3. Model evaluation
In order to assure that the estimates provided by AERIS are reasonable, it is essential to carry out an analysis on its consistence and robustness as to evaluate its “fitness-for-purpose”. No universal consensus has been reached so far on good practices to evaluate model performance (Chemel et al., 2010 and Alexandrov et al., 2011). Aiming to be as rigorous as possible, the evaluation was conducted for two relevant aspects as shown in detail in the following sections. First, the congruence of the modelling approach was examined through an additivity test where the outputs of the AQMS were used as reference for assessment purposes. Then, in order to check the accuracy of AERIS estimates, its performance was compared: (i) against air-quality observations for year 2011 and (ii) against the predictions of the AQMS for a future emission scenario in 2014. It is worth noting that the results of the AQMS have been already contrasted against observations independently (de la Paz et al., 2013).