A three-dimensional Eulerian transport and transformation model driven by observation-derived synoptic meteorological data has been applied to calculate mixing ratios (MRs) of sulfate and SO2 and wet deposition of sulfate over the North Atlantic and adjacent continental regions for 1-month periods in each of four seasons in 1986-1987. Model performance is evaluated by comparison of grid-cell average (1.125 degrees) modeled MRs for sulfate (24-hour average) and SO2 (6- and 24-hour average) in the lowest model level (surface to similar to 65 m) to surface MRs observed at monitoring stations in North America and Europe. For sulfate similar to 8000 model-observation comparisons were made employing similar to 10,000 individual measurements; for 24-hour SO 2, 21,000 comparisons (54,000 measurements) and for 6-hour SO2, similar to 71,000 comparisons (211,000 measurements). Subgrid variation of observed MRs is inferred from the spread of multiple simultaneous measurements within individual grid cells. The median spread of the observed MRs is a factor of 1.5 for 24-hour sulfate and 2.2 for 24-hour SO2. The median spread between observed and modeled MRs is a factor of 2.3 for sulfate and 2.1 for 24-hour SO2, comparable to that for the observations themselves. This suggests that much of the departure between modeled and observed MRs can be attributed to subgrid spatial variation and nonrepresentative sampling of model grid cells at the stations used for the comparisons, For SO2 the median ratio of modeled to observed MRs is 0.97, with little seasonal variation, somewhat lower in North America but considerably higher in Europe; little difference was evidenced in comparisons of 6-hour averages versus 24-hour averages. For sulfate the median ratio is 0.51, with the range for the four simulation periods 0.36 to 0.66, lowest in January-February 1987, and with comparable values for Europe and North America. For all four simulations the time series of 24-hour average modeled MRs at most locations rather closely reproduce the magnitudes and temporal episodicity of the observed sulfate and SO2 MRs. Analysis of correlations of observed and modeled MRs was carried out for all grid cell locations for which at least 25 days of observations were available in a simulation period; 76% of 203 correlations for 24-hour sulfate and 51% of 526 correlations for 24-hour SO2 were significant at the 95% confidence level. The superior model performance in this respect for sulfate is attributed to the lower subgrid variation in the mixing ratio of this mainly secondary atmospheric species versus the mainly primary emitted species SO2. Comparisons of modeled and observed sulfate wet deposition (concentration times precipitation amount) for similar to 300 daily and similar to 1100 weekly samples, all in North America, indicate a median spread between modeled and observed deposition of a factor of 2.6 and a median ratio of modeled to observed deposition of 0.82. The major contributor to model underestimation of sulfate MR in air is tentatively attributed to the lack of representation in the model of the aqueous-phase conversion of SO2 to sulfate in nonprecipitating clouds.
