Atmospheric sulfur cycle simulated in the global model GOCART: Comparison with field observations and regional budgets

We present a detailed evaluation of the atmospheric sulfur cycle simulated in the Georgia Tech/Goddard Global Ozone Chemistry Aerosol Radiation and Transport (GOCART) model. The model simulations of SO2, sulfate, dimethylsulfide (DMS), and methanesulfonic acid (MSA) are compared with observations from different regions on various timescales. The model agrees within 30% with the regionally averaged sulfate concentrations measured over North America and Europe but overestimates the SO2 concentrations by more than a factor of 2 there. This suggests that either the emission rates are too high, or an additional loss of SO2 which does not lead to a significant sulfate production is needed. The average wintertime sulfate concentrations over Europe in the model are nearly a factor of 2 lower than measured values, a discrepancy which may be attributed largely to the sea-salt sulfate collected in the data. The model reproduces the sulfur distributions observed over the oceans in both long-term surface measurements and short-term aircraft campaigns. Regional budget analyses show that sulfate production from SO2 oxidation is 2 to 3 times more efficient and the lifetimes of SO2 and sulfate are nearly a factor of 2 longer over the ocean than over the land. This is due to a larger free tropospheric fraction of SO2 column over the ocean than over the land, hence less loss to the surface. The North Atlantic and northwestern Pacific regions are heavily influenced by anthropogenic activities, with more than 60% of the total SO2 originating from anthropogenic sources. The average production efficiency of SO2 from DMS oxidation is estimated at 0.87 to 0.91 in most oceanic regions.

These explanations include too much wet deposition or vertical ventilation, and too little SO•. oxidation. While all of the above are possible, none of them seems sufficient to account for sometimes a factor of 10 difference between the model and the measurements. The same discrepancy was also reported in the EMEP modeling and measurements comparison studies [Schaug et al., 1993;Iversen, 1993], with a systematic low bias in the model at coastal sites. It was suggested that contributions from uncounted sources could be one possible explanation [Schaug et al., 1993].
Here we suggest that a sea-salt component in the data is likely an important source of the discrepancy, The sea-salt contribution is expected to be most significant in winter when sea-salt emissions over the North Atlantic are higher due to strong winds. The precipitation measurements in the EMEP network also indicate that sea-salt sulfate could contribute up to 50-90% in winter at some locations. To test the possible seasalt contrib•tion to the sulfate aerosol data, we use the preliminary results of the GOCART model simulated monthly averaged sea-salt concentrations below 800 m for 1990 (P. Ginoux, unpublished results, 2000), assuming that 7% of the sea salt mass is sulfate. The modified model results, that is, non-sea-salt sulfate plus sea-salt sulfate, are shown in the dash-dotted lines in Figure  5. This simple modification res•llts in a much better agreement between the modeled and observed sulfate in winter at, for example, High •h•fftes (United Kingdom). Vanhill (Sweden), and Faeroeme-Akraber (Denmark). It should be noted. however. that what we try to demonstrate here is a possible magnitude rather than a quantitative assessment of the sea-salt sulfate amount in the EMEP data. A quantitative correction will require airborne sea-salt measurements at the EMEP sites.

Regional Budget
The generally good simulations of sulfur species provide the base for analyzing the regional budgets with the  Table 1 shows the column budget of SO2 and sulfate in the polluted regions of NAM, EUR, and EAS, defined in Figure 21. Anthropogenic emission is the predominant source of SO2 there, with the highest emission rate over Europe. However, nearly half of the SO2 emitted is dry-deposited to the surface. Dry deposition dominates the removal of SO2 over the continent particularly in winter when SO2 oxidation is slow. The sulfate production efficiency, which is defined as the amount of sulfate produced relative to the amount of SO2 emit-  (Table 1). The export flux is much lower than that estimated for eastern North America in previous studies ranging from 30% to 59% [Wojcik and Chang, 1997; and references therein]. The major difference is that the dry deposition loss for SO2 in this study is much higher (50%) than that in the previous estimates (17-36%). However, as we have discussed in section 3, the model results of SO2 over the polluted regions suggest that a higher SO• loss rate is required. The net export for the EUR region is 23%, which is within the wide range of previous estimates in the literature (8-80% [see Wojcik and Chang, 1997]). Table 2 is the column sulfur budget over the four oceanic regions defined in Figure 21. The NAL region receives a significant amount of anthropogenic SO2 from the neighboring continents, that is, about 40% of the total SO• source. This influx may be overestimated, if the modeled SO2 concentrations are in fact an overestimate (section 3). Ship emissions account for another 24% of the anthropogenic SO• source in the NAL. Capaldo et al. [1999] estimated that 40-80% of the surface SO2 over the North Atlantic region in July is from ship emission. We have examined the July surface SO2 budget of the NAL region and estimate that 65% of the SO2 source in the model surface layer (0-50 m) is from ship emission. However, the large contributions of ship emission to SO2 concentrations are likely limited to near the surface; we have found that in the lowest 2-km column, DMS oxidation is a factor of 1.75 higher than ship emission as the SO2 source over NAL in July. Therefore it appears in our model that both the continental outflow and DMS oxidation (36%) are more important than ship emission as the SO2 and sulfate source over the North Atlantic region. Similar to NAL, the anthropogenic input from continents is the largest source of SO2 in the NWP region (56•). Ship emissions contribute only 5% of the SO2 source. Accordingly, the total anthropogenic source of SO2 over the NWP region is 61%, similar to the fraction of 64% over the NAL. The remaining 39% of SO2 is from DMS oxidation (32%) and volcanic emission (7%).

Sulfur Budget in the Oceanic Regions
Unlike the polluted regions of NAL and NWP, Table 2 shows that only 12% of the SO2 source in the SWP region is from outside of the column. While DMS oxidation is the most important SO2 source (53%), volcanic emission is also significant; it contributes more than 30% of the SO2 source in the SWP region. The TEP is the cleanest region among the four with little direct anthropogenic emission. With only 7% of SO2 from volcanic emissions and 16% imported, DMS oxidation accounts for 77% of the regional SO2 source.
The   in-cloud oxidation and/or dry deposition rates are sufficiently higher than those in the real world that they could be counted as total SO2 heterogeneous oxidation and total deposition.

Conclusions
We have presented a detailed evaluation of the atmospheric sulfur cycle simulated in the GOCART model.
We have compared the simulated S02, sulfate, DMS, and MSA concentrations with the observations over polluted continental source regions, anthropogenically modified oceans, and in remote environments. The comparisons have been conducted on various time scales from multiyear surface networks to short-term aircraft campaigns. The model in general reproduces the observed spatial and temporal distributions and captures the local and regional features.
Over the polluted regions of North America and Europe where the sources are thought to be well understood, the model reproduces the seasonal variations of S02 but overestimates the atmospheric level of S02 by more than a factor of 2 on average. On the other hand, the modeled sulfate levels agree with the observations within 30% as the sulfate production rates are controlled by the oxidant concentrations. The model results suggest that either the S02 emission rates are too high or an increased S02 loss which does not lead to significant sulfate production is required. Such a loss could include an increase of the calculated S02 dry deposition rate over the land, or an additional deposition We have estimated that about 16%, 23%, and 11% of the anthropogenically emitted sulfur in North America, Europe, and eastern Asia are transported out to the neighboring oceans. This anthropogenic outflow has a large impact on the sulfur budget over the oceans, especially in the North Atlantic and northwestern Pacific regions. Our budget calculations indicate that about 60% of total SO2 over these two regions is of anthropogenic origin. Ship emission is estimated to contribute 65• of SO2 at the surface of the North Atlantic; but the impact of ship emissions is concentrated mostly near the surface. Within the atmospheric column over the North Atlantic, DMS oxidation is 50% more important than ship emissions as the SO2 source. In contrast to the polluted North Atlantic and northwestern Pacific, the predominant SO2 source in the southwestern and tropical eastern Pacific is from DMS oxidation.
We have estimated that the production efficiency of SO2 from DMS oxidation is 0.87 to 0.91 in most oceanic regions.
These values are higher than those estimated in some 0-D model studies, a discrepancy which is likely due to the transport or loss parameters assumed in the 0-D models.
The sulfate production efficiency from SO2 oxidation is much lower over the polluted continents than over the oceans. With about half of the SO2 emitted over the land being deposited to the surface, only 24% to 33% of them are oxidized to produce sulfate over the land. By contrast, 53% to 66% of the SO2 over the oceans converts to sulfate in the atmosphere. This is because of a larger free tropospheric fraction of column SO2 over the oceans than over the land, hence less surface loss.
The lifetimes of SO2 and sulfate over the oceans are nearly twice as long as those over the land.
The model-simulated sulfur concentrations are generally in agreement with the observations under a variety of conditions and on different spatialscales and timescales. However, the parameters used in the model, such as wet scavenging, DMS emission, and chemical mechanisms, are highly simplified mainly due to the lack of better knowledge or resource, and some potentially important pathways, such as SO2 loss in seasalt aerosols, are not included in the current model. Therefore large uncertainties still exist in quantifying each process involved in the atmospheric sulfur cycle.
Nonetheless, the detailed evaluation of the GOCART model provides a solid base for investigating the processes that control the sulfur distributions in the atmosphere, analyzing the relationship that exists between the sulfate aerosol and its precursors, and estimating the forcing that sulfate aerosol exerts on global climate.