PREV'AIR, A Platform for Air Quality Monitoring and Forecasting

The PREV'AIR system is one among the few systems for air pollution forecasting currently running in Europe. Since 2003, observation data and model outputs are displayed daily on a web site (http://www.prevair.org/en/) in order to monitor and forecast surface chemical concentrations fields. Results are ozone, NO2, PM2.5 and PM10 maps. Statistical scores are also computed (maxima and daily average) and comparisons to the whole set of available near-real time surface measurements are performed. In summertime, alert criteria based on ozone concentration thresholds are evaluated ; in case of a major pollution risk, informations are sent to national authorities and televisions. In addition, the available ozone and PM10 observations data are used based on an optimal interpolation method to build 'analysis' maps integrating all sources of information. We first present the forecast system and the statistical scores computed for previous years: for example, for surface ozone concentrations forecasts during the whole summer of 2004, the bias is less than 5 ug/m-3, with a RMSE less than 19 ug/m-3 and a correlation factor of 0.82. This shows that the PREV'AIR system is able to forecast a large majority of the pollution events observed over Europe. An analysis of the causes of these errors is given. This gives further insights into which efforts have to be done in air quality modeling. In the framework of the GMES PROMOTE and GEMS European projects, PREV'AIR will quickly evolve in the next few years: we present, in the second part, the current projects concerning the models used. This includes developments both for meteorology and chemistry-transport modeling (CTM), particularly for aerosols (including dust). We also discuss the implementation of variational methods to assimilate satellite data. Finally, the feasability of using several models to deliver a single forecast (multi model approach) will be investigated.