**4. Conclusions**

The Japanese air quality model intercomparison study, J-STREAM, has shown that although SO42– is generally well-captured by the models, concentrations of SO42– were underestimated during winter. In previous studies, the modeled SO42– concentrations were revised by focusing on the Feand Mn-catalyzed oxidation pathways and highlighting the importance of developing an emission inventory for trace metals over Asia in the first phase [3]. We further increased production of SO42– by the addition of an aqueous-phase NO2 oxidation pathway in the second phase [4]. To further improve the modeling performance of gas-phase oxidation, three SCI were incorporated in the third phase of J-STREAM and the KMT option available in the latest CMAQ model was examined. A difference between the KMT and the base-case simulation was found over the Sea of Okhotsk, and the absolute differences in SO42– concentrations were less than −0.1 μg/m3, giving essentially similar results with a test case for KMT over the U.S.A. Most previous studies have treated SCI in bulk; however, in this study, three SCI were treated separately with each SCI and incrementally tested and where the dependency of the rate constant of SCI1 with H2O was also examined by performing sensitivity simulations with a high and low rate constant. It was found that only when the lower rate constant for the SCI1 + H2O reaction was used that the production of SO42– was increased by SCI1, and the importance of the value of the rate constant of SCI1 with H2O for the Asian region was highlighted. This finding was consistent with a previous study conducted for the U.S.A. It was further demonstrated in the present study that the key role of SCI3 is to increase SO42– production because this reaction is independent of the rate constant of SCI1 with H2O. It was established that the explicit treatment of each SCI is required to enable clarification of the role of SCI on SO42– production.

In addition to the investigation of model performance, the third phase of J-STREAM included an intercomparison study on source sensitivities. Four major domestic sources (transportation, stationary combustion, fugitive VOC, and agricultural NH3) were investigated as source groups. The source sensitivities were estimated based on the traditional sensitivity simulation approach whereby a 20% emission reduction was calculated, the result of which was subtracted from the base-case simulation. It was clarified that the winter haze episode at the Tokyo site was generally dominated by emission sources from outside Japan, and the haze was enhanced by the domestic emission sources of transportation and fuel combustion. The estimations of source contributions were nearly the same between the base-case CMAQ simulation and KMT. With the chemistry updates involving the aqueous-phase Fe- and Mn-catalyzed oxidation reactions and NO2 oxidation, it was found that these revisions led to an increase in transboundary impacts. In the case of the chemistry updates with the inclusion of SCI, it was shown that the change to fugitive VOC emissions could impact SO42– concentrations by influencing O3 which in turn influences SCI.

As a result of conducting the first, second and third phases of J-STREAM, we have successfully demonstrated a means to enhance simulated SO42– production during the winter when underestimations of SO42– concentrations have been problematic. Given recent drastic reductions in SO2 emissions from China, further declines in SO42– can be expected, and reactive nitrogen will continue to play an important role in this process due to the abundance of freely available NH3. Because of the difficulty of producing reliable simulation models for reactive nitrogen species because of their semi-volatile nature, it is first necessary to establish accurate simulations for SO42–. The means to enhance SO42– production has been demonstrated for a single winter haze episode, and further tests on other haze episodes should be performed. Furthermore, the incorporation of SCI in this study suggests sensitivity to fugitive VOC sources that do not include direct SO2 emissions but can change O3 concentrations, and this effect should be tested in other seasons.

**Author Contributions:** S.I. coded the additional gas- and aqueous-phase reactions into the CMAQ model, conducted the model simulations and the comparison of models with observations, and wrote the manuscript; K.Y. is the sub-leader of the model intercomparison and prepared the meteorological inputs and initial and boundary conditions; H.H. is the sub-leader of the inorganic aerosol measurements and conducted the ACSA observations at Mukoujima, Tokyo; S.C. is the leader of the J-STREAM project and prepared the emission inputs and discussed the model intercomparison results.

**Funding:** This research was funded by the Environment Research and Technology Development Fund (5-1601) of the Environmental Restoration and Conservation Agency.

**Acknowledgments:** This research was supported by the Environment Research and Technology Development Fund (5-1601) of the Environmental Restoration and Conservation Agency.

**Conflicts of Interest:** The authors declare no conflict of interest.
