In general term a distinction between process-oriented models and statistical models can be made. Process oriented models are based on the description of physical/chemical processes: starting with emissions, atmospheric advection and dispersion, chemical transformation and deposition is calculated. This type of models is able to give a description of cause-effect relations. Statistical models are valuable tools in estimating present air quality by means of interpolation and extrapolation of measuring data.

Although atmospheric models are indispensable in air quality assessment studies, their limitations should always be taken into account. Once a model has been developed, the further application of the model will be relatively cheap; however, collecting the necessary input data might be cumbersome. Models can be used for estimating past, present and future air quality, provided that information on emissions is available.

The contribution of source regions, economical sectors etc. to the ambient levels can be easily deduced from model calculations. Uncertainties in model results may be large; uncertainties are both introduced by the model concept and by the input parameters (emission data, meteorology). The model results may be representative to a limited degree. In most models an implicit spatial and temporal average is introduced which may disable a direct comparison with measurements at one location at a given moment.

Different type of models
A wide range of different models have been published in scientific papers and even a larger number of unpublished models and special model versions exist. Models can be distinguished on many grounds: e.g. the underlying physical concepts, the temporal and spatial scale, type of component. Contemporary air pollution models deal with "conventional" primary pollutants (mainly SO2, CO, NOx and VOC). Already at present the need is recognised to extend the models to include heavy metals and persistent organic pollutants (POPs). Modelling of visibility and particulate concentrations (PM10) are among the most important current model development trends.

There are no well-defined requirements with respect to model documentation. This documentation should at least consist of a user manual, short technical description and the results of sensitivity and validation tests.

Four typical applications

Regulatory purposes
Model results are used in issuing emission permits (usually for single sources) or for environmental impact studies related to, for example, industrial plants and new roads. In general terms, models in this application area have to provide spatial distribution of high episodic concentrations and of long-term averaged concentrations for comparison with air quality guidelines. A wide range of pollutants is modelled (e.g. SO2, NO2, suspended particles, but also toxic substances like heavy metals and organics). In some situations the desired model output should include information on odorous components. It is indisputable that this might be beyond the scope of most of the models since the models used at present are not very suitable for handling concentration fluctuations in a proper manner.

In the framework of a European ad hoc initiative on Harmonisation within Atmospheric Dispersion Modelling for Regulatory Purposes standardised methods e.g. tools for model evaluation: reference data set, software and protocols for model evaluation are being developed (Olesen and Mikkelsen, 1992) : www.dmu.dk/AtmosphericEnvironment/Harmoni/MEpapers.htm

Policy support
The effect of abatement measures has to be forecasted by the models. This may require that the model also give reliable results under pollution conditions, which differ strongly from the present situation.

Use of atmospheric models in combination with models for other compartments (e.g. soil, water but also emission modules) in order to obtain a more integrated approach is becoming more and more important. For practical reasons this might imply that more simplified models without losing essential information has to be developed.

Public information
In public information the role of models is expected to grow. Requirements for models for public information parallel to a large extent those for policy support as far as it concerns assessment studies. On-line information to the public will be needed concerning air quality. Occurrence of smog episodes will have to be forecasted. 

Such forecasts are already available for several cities in Norway. 
See: www.nilu.no/niluweb/services/svo/

Scientific research
Among the major objectives for research type models are the description of dynamic effects and the simulation of complex chemical processes involving air pollutants. Until very recently, this type of models proved in most cases not to be suitable for practical applications: their requirement on computational effort was too high for application in the above three fields. Thanks to the tremendous computer hardware development, however, the situation is rapidly changing in favour of complex research type models. Hence, models of this type are not only valuable for identifying limitations and gaps in simpler policy oriented models; they could represent the proper policy supporting models in the near future. For this reason, research type models are also discussed, at least partially, in the present report.

Note: All external links opens in new window