense grid of receptors that moreover can be complemented with an irregular grid where necessary.
Also regarding the temporal resolution, Gaussian models are more variable. They offer computations for long periods (up to years), but the results are often devalued by the use of low-quality meteorological data. Another issue is the interpretation of e.g. yearly averages. On the other hand, Eulerian and Lagrangian models generally dispose of meteorological data of relatively high quality, and their application on too long temporal periods would probably mean an unbearably long time of computation.

- Ability to model various pollutants

As specified above, only Eulerian models are able to simulate the reactions of modeled pollutants in the atmosphere faithfully enough. The possibility to edit additional parameters describing the behavior of pollutants differs among models in dependence on their perfection and version.

Go up

References

Bubník, J. et al.: SYMOS´97 - Systém modelování stacionárních zdrojů, Metodická příručka. Nakladatelství ČHMÚ, Praha, 1998. 60 s. ISBN 80-85813-55-6

Brunclík, T.: Aplikace GIS pro výpočet rozptylu emisí v atmosféře. [Disertační práce] [online]. ©2000, poslední revize 2.10.2006 [citováno 2008-03-14]. Dostupné z: http://webak.upce.cz/~uozp/rozptyl/download/dprac.pdf

Caputo, M., M. Giménez, et al. (2003). "Intercomparison of atmospheric dispersion models." Atmospheric Environment 37(18): 2435-2449.

Hirtl, M. and K. Baumann-Stanzer (2007). "Evaluation of two dispersion models (ADMS-Roads and LASAT) applied to street canyons in Stockholm, London and Berlin." Atmospheric Environment 41(28): 5959-5971.

Holmes, N. S. and L. Morawska (2006). "A review of dispersion modelling and its application to the dispersion of particles: An overview of different dispersion models available." Atmospheric Environment 40(30): 5902-5928.

Jacobson, M.Z. (2005). Fundamentals of Atmospheric modeling, Cambrigde University Press, New York, ISBN 0-521-83970-X

Stein, A. F., V. Isakov, et al. (2007). "A hybrid modeling approach to resolve pollutant concentrations in an urban area." Atmospheric Environment 41(40): 9410-9426.

Touma, J. S., V. Isakov, et al. (2006). "Air Quality Modeling of Hazardous Pollutants: Current Status and Future Directions." Journal of the Air & Waste Management Association (1995) 56(5): 547-558.

Jančík, P. (2003): Modely znečišťování ovzduší a GIS. Životné prostredie 37, 39-41

Zhang, Q., Y. Wei, et al. (2008). "GIS-based emission inventories of urban scale: A case study of Hangzhou, China." Atmospheric Environment 42(20): 5150-5165

Useful links

Wikipedia, modeling of dispersion in the atmosphere
http://en.wikipedia.org/wiki/Air_dispersion_modeling

Web pages of the German Lagrangian dispersion model LASAT
http://www.janicke.de/en/lasat.html

Web pages of the German Lagrangian dispersion model AUSTAL 2000, developed from the LASAT model (free download)
http://www.austal2000.de/de/home.html

Internetové stránky amerického eulerovského modelu CMAQ (volně ke stažení).
http://www.cmaq-model.org

Web pages of the American Eulerian model CAMx (free download)
http://www.camx.com

Web pages of the French dispersion model CHIMERE
http://www.lmd.polytechnique.fr/chimere/

Web pages of the British dispersion model ADMS 4
http://www.cerc.co.uk/software/adms4.htm

Web pages of American Gaussian dispersion models CALPUFF and AERMOD
http://www.epa.gov/scram001/dispersion_prefrec.htm

List of other dispersion models on Wikipedia
http://en.wikipedia.org/wiki/List_of_atmospheric_dispersion_models

Go up