Trends of inorganic and organic aerosols and precursor gases in Europe: insights from the EURODELTA multi-model experiment over the 1990–2010 period

Abstract. In the framework of the EURODELTA-Trends (EDT) modeling experiment, several chemical transport models (CTMs) were applied for the 1990–2010 period to investigate air quality changes in Europe as well as the capability of the models to reproduce observed long-term air quality trends. Five CTMs have provided modeled air quality data for twenty-one continuous years in Europe using emission scenarios prepared by IIASA/GAINS and corresponding year-by-year meteorology derived from ERA-interim global reanalysis. For this study, long-term observations of particle sulfate (SO 4 2− ), total nitrate (TNO 3 ), total ammonium (TNH x ) as well as sulfur dioxide (SO 2 ) and nitrogen dioxide (NO 2 ) for multiple sites in Europe were used to validate the model results. The trends analysis was performed for the full twenty-one years (referred to as PT), but also for two 11-year sub-periods: 1990–2000 (referred to as P1) and 2000–2010 (referred to as P2). The experiment revealed that the models were able to reproduce the faster decline in observed SO 2 concentrations during the first decade, i.e. 1990–2000, with a 64–76 % mean relative reduction in SO 2 concentrations indicated by the EDT experiment (range of all the models) versus an 82 % mean relative reduction in observed concentrations. During the second decade, P2, the models estimated a mean relative reduction in SO 2 concentrations of about 34–54 %, which was also in line with that observed (47 %). Comparisons of observed and modeled NO 2 trends revealed a mean relative decrease of 25 % and between 19–23 % (range of all the models) during the P1 period, and 12 % and between 22–26 % (range of all the models) during the P2 period, respectively. Comparisons of observed and modeled trends in SO 4 2− concentrations during the P1 period indicated that the models were able to reproduce the observed trends at most of the sites, with a 42–54 % mean relative reduction indicated by the EDT experiment (range of all models) versus a 57 % mean relative reduction in observed concentrations, and with good performances also during the P2 and PT periods. Moreover, especially during the P1 period, both modeled and observational data indicated smaller reductions in SO 4 2− concentrations compared with its gas-phase precursor (i.e. SO 2 ), which could be mainly attributed to increased oxidant levels and pH-dependent cloud chemistry. An analysis of the trends in TNO 3 concentrations indicated a 28–39 % and 29 % mean relative reduction in TNO 3 concentrations for the full period for model data (range of all the models) and observations, respectively. Further analysis of the trends in modeled HNO 3 and particle nitrate (NO 3 − ) concentrations revealed that the relative reduction in HNO 3 was larger than that for NO 3 − during the P1 period, which was mainly attributed to an increased availability of “free-ammonia”. By contrast, trends in modeled HNO 3 and NO 3 − concentrations were more comparable during the P2 period. Also, trends of TNH x concentrations were, in general, under-predicted by all models, with worst performance for the P1 period than for P2. Trends in modeled anthropogenic and biogenic secondary organic aerosol (ASOA and BSOA) concentrations together with the trends in available emissions of biogenic volatile organic compounds (BVOCs) were also investigated. A strong decrease in ASOA was indicated by all the models, following the reduction in anthropogenic NMVOCs precursors. Biogenic emission data provided by the modeling teams indicated a few areas with statistically significant increase in isoprene emission and monoterpene emissions during the 1990–2010 period over Fennoscandia and Eastern European regions (i.e. around 14–27 %), which was mainly attributed to the increase of surface temperature. However, the modeled BSOA concentrations did not linearly follow the increase in biogenic emissions. Finally, a comprehensive evaluation against positive matrix factorization (PMF) data, available during the second period (P2) at various European sites, revealed a systematic under-estimation of the modeled SOA fractions of between a factor of 3 to 11, on average, most likely because of missing SOA precursors and formation pathways, with reduced biases for the models that accounted for chemical aging of semi-volatile SOA components in the atmosphere.