Emission -> Transmission -> Immission

You need emission data (e.g. flow rate + concentration, possibly as a time series), transmission data (e.g. statistic of wind speed and atmospheric stability or similar, geometry of buildings or lanscape, depending on the model) and a critical receptor and possibly a background concentration (immission from sources outside the model).

The complexity of the input data depends on the purpose of the study and on the requirements of the model. The more you have, the better. But usually, the availability of input data is the problem of a dispersion analysis. You may have to rely on assumptions. If there is a "best practice guide", this will give you the minimum requirements of input data.