Performance criteria for the benchmarking of air quality model regulatory applications: the 'target' approach

The definition of appropriate performance criteria is one of the key issues for the benchmarking of air quality models in regulatory applications. As part of the FAIRMODE benchmarking activities (Thunis et al., 2010), suitable criteria for air quality modelling in the frame of the EU air quality directive (AQD) 2008 are proposed and tested. The suggested approach builds on the target indicator (Jolliff et al., 2009) as support to the relative directive error, the current official statistical parameter as defined in the AQD (EEA, 2011), for quantitatively estimating model performances in air quality modelling applications. This study describes the advantages of using the target compared to the actual limitations of RDE and addresses the main links between the target and some 'traditional' statistical indicators (MFB, R, FAC2, σ). It also describes the application of this methodology to NO2, O3 and PM10 concentrations on three different model-observations datasets. Among these datasets two focus on the urban areas of Madrid and London and include modelled results provided by the air quality models CMAQ and ADMS-Urban for years 2007 and 2008 respectively. One other dataset (POMI) covering the Po valley and including multiple model results has also been tested for year 2005