Nowadays we are able to forecast the levels of contamination in a region by using the air quality (AQ) models. These models can reproduce the main agents that influence the air pollution, like the meteorological situation, the emission sources, the dispersion of pollutants or the chemical reactions in the atmosphere. The result is the local concentration of pollutants in a set of points.
The AQ models could be classified by:
Dispersion models (ex. CALPUFF, AERMOD, etc) that simulate the transport of pollutants from the sources of emissions.
Photochemical models (ex. CMAQ, CAMx, etc) that reproduce the physical-chemical mechanisms in a 3D grid at a different scales.
The most advanced models are the photochemicals, however, they also need more data requirements. In order to obtain a trusted result, it is essential that the underlying input data would be realiable too, namely the meteorology and the emissions data. In that sense the AQ models are usually coupled with other numerical models that provide the input data that the former needs. This makes a modelization system built from:
A meteorological model, that provides the atmospheric parameters.
An emissions model, that provides the emissions rate at all sources (called “emissions inventory”).
An AQ model, that computes the concentration of pollutants from the data generated by the meteo and emissions models.
Since the meteorological information is provided from mesoscalar models (ex: WRF or HIRLAM), then the key factor is to develop a complete and exact emissions inventory.
But how can we compute the emissions rates when observational data is not available? In some cases, like industrial sources, these emissions could be known previously. But this information is rarely available, as for example traffic emissions. To solve that, the main environmental institutions in the world (like the EEA in Europe, or the EPA in USA) have defined technical procedures to build emissions inventories, computed from the characterization of activities. For instance, we could calculate the emissions due to road transport from the intensity of traffic, type of road, distribution and typology of vehicles, average speed, fuel composition or temperature at surface.
From our point of view, an optimal emissions models should support these features:
A suitable system must compute anthropogenic (ex: transport, industry, mining, …) and natural sources (ex: dust from erosion, marine spray, biogenic emissions, …).
The emissions model must have enough temporal resolution (hourly). Moreover it must support all spatial configurations (in terms of resolution and vertical levels).
The architecture must ease the implementation of new activities or methods.
When needed, the model must ingest meteorological data for computing suitable emissions rates.
The system must be able to export the model results to a set of AQ models (CMAQ, CALPUFF or AERMOD).
Finally, we would like to highlight the importance of developing an up-to-date model. From one year to another, the science improves and also the associated procedures to make the computations. The legal framework changes, cities progress and the emissions inventories too. So, it is important that the model could adapt to a fast-changing scenario in order to maximize its utility, like operative forecasts, as a tool for Environmental Impact Studies, to analyze correction measures versus pollutation, or to make sensibility studies.