Predictability of European air quality:: Assessment of 3 years of operational forecasts and analyses by the PREV'AIR system

For the first time, the long-term evaluation of an operational real-time air quality forecasting and analysis system is presented, using error statistics over 3 consecutive years. This system, called PREV'AIR, is the French air quality forecasting and monitoring system. It became operational in 2003 as a result of a cooperation between several public organizations. The system forecasts and analyzes air quality throughout Europe, with a zoom over France, for regulatory pollutants: ozone (O-3), particulate matter with diameter smaller than 10 mu m (PM10), and nitrogen dioxide (NO2). The ability of PREV'AIR to forecast, up to 3 days ahead, photochemical and particle pollution over the domains considered is demonstrated: daily ozone maxima forecasts correlate with observations with 0.75-0.85 mean coefficients; U. S. Environmental Protection Agency acceptance criteria relative to the forecast accuracy for high concentrations and daily maxima are met for more than 90% of the measurement sites. For NO2 and PM10, the performance corresponds to the state of the art. The contribution of weather forecast errors to air quality predictability is addressed: ozone daily maxima forecast errors are not dominated by meteorological forecast errors; for rural stations, only 6% (15% and 25%, respectively) of the error variance is due to meteorological forecast errors on the first 24 (48 and 72, respectively) hours. The Model Output Statistics procedure, implemented in PREV'AIR, is proved to improve ozone forecasts, especially when photochemical pollution episodes occur. The PREV'AIR real-time analysis procedure, based on a kriging method, provides an accurate and comprehensive description of surface ozone fields over France.