It depends upon what’s your objectives are. If you want to identify sources, regional transport of pollutants, local street canyon effects or role of meteorological factors in pollutant dispersion there are various models available each have its own advantage and disadvantages. Another very important thing is what kind of data you are working on (satellite data, real-time air quality monitoring data or long-term historical data, personally collected data) and resolution of the data. You can use existing model or you can create your own models.
Go through these attached papers, Hope it may help you.
The models used in air pollution research consist of a system of mass conservation equations with terms representing the chemical transformations of gases and aerosols suspended in the atmosphere. The equations can be expressed in Eulerian or Lagrangian reference frames, the former being the most common in recent applications, leading to a set of advection-diffusion-reaction equations. The surface deposition and emission processes in such a system are represented by the lower boundary conditions, while emissions from volume sources are parametrized as forcing terms on the right side of the transport equation. The typical air pollution model consists of a large set of equations of this type, very often of the order of 100. The separate problem is the specification of emission data and meteorological fields, including wind, temperature , the distribution of hydrometeors and the surface properties. The complexity of meteorological processes is one of the causes of the inclusion of all chemical transport equations in the numerical weather prediction model. One of the best known models of this type is a WRF / chem system, the description is available in the user guide:
https://ruc.noaa.gov/wrf/wrf-chem/Users_guide.pdf
Some additional information on models used in air quality research are also provided in a US EPA website
It basically depends on what is your main objective. For example, for regulatory purposes Gaussian based plume models are used. Now, even in Gaussian based models there are both analytical and numerical models with numerical models having the ability to handle different and more detailed input parameters. Other air dispersion models based on eularian, lagrangian and CFD model based systems. These models basically handle local scale prediction of pollutants. other mesoscale models like CMAQ, WRF predict pollution concentrations on regional scale
You need to narrow down your problem by asking some question to select the appropriate model for your case. For example What is your emission source? Point? Area? Volume? line?/ or What are the available data in the study area regrading emission sources and atmospheric parameters?/ and other question like your aim of study, your desired accuracy and ....
You can find a list of models in the following link: