The paper presents the adaptation procedure concerning the application of an ANN-based Air Quality forecasting model that was already developed for the city of Gdansk, Poland, and is now tested for the city of Thessaloniki, Greece. For Gdansk the model has taken into account the city’s meteorological parameters, which have been implemented using a one-way neural network for ease of learning, as well as the concentration levels of the pollutant of interest, PM10. In the process of teaching the network, four methods of propagation have been used (Back Propagation, Resilient Propagation, Manhattan Propagation, and Scaled Conjugate Gradient) for the purpose of choosing the best method. Results were then compared with real values which define the full network configuration (minimizing the forecast error). The model was then subjected to a process of adaptation for Thessaloniki. The data acquired through the process of adaptation regarding the PM10 levels used for both model training and testing purposes.