Four ensemble methods are applied to two annual AQMEII datasets and their performance is compared for O₃, NO₂ and PM₁₀. The goal of the study is to quantify to what extent we can extract predictable signals from an ensemble with superior skill at each station over the single models and the ensemble mean. The promotion of the right amount of accuracy and diversity within the ensemble results in an average additional skill of up to 31 % compared to using the full ensemble in an unconditional way.