PM measurement campaign HOVERT in the Greater Berlin area: model evaluation with chemically specified particulate matter observations for a one year period
1Laboratoire Interuniversitaire de Systèmes Atmosphériques (LISA), CNRS UMR 7583, University Paris 7, University Paris 12, 61 Av. du Général de Gaulle, 94000 Créteil, France
2Freie Universität Berlin, Institut für Meteorologie, Carl-Heinrich-Becker-Weg 10, 12165 Berlin, Germany
3BTU Cottbus, Lehrstuhl für Luftchemie und Luftreinhaltung, Burger Chaussee 2, 03044 Cottbus, Germany
Abstract. The HOVERT (Horizontal and VERtical Transport of Ozone and particulate matter) campaign held in the Berlin Brandenburg area in Eastern Germany from September 2001 to September 2002 allowed to collect a unique data set of the aerosol chemical speciation (daily averages) at traffic, urban and rural sites. These observations are used for a thorough evaluation of the aerosol part in the REM-CALGRID model (RCG) developed at the Free University of Berlin (FUB). For inorganic ions (sulphate, nitrate and ammonium), simulated annual averages agree to observations within ±30% at more than half of the sites and always within a factor of two. Correlation coefficients are larger than in previous studies for SO42- and NH4+ (>0.7). For nitrate, less elevated correlations, 0.4–0.7 in the cold season, 0.2–0.4 in the warm season, are encountered. To our knowledge, this is one of the first comparisons of air quality model simulated elemental carbon (EC and OC) with daily observations for a whole year. It suggests an overestimation of EC and OC emissions in the Berlin area (through a scaling techniques between EC, OC and NOx and when assuming correct NOx emissions), and an underestimation of EC and OC at rural sites. Secondary organic aerosol (SOA) formation, recently introduced into the model (SORGAM module, Schell et al., 2001), is simulated as a very variable process, SOA levels varying from close to zero for most days to more than 5 µg/m3. Correlation between simulated SOA to observed OC is about 0.6, indicating that simulated variability partly corresponds to reality.