An Advanced Modeling Study on the Impacts and Atmospheric Implications of Multiphase Dimethyl Sulfide Chemistry

An advanced modeling study on the impacts and atmospheric implications of multiphase dimethyl sulfide chemistry

Erik Hans Hoffmanna, Andreas Tilgnera, Roland Schrödnerb,1, Peter Bräuera,2, Ralf Wolkeb, and Hartmut Herrmanna,3

aAtmospheric Chemistry Department (ACD), Leibniz Institute for Tropospheric Research (TROPOS), D-04318 Leipzig, Germany; and bModeling of Atmospheric Processes Department (MAPD), Leibniz Institute for Tropospheric Research (TROPOS), D-04318 Leipzig, Germany

Edited by Barbara J. Finlayson-Pitts, University of California, Irvine, CA, and approved August 12, 2016 (received for review April 20, 2016)

Oceans dominate emissions of dimethyl sulfide (DMS), the major gas phase, SO2 can be oxidized to gaseous H2SO4,whichmay natural sulfur source. DMS is important for the formation of non-sea condense on existing particles or contribute to new particle for- 2− salt sulfate (nss-SO4 ) aerosols and secondary particulate matter mation in low-condensation sink environments. Other loss pro- over oceans and thus, significantly influence global climate. The cesses for SO2 include dry deposition or aqueous-phase reactions. mechanism of DMS oxidation has accordingly been investigated in MSA mainly condenses on existing particles contributing to aerosol several different model studies in the past. However, these studies particle mass, and it is mainly removed by dry and wet deposition. had restricted oxidation mechanisms that mostly underrepresented The applied parameterizations usually consider only gas-phase re- important aqueous-phase chemical processes. These neglected but actions and should be treated with caution, because the mechanism highly effective processes strongly impact direct product yields of of DMS oxidation to SO2 and MSA is not completely understood DMS oxidation, thereby affecting the climatic influence of aerosols. (14). Various model studies have found significantly varying yields To address these shortfalls, an extensive multiphase DMS chemistry of DMS oxidation products (15). For example, the study by von mechanism, the Chemical Aqueous Phase Radical Mechanism DMS Glasow and Crutzen (5) applied different mechanistic assumptions Module 1.0, was developed and used in detailed model investiga- and calculated conversion efficiencies of DMS into SO2 of between tions of multiphase DMS chemistry in the marine boundary layer. 0.14 and 0.95 in the marine boundary layer (MBL). Predictions of The performed model studies confirmed the importance of aqueous- sulfate aerosol formation caused by DMS oxidation and their cli- phase chemistry for the fate of DMS and its oxidation products. mate impact calculated by CTMs and GCMs are, therefore, highly Aqueous-phase processes significantly reduce the yield of sulfur di- uncertain. The formation of other stable DMS oxidation products, oxide and increase that of methyl sulfonic acid (MSA), which is needed mainly MSA, lowers the SO2 yield and constitutes the main source to close the gap between modeled and measured MSA concentra- of uncertainty to aerosol formation, because MSA production tions. Finally, the simulations imply that multiphase DMS oxidation predominantly leads to growth of existing particles and suppresses produces equal amounts of MSA and sulfate, a result that has sig- new particle formation (16). Recent studies suggest that MSA can 2− significantly assist cluster formation between H2SO4 and amines nificant implications for nss-SO4 aerosol formation, cloud condensation nuclei concentration, and cloud albedo over oceans. Our and thereby, contribute to new particle formation, although it is a less potent clustering agent than H SO (17, 18). Hence, a findings show the deficiencies of parameterizations currently 2 4 used in higher-scale models, which only treat gas-phase chemistry. Overall, this study shows that treatment of DMS chemistry in Significance both gas and aqueous phases is essential to improve the accuracy of model predictions. Climate models indicate the importance of dimethyl sulfide (DMS) oxidation in new aerosol particle formation and the marine multiphase chemistry | dimethyl sulfide | multiphase modeling | activation of cloud condensation nuclei over oceans. These CAPRAM | marine aerosols effects contribute to strong natural negative radiative forcing and substantially influence the Earth’s climate. However, the − DMS oxidation pathway is not well-represented, because ear- aseous sulfuric acid (H2SO4) and aqueous sulfate (HSO4 , 2− lier model studies only parameterized gas-phase DMS oxida- GSO4 ) contribute to the formation of new aerosol particles as well as secondary particulate matter and are, thus, important tion and neglected multiphase chemistry. Here, we performed for human health and the Earth’s climate (1). Globally, anthro- the most comprehensive current mechanistic studies on multiphase DMS oxidation. The studies imply that neglecting mul- pogenic sulfur emissions in the form of sulfur dioxide (SO2)dominate atmospheric production of gaseous H SO and particle-phase tiphase chemistry leads to significant overestimation of SO2 2 4 production and subsequent new particle formation. These sulfate. However, the main natural source of sulfur is the oxidation findings show that an advanced treatment of multiphase DMS of dimethyl sulfide (DMS) emitted by oceans (2), which is the most − chemistry is necessary to improve marine atmospheric chem- important precursor for non-sea salt sulfate (nss-SO 2 )aerosols 4 istry and climate model predictions. over the open ocean (3). Sulfate aerosols strongly influence the

climate both by direct negative radiative forcing (4) and as a Author contributions: E.H.H., A.T., R.S., P.B., R.W., and H.H. designed research; E.H.H., A.T., dominant source of cloud condensation nuclei (CCN) over the R.S., P.B., R.W., and H.H. performed research; E.H.H., A.T., and H.H. analyzed data; and E.H.H., open ocean (5). Because oceans cover about 70% of Earth’s A.T., and H.H. wrote the paper. surface (6) and have generally low albedo, DMS oxidation plays a The authors declare no conflict of interest. major role in influencing the natural radiative forcing of sulfate This article is a PNAS Direct Submission. aerosols as well as cloud properties (3). 1Present address: Centre for Environmental and Climate Research, Lund University, Investigations of the effect of DMS oxidation on natural sulfate SE-22362 Lund, Sweden. aerosol concentrations and cloud and aerosol properties require an 2Present address: Wolfson Atmospheric Chemistry Laboratories, Department of Chemistry, accurate, reduced DMS oxidation scheme in chemical transport University of York, York YO10 5DD, United Kingdom. models (CTMs) and global climate models (GCMs). Current pa- 3To whom correspondence should be addressed. Email: [email protected]. rameterizations use fixed yields of SO2 and methyl sulfonic acid This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10. 2− (MSA) to calculate new nss-SO4 aerosol formation (7–13). In the 1073/pnas.1606320113/-/DCSupplemental.

11776–11781 | PNAS | October 18, 2016 | vol. 113 | no. 42 www.pnas.org/cgi/doi/10.1073/pnas.1606320113 Downloaded by guest on September 30, 2021 better estimate of MSA and SO2 yields is necessary for improved chemical regimes (cloud-free and cloudy MBL). Finally, implica- 2− nss-SO4 aerosol predictions in CTMs and GCMs. tions of this study on climate model predictions are discussed. More detailed mechanistic studies of the DMS oxidation mechanism are needed to overcome these uncertainties. In Results and Discussion particular, model studies that exclude multiphase chemistry and Model Simulations. Studies with the box model SPACCIM (27) were only treat gas-phase DMS chemistry do not reproduce observed performed to investigate multiphase DMS chemistry in the MBL. MSA aerosol concentrations (16). In this context, it is important The chemical mechanism applied combines the MCMv3.2 (25, 26) to note that altocumulus and altostratus clouds cover more than for the description of the gas phase, the CAPRAM4.0α for the de- 20% and that stratus and stratocumulus clouds cover nearly 30% scription of the aqueous phase, and marine chemistry described by of the oceans. The residence time of an air parcel in these clouds the multiphase mechanism HM2.1 and the newly developed is between 3 and 4 h (19). The review of Barnes et al. (14) CAPRAM DM1.0. The HM2.1 was adapted from the CAPRAM concluded that multiphase DMS chemistry must also be taken Halogen Module 2.0 (24) for the use with the MCMv3.2. Mul- into account to determine the oxidation of both DMS and its tiphase DMS chemistry in the MBL under pristine ocean condi- oxidation products. Early investigations of multiphase DMS tions was investigated based on the model scenario from the work chemistry were limited to 10 (20) and 7 (5) aqueous-phase re- by Bräuer et al. (24). This scenario was extended for multiphase actions. However, kinetic studies have revealed many other sig- DMS chemistry by implementing the DMS emission rate from the nificant reactions of DMS and its oxidation products in the work by Lana et al. (28) as well as new initial concentrations and aqueous phase (14). Zhu et al. (21) applied a trajectory ensemble deposition rates for DMS oxidation products (Materials and Methods model to show the importance of aqueous-phase DMS chemistry and SI Appendix). Other than the base run (termed as full), different in cloud droplets. Their multiphase mechanism was restricted to sensitivity runs were performed to study the influence of various nine DMS aqueous-phase reactions and neglected the contri- chemical subsystems on DMS product yields. All runs had identical butions of other important chemical subsystems, especially hal- meteorological conditions. An outline of the different sensitivity ogen chemistry and aqueous-phase chemistry in deliquesced runs is given in Table 1. particles. Such small aqueous-phase mechanisms in box or local The DM1.0 was developed by an extensive literature study using model studies exclude important DMS aqueous-phase reaction the most recent kinetic and mechanistic data as well as the state of pathways. Despite the above-mentioned call by Barnes et al. (14) the art mechanism MCMv3.2. It contains 103 gas-phase reactions, to incorporate more detailed multiphase DMS schemes in nu- five phase transfers, and 54 aqueous-phase reactions and is merical models, no additional progress in DMS multiphase chem- designed for varying atmospheric conditions ranging from a istry modeling has been made. As a result, our understanding of clean, remote marine atmosphere to polluted coastal conditions. DMS oxidation to MSA and SO2 and its contribution to aerosol production and mass is up to now limited. For example, measured Multiphase DMS Oxidation. A key issue in DMS chemistry is the MSA concentrations can only be reproduced using simplified model contribution of different oxidants to its oxidation in both the gas and assumptions neglecting important reaction pathways (22). Further- aqueous phases. The most important reaction pathways (those that more, Berresheim et al. (23) indicated that modeled gaseous H2SO4 contribute at least 5% to the total sink flux of multiphase DMS to measured concentrations does not agree with field measure- chemistry) including the mean reaction fluxes of each pathway over ments in coastal regions. These authors concluded that a better the whole simulation are presented for the full run in Fig. 1. It in- understanding of sulfur trioxide (SO3) formation from DMS, an cludes the main chemical processes in the gas and aqueous phases intermediate that rapidly forms H2SO4 with water vapor, could help as well as phase transfer interactions. The percentage contributions to solve this gap (23). Overall, a better mechanistic understanding outlined below generally correspond to the averages over the whole of DMS oxidation to SO2,H2SO4, and other stable oxidation simulation time. products, especially MSA, is needed to improve model predictions In the gas phase, oxidation of DMS occurs by either an on DMS-related climate impacts. H-abstraction or an addition pathway. The abstraction pathway Therefore, this study investigates the role of multiphase DMS leads to a peroxyl radical in the first generation, whereas the chemistry in the MBL of the open ocean. For this purpose, a addition pathway mainly results in DMSO formation. The ad- comprehensive multiphase DMS chemistry mechanism, the CAP- dition pathway dominates over the abstraction pathway, mainly RAM DMS Module 1.0 (DM1.0), was developed and coupled to because of the reaction with BrO, which accounts for 46% of the multiphase chemistry mechanism Master Chemical Mechanism, DMS oxidation averaged over the whole model run. The re- version 3.2 (MCMv3.2)/Chemical Aqueous Phase Radical Mecha- maining gas-phase oxidation is mostly triggered by reaction with nism 4.0α (CAPRAM4.0α) + CAPRAM Halogen Module 2.1 the chlorine atom (Cl; 3% by addition and 15% by H-abstraction) (HM2.1) (24–26). An open ocean scenario was simulated using the and the hydroxyl radical (OH; 9% by addition and 11% by Spectral Aerosol Cloud Chemistry Interaction Model (SPACCIM) H-abstraction). These results show the importance of DMS oxida- (27). To study the differences between multiphase DMS chemistry tion by halogens, which are mainly ignored in current parameteri- in clouds and deliquesced particles, a model scenario with non- zations of CTMs and GCMs. Neglecting this chemistry will lead to

permanent clouds is applied. Compared with studies considering inaccurate predictions of the formation of new aerosol particles, the EARTH, ATMOSPHERIC, either cloud or noncloud conditions, the use of a nonpermanent activation of CCN, and natural radiative forcing (5, 29). AND PLANETARY SCIENCES cloud scenario allows more realistic simulations and the study of Clouds significantly lower the gas-phase concentrations of OH, processed aerosols. The net effect of including aqueous-phase Cl, and BrO (24, 30) and modify the prevailing chemistry, such chemistry on the conversion of DMS into SO2 and the formation of that aqueous-phase reactions take over and dominate the oxida- rather stable particulate products is evaluated in these two different tion of DMS and its oxidation products. During cloud periods,

Table 1. Overview of the sensitivity runs Model run Specification

full MCMv3.2, CAPRAM4.0α, HM2.1, and DM1.0 woIodine MCMv3.2, CAPRAM4.0α, DM1.0, and HM2.1 without iodine chemistry woHM2 MCMv3.2, CAPRAM4.0α, and DM1.0

O3 MCMv3.2, CAPRAM4.0α, HM2.1, and DM1.0; high initial concentrations of NO2,O3, and HNO3 woCloud MCMv3.2, CAPRAM4.0α, HM2.1, and DM1.0; without cloud periods woAqua MCMv3.2, CAPRAM4.0α, HM2.1, and DM1.0; without aqueous-phase chemistry of DMS and its oxidation products

Hoffmann et al. PNAS | October 18, 2016 | vol. 113 | no. 42 | 11777 Downloaded by guest on September 30, 2021 Addition H-Abstraction OH, Cl S HO O OH OH DMS 2 S O O2 3CH CH2 CH CH S S 3 2 O CH SOH 3CH CH3 3CH CH3 54.6 O 3 38.1 43.0 131.1 O OH CH3 NO,

4.8 RO OH 2 hν 35.7 78.8 S 42.3 CH OH 3 OH 2.5 BrO DMS S 4.0 S 3CH CH CH CH S 3 3 3CH CH2

38.1 OH, O 233.0 O hν

O3,aq

O 81.4 27.6

35.3 3 S CH O 3CH S DMSO 3 64.1 O 73.6 DMSO O O2 S O3 3CH CH3 S

CH CH 55.6

3 3 67.6 OH 150∙102 OH aq 23.4 S O 100 SO 162.0 2 3CH O 127.0 90 MSIA 80 156.0 MSIA 7.5

O 32.0 O 70 S 3CH OH S 60 3CH OH CH 82.4 3 50 CH3 S SO OH , OO 2 S aq 40 13.4 OO - O Cl2 - 2 CH3SO2 30 73.3 9.4 O 20 - 3,aq CH3SO2 CH O3 3 CH3 119.5 10

SO3 64.1 OOS OOS 73.3 CH SO - 25.2 7.5 O O 3 2 - O Cl2 5.0 0.1 73.3 9.4 <2.5 CH SO - O- 3 2 MSA MSA + H O SO O CH OH CH CH HO2 CH 2,aq 2 aq 3 3 3 21.9 - SO3 CH OOS OOS OOS OOS SO3 4.2 2 O 4.4 4.4 4.6 4.3 O- OH OH O O

− − Fig. 1. Depiction of mean multiphase source and sink fluxes (in 102 molecules cm 3 s 1) of the full scenario run throughout the whole model. The temperature is 280 K during noncloud conditions and 276.5 K in clouds. Only oxidation fluxes exceeding 5% of the total flux are included. The blue area highlights aqueous-phase reactions, and stable compounds are shaded yellow. The width and color of arrows indicate the magnitude of the mass flux, and dashed arrows represent phase transfer processes. Numbers above and below the arrows give the mean fluxes, and chemical species at the beginning of the small curved arrows below and above the thicker arrows describe the main reactants.

78% of DMS (in all phases) is oxidized in the aqueous phase by sulfinic acid (MSIA), and MSA. For the sake of completeness, the ozone (O3)toDMSO(SI Appendix, Multiphase Chemistry of DMS modeled evolution of the concentration–time profiles of DMS, and Its Oxidation Products, and Figs. S4 and S5). This reaction DMSO, MSIA, and MSA along with comparisons with field mea- contributes 7% to multiphase DMS oxidation over the whole surements and mass flux analyses are discussed in detail in SI model run. Although consistent with the previous simulations by Appendix, Multiphase Chemistry of DMS and Its Oxidation Products. von Glasow and Crutzen (5), this finding is unexpected, because less than 0.01% of DMS partitions to the aqueous phase (Fig. 2). Processing of the Main DMS Oxidation Products. The additional The contribution is explained by the very fast kinetics of the O3 conversion of DMS oxidation products must also be considered to reaction, so that a large flux is established because of the reactive determine the effect of the overall DMS oxidation on natural uptake of DMS. During in-cloud periods, the overall rate of DMS radiative forcing and thus, Earth’s climate. In this section, we first degradation decreases, with the result that the gas-phase DMS consider the additional processing of H-abstraction products and concentration increases slightly. then, examine the reactions of DMS products of the addition In conclusion, this study has shown that halogens contribute pathway (Fig. 1). significantly to DMS oxidation and that, during cloud periods, the The main product of the H-abstraction pathway is SO2, whereas aqueous-phase oxidation of DMS by O3 is dominant, whereas the only small amounts of MSA and SO3 are formed. Furthermore, overall degradation decreases. An overview of the fractional con- accumulation of reservoir gases, especially S-methyl thioformate tributions of different oxidants in the DM1.0 to DMS oxidation in (CH3SCHO), may reduce the conversion of DMS to SO2 because the gas and aqueous phases is given in SI Appendix,TableS13along of CH3SCHO deposition to the ocean. In principle, the under- with fractional contributions to the oxidation of DMSO, methyl standing is that SO2 forms gaseous H2SO4 from OH oxidation.

11778 | www.pnas.org/cgi/doi/10.1073/pnas.1606320113 Hoffmann et al. Downloaded by guest on September 30, 2021 The simulation results imply that DMSO2 formed through the 1 OH-addition pathway or DMSO oxidation is only a minor product of DMS oxidation in a marine environment. Therefore, DMSO2 oxidation is not further considered in this study. 10-2 MSIA is the major DMSO oxidation product, and about 10–40% of MSIA partitions into marine aerosols. MSIA has a diurnal profile, showing stronger partitioning at night (gray shaded bars in 10-4 Fig. 2). The high solubility and formation in the aqueous phase increase the importance of aqueous-phase chemistry for the MSIA -6 fate. MSIA is almost completely degraded during daytime in the 10 aqueous phase. Oxidation is decreased during night, explaining the diurnal partitioning profile in Fig. 2. The main oxidants are dis-

Aqueous phase fraction − solved O ,OH,andCl . The strongest sink flux is the reaction of 10-8 3 2 − MSIA and dissociated methyl sulfinic acid (MSI )withO3 (42%) to form MSA. The DM1.0 is the first mechanism that has imple- − − 10-10 mented these reactions. Cl2 and OH react mainly with MSI . 42 28 32 63 40 44 84 These two oxidations have to be differentiated between oxidation in − Model time / hours deliquesced particles and cloud droplets. The oxidation by Cl2 is DMS DMSO DMSO2 MSIA MSA dominant in deliquesced particles, whereas oxidation by OH dominates in cloud droplets (SI Appendix,TableS13). These results are Fig. 2. Partitioning of DMS, DMSO, DMSO2, MSIA, and MSA into the aqueous clearly different from the assumption in the work by Zhu et al. (21), phase of deliquesced particles and cloud droplets throughout the second − in which the cloud droplet Cl2 concentration was kept constant. model day. The y axis is logarithmic. Gray shaded bars denote the night pe- − The total contributions of Cl2 and OH to aqueous-phase MSIA riods, and blue bars denote the cloud periods. oxidation are 10% and 19%, respectively. As explained in SI Appendix, Model and Mechanism Description, both reactions occur

• via an electron transfer reaction leading to high amounts of However, our results indicate that the thermal decay of CH3SO3 • − CH3SO2(O2 ), which reacts further with MSI to form dissolved into SO3 represents the main contributor to gaseous H2SO4 over MSA and CH SO • (31). CH SO • mainly decomposes to CH and • 3 3 3 3 − 3 the open ocean. Gaseous CH3SO3 is predominantly formed from SO , but small reaction fluxes with MSI occur, also forming MSA. • 3 − − gaseous CH3SO2 , which is mainly generated by the isomerization of • As a result, the oxidation of MSI by Cl2 and OH leads to a net gaseous CH3S(OO ) (Fig. 1). In the atmosphere, SO3 reacts rapidly destruction of at least two MSIA molecules. Overall, the reactions − with water vapor to form gaseous H2SO4. In total, 94% of the of MSI with CH SO (O •) and CH SO • account for 29% of • 3 2 2 3 3 − gaseous H2SO4 is formed because of thermal decay of CH3SO3 . MSIA depletion. Because of these reactions, MSI oxidation by • − Consequently, the isomerization of gaseous CH3S(OO )isvery Cl and OH produces dissolved MSA and SO in nearly equal 2 − 3 important in determining gaseous concentrations of H2SO4 (addi- SI Appendix amounts and oxidizes more MSIA/MSI than the O3 reactions. tional details are in ). However, only 4% of DMS is In total, 58% of the modeled MSA is produced by the aqueous- − converted into gaseous SO3, whereas 24% of DMS is converted into phase reactions of MSIA/MSI with O , and 40% is caused by the − 3 SO2 (Table 2). The strongly reduced importance of SO2 to gaseous aqueous-phase reactions of MSI with its oxidation products. In H2SO4 formation results from the presence of noon clouds, which contrast, the gas-phase formation only accounts for 2% of MSA. enhance uptake processes and significantly reduce both the OH and However, the gas-phase MSA as well as gaseous H2SO4 pro- SO2 concentrations, thereby suppressing gaseous production of duction depend strongly on the rate constant of the thermal decay H SO . Daytime clouds, therefore, greatly reduce the significance • 2 4 of CH3SO3 , which is very uncertain (details are in SI Appendix, of SO2 as an intermediate species in the formation of gaseous Model and Mechanism Description). Hence, additional laboratory H2SO4. Still, even in the woCloud run, in which no clouds are investigations of this thermal decay are warranted. MSA remains implemented, 80% of H2SO4 is formed via the thermal de- • almost entirely in the aqueous phase throughout the whole model composition of CH3SO3 . This result suggests that, other than the • run (Fig. 2). Our results, therefore, indicate that only aqueous- gas-phase oxidation of SO2, the thermal decomposition of CH3SO3 phase oxidation of MSA is important. Because of its high stability, is a main contributor to new particle formation in the MBL in the − • only 2% of MSA is oxidized further by Cl2 and dissolved OH. presence of clouds. Hence, the thermal decomposition of CH3SO3 Hence, MSA sinks do not compete with its production, and MSA may explain the gap between modeled and measured gaseous accumulates in the aerosol phase. Because the tropospheric life- H2SO4 concentrations in coastal regions as it has been proposed time of marine aerosols is between 1 and 7 d (32), wet and dry before (23). deposition are likely the dominant tropospheric sinks of MSA. DMSO, formed via the oxidation of DMS by OH and BrO, is a Overall, modeling of DMS oxidation shows that the H-abstraction key intermediate of the addition pathway. The subsequent oxidation pathway in the gas phase leads predominantly to SO2 and that the EARTH, ATMOSPHERIC, • of DMSO strongly alters new aerosol formation in the MBL (5), thermal decay of gaseous CH3SO3 is the main contributor to AND PLANETARY SCIENCES because it reduces the yield of SO2 from DMS oxidation by forming gaseous H2SO4 formation, whereas the addition pathway leads to additional stable oxidation products, like dimethyl sulfone (DMSO2) MSA involving multiphase processes. Furthermore, DMS oxida- andMSA.DMSOismostlyoxidizedbyOHtoMSIAinboththegas tion leads to MSA and sulfate production in nearly equal amounts. and aqueous phases. Oxidation by halogens is of minor importance. The large fraction of MSA formed through aqueous-phase pro- Fig. 2 presents the partitioning of DMSO, DMSO2,MSIA,andMSA cesses is different from parameterizations of the DMS addition in deliquesced particles and cloud droplets. About 1% and 2% of pathway currently applied in GCMs mostly implemented after the DMSO and DMSO2, respectively, partition into deliquesced parti- work by Chin et al. (8). cles. Because of the less effective partitioning of DMSO into aerosols

(Fig. 2) compared with the more oxidized products MSIA and MSA, SO2 and MSA Yields and Atmospheric Implications. The contribution the aqueous-phase oxidation mainly occurs in cloud droplets. It is of DMS oxidation to natural radiative forcing is strongly influenced also noteworthy that, although the production of DMSO is sup- by the yields of SO2 and MSA. Fig. 1 shows the importance of both pressed in cloud droplets, its oxidation is increased (SI Appendix,Fig. gas- and aqueous-phase reactions in determining product yields. S7 and Table S13). This effect is explained by the partitioning of Multiphase chemistry causes a shift in the product distribution of DMSO. With the onset of cloud formation, DMSO and DMSO2 DMS chemistry from SO2 toward MSA. This result highlights the partition almost completely into cloud droplets because of their high deficiencies of current parameterizations in climate models. Apart Henry’s Law coefficients (14). from multiphase chemistry in general, the results presented in Fig. 1

Hoffmann et al. PNAS | October 18, 2016 | vol. 113 | no. 42 | 11779 Downloaded by guest on September 30, 2021 Table 2. Summary of effective conversion yields of DMS into (i)SO2 and (ii) MSA, and (iii) conversion of MSA to S(VI) in the gas and aqueous phases for all sensitivity runs

Sensitivity run DMS→SO2 (%) DMS→MSA (%) MSA→S(VI) (%)

full 23.6 41.0 2.2 O3 56.5 14.9 34.4 woIodine 20.5 46.9 2.5 woHM2 30.7 29.2 3.9 woCloud 32.7 30.3 0.8 woAqua 60.3 0.9 0.1

also highlight specific chemical subsystems (e.g., halogen chemistry Conclusions or chemistry in cloud droplets) that strongly influence DMS con- Atmospheric multiphase chemistry simulations of a pristine ocean version yields. Different sensitivity runs (Table 1) were performed scenario show that the conversion of DMS to MSA is strongly to investigate the influence of these chemical subsystems on DMS underestimated when DMS aqueous-phase chemistry is omitted conversion. Table 2 shows the results of the sensitivity runs for (i) and correspondingly, that formation of SO is overestimated. ii 2 the total yield of SO2,( ) the yield of total MSA from DMS, and Halogen–DMS interactions are also essential for the conversion iii ( ) the conversion of MSA into sulfur VI [S(VI)]. High background efficiency of DMS to SO2 and MSA. The simulations show that, concentrations of O3 and NO2 enhance the production of SO2 and despite the low solubility of DMS, marine clouds can have a major decrease the conversion into MSA. Under these conditions, oxi- influence on DMS oxidation product yields and formation and 2− dation of MSA increases strongly. A similar, albeit weaker, effect growth of new nss-SO4 aerosols. Overall, halogen–DMS interac- appears when halogen chemistry is limited. The decreased conver- tions and DMS aqueous-phase chemistry have a strong impact on: sion to MSA in the woHM2 run arises from the orders of magni- (i) the conversion of DMS into SO2 and MSA, (ii) the aging of 2− tude lower gas-phase concentrations of BrO and Cl owing to the marine aerosols, (iii) the production of nss-SO4 aerosols, and restricted halogen chemistry schemes in the MCMv3.2 and (iv) the radiative properties of marine clouds and aerosols. Hence, the CAPRAM4.0α. The higher MSA oxidation is caused by the neglecting multiphase DMS chemistry in CTMs and GCMs greatly slightly higher OH concentration. The higher OH concentration increases the uncertainties of model predictions. Implementation of arises from reduced ozone depletion by halogens, which enhance a near-explicit multiphase DMS mechanism, such as the DM1.0, is the photolysis of O3 and therefore, OH production. Thus, reactive computationally expensive. Therefore, future work will focus on the Cl and bromine species strongly affect DMS oxidation and conver- reduction of the DM1.0 for application in CTMs and parameteri- sion efficiency into SO2 and MSA. However, knowledge of tropo- zation development for application in GCMs. spheric halogen chemistry is still quite restricted and needs additional investigation as Simpson et al. (33) have outlined in detail. Materials and Methods In a cloud-free MBL, the MSA yield decreases, and SO2 and Model Description. Marine multiphase chemistry simulations were carried out with MSA are produced in equal amounts. This effect results from the the box model SPACCIM [details are in the work by Wolke et al. (27)]. SPACCIM higher production of Cl and BrO in a cloud-free MBL, because allows the complex chemical processing in clouds and deliquesced particles to be clouds strongly suppress halogen activation. Cl activation in par- investigated by combining a detailed microphysical model with a fine-resolved ticular is reduced during cloud periods (24). The H-abstraction particle/droplet spectrum and a complex multiphase chemistry model. The applied pathway is more important under these circumstances. Overall, chemical mechanism combined the MCMv3.2, the CAPRAM4.0α, the HM2.1, and marine clouds shift the DMS oxidation toward a higher production of MSA. Neglecting aqueous-phase DMS chemistry (as in most GCMs) leads to an overestimation of SO2 formation and an underestimation of MSA production. The majority of MSA is AB formed and resides in the aqueous phase. This is an important result, because the described aqueous-phase chemistry does not lead to the formation of new particles but rather, results in an increase of aerosol mass of already existing particles. This crucial finding implies that not considering multiphase DMS chemistry in SO 0.60 0.01 MSA SO 0.24 0.41 MSA CTMs and GCMs will strongly affect the modeled number con- 2 2 centration and mass of marine aerosols and thereby, modify radiative forcing and climate predictions. The supposed influences are schematically depicted in Fig. 3. Ignoring multiphase DMS 2− chemistry could lead to an overestimation of the nss-SO4 aerosol concentration and the number of CCNs. Accordingly, predicted cloud albedos would be higher under these conditions, and GCMs would strongly overestimate the negative natural radiative forcing DMS emission (Tg sulfur yr-1) by clouds. 17.6 - 34.4 This study shows that the role of DMS on Earth’s climate is still not well-understood, despite many global model studies. Our simulations suggest that enhanced DMS emission results in higher particulate mass but not necessarily to appreciably higher aerosol Fig. 3. Schematic depiction of the multiphase DMS conversion efficiency to number concentrations. The model findings are also relevant to SO2 and MSA for (A) the sensitivity run without aqueous-phase DMS chemistry and (B) the model run including aqueous-phase DMS chemistry. The supposed geoengineering concepts based on aerosol production via DMS changes to cloud albedo and related radiation effects are also shown. The blue generated by ocean fertilization. Previous calculations on the effi- circle in B represents aqueous-phase chemistry in deliquesced particles and ciency of this approach often assumed a total DMS to SO2 yield of cloud droplets by the simulation. The red plus and minus signs represent

unity (34). The results of advanced multiphase modeling of DMS changes of the DMS conversion yields into SO2 and MSA between the two oxidation indicate that SO2 production is likely to be greatly sup- runs. The numbers represent the modeled yields of SO2 and MSA. The yellow pressed and that changes in cloud albedo will likely be weaker arrows represent the supposed effects on incoming and reflected solar radi- than expected. ation. Tg, teragram.

11780 | www.pnas.org/cgi/doi/10.1073/pnas.1606320113 Hoffmann et al. Downloaded by guest on September 30, 2021 the DM1.0. In total, 10,212 species and 21,928 reactions were included. The sim- by 23 reactions of reactive halogen compounds and 14 reactions of non- ulationsstartedonJune19at45°Nandranfor108hwithanassumedtem- halogen compounds. Overall, the gas-phase mechanism scheme of the DM1.0 perature of 280 K. In the model scenario, an air parcel is moved along a predefined contains 103 reactions, including nine photolysis processes. trajectory starting at 850 hPa. The scenario includes eight cloud passages, which The SPACCIM calculates phase transfer processes according to the approach

are achieved by lifting the air parcel from 850 to 800 hPa for about 2 h at noon by Schwartz (37) by considering the gas-phase diffusion coefficient Dg,the and midnight, respectively. Because of the adiabatic cooling as the air parcel rises, mass accommodation coefficient α,andtheHenry’s Law coefficient HA.These the temperature falls to about 276.5 K. The in-cloud residence time of about 2 h parameters were implemented in the DM1.0 for uptake of DMS, DMSO,

was chosen based on the calculations by Pruppacher and Jaenicke (35). Noon and DMSO2, MSIA, and MSA. midnight clouds were chosen to study differences in nighttime and daytime clouds The aqueous-phase mechanism of the DM1.0 contains many more interme-

on multiphase chemistry. diate steps and reaction pathways of DMS, DMSO, DMSO2, MSIA, and MSA than any other DMS aqueous-phase mechanism. Rate constants for these reactions Mechanism Development. Based on the gas-phase reactions considered in the were chosen from recent measurements. No literature data were available for MCMv3.2, an extended multiphase DMS chemistry mechanism, the DM1.0, some types of reactions (such as oxygen addition at carbon-centered radicals), has been developed in this work. For this purpose, an extensive aqueous- and they were estimated according to the treatment of this reaction kind in the phase chemistry mechanism was designed using the most recent kinetic and CAPRAM (38). Overall, the aqueous-phase mechanism of the DM1.0 contains 49 mechanistic data. reactions and five dissociations and represents the most detailed aqueous-phase The gas-phase reaction scheme in the DM1.0 is largely based on the DMS gas- DMS chemistry mechanism to date. phase oxidation mechanism of the MCMv3.2. The MCMv3.2 contains one of the Briefly, the module described explicitly the tropospheric oxidation of DMS and − most detailed gas-phase DMS oxidation mechanisms and is able to reproduce its oxidation products by several oxidants (mainly OH, NO3,Cl,ClO,Br,BrO,Cl2 , − experimental findings (36). However, the MCMv3.2 does not contain DMS and SO4 ) in the gas and aqueous phases. For additional information, the reader reactions with halogen species and is also missing some reactions that are is referred to SI Appendix, Model and Mechanism Description. potentially important in the marine environment (e.g., reactions with the hydroperoxyl radical). These reactions can be critical for the multiphase ACKNOWLEDGMENTS. We thank Dr. Dean Venables for helpful discussions chemistry in the MBL (5). The chemical scheme in the MCMv3.2 was extended and comments on the manuscript.

1. von Schneidemesser E, et al. (2015) Chemistry and the linkages between air quality 21. Zhu L, Nenes A, Wine PH, Nicovich JM (2006) Effects of aqueous organosulfur and climate change. Chem Rev 115(10):3856–3897. chemistry on particulate methanesulfonate to non–sea salt sulfate ratios in the 2. Andreae MO (1990) Ocean-atmosphere interactions in the global biogeochemical marine atmosphere. J Geophys Res Atmos 111(D5):D05316. sulfur cycle. Mar Chem 30(1-3):1–29. 22. Mihalopoulos N, Kerminen VM, Kanakidou M, Berresheim H, Sciare J (2007) Forma- 3. Charlson RJ, Lovelock JE, Andreae MO, Warren SG (1987) Oceanic phytoplankton, tion of particulate sulfur species (sulfate and methanesulfonate) during summer over atmospheric sulfur, cloud albedo and climate. Nature 326(6114):655–661. the Eastern Mediterranean: A modelling approach. Atmos Environ 41(32):6860–6871.

4. Myhre G, et al. (2013) Anthropogenic and natural radiative forcing. Climate Change 23. Berresheim H, et al. (2014) Missing SO2 oxidant in the coastal atmosphere? - obser- 2013: The Physical Science Basis. Contribution of Working Group I to the Fifth vations from high-resolution measurements of OH and atmospheric sulfur com- Assessment Report of the Intergovernmental Panel on Climate Change, eds Stocker TF, pounds. Atmos Chem Phys 14(22):12209–12223. et al. (Cambridge Univ Press, Cambridge, United Kingdom). 24. Bräuer P, Tilgner A, Wolke R, Herrmann H (2013) Mechanism development and 5. von Glasow R, Crutzen PJ (2004) Model study of multiphase DMS oxidation with a modelling of tropospheric multiphase halogen chemistry: The CAPRAM Halogen focus on halogens. Atmos Chem Phys 4:589–608. Module 2.0 (HM2). J Atmos Chem 70(1):19–52. 6. Bosch T, et al. (2010) World Ocean Review 2010: Living with the Oceans (Mare, 25. Saunders SM, Jenkin ME, Derwent RG, Pilling MJ (2003) Protocol for the development Hamburg, Germany). of the Master Chemical Mechanism, MCM v3 (Part A): Tropospheric degradation of 7. Bopp L, et al. (2004) Will marine dimethylsulfide emissions amplify or alleviate global non-aromatic volatile organic compounds. Atmos Chem Phys 3:161–180. warming? A model study. Can J Fish Aquat Sci 61(5):826–835. 26. Jenkin ME, Saunders SM, Wagner V, Pilling MJ (2003) Protocol for the development of 8. Chin M, et al. (1996) A global three-dimensional model of tropospheric sulfate. the Master Chemical Mechanism, MCM v3 (Part B): Tropospheric degradation of ar- J Geophys Res Atmos 101(D13):18667–18690. omatic volatile organic compounds. Atmos Chem Phys 3:181–193. 9. Gondwe M, Krol M, Gieskes W, Klaassen W, de Baar H (2003) The contribution of 27. Wolke R, et al. (2005) SPACCIM: A parcel model with detailed microphysics and

ocean-leaving DMS to the global atmospheric burdens of DMS, MSA, SO2, and NSS complex multiphase chemistry. Atmos Environ 39(23-24):4375–4388. 2 SO4 -. Global Biogeochem Cycles 17(2):1056. 28. Lana A, et al. (2011) An updated climatology of surface dimethlysulfide concentrations 10. Jones A, Roberts DL, Woodage MJ, Johnson CE (2001) Indirect sulphate aerosol forcing in a and emission fluxes in the global ocean. Global Biogeochem Cycles 25(1):GB1004. climate model with an interactive sulphur cycle. J Geophys Res Atmos 106(D17):20293–20310. 29. Breider TJ, Chipperfield MP, Mann GW, Woodhouse MT, Carslaw KS (2015) Sup- 11. Kloster S, et al. (2006) DMS cycle in the marine ocean-atmosphere system - a global pression of CCN formation by bromine chemistry in the remote marine atmosphere. model study. Biogeosciences 3(1):29–51. Atmos Sci Lett 16(2):141–147. 12. Mahajan AS, et al. (2015) Quantifying the impacts of an updated global dimethyl sulfide 30. Tilgner A, Bräuer P, Wolke R, Herrmann H (2013) Modelling multiphase chemistry in climatology on cloud microphysics and aerosol radiative forcing. JGeophysResAtmos deliquescent aerosols and clouds using CAPRAM3.0i. J Atmos Chem 70(3):221–256. 120(6):2524–2536. 31. Flyunt R, Makogon O, Schuchmann MN, Asmus K-D, von Sonntag C (2001) OH-Radical- 13. Thomas MA, et al. (2010) Quantification of DMS aerosol-cloud-climate interactions induced oxidation of methanesulfinic acid. The reactions of the methanesulfonyl using the ECHAM5-HAMMOZ model in a current climate scenario. Atmos Chem Phys radical in the absence and presence of dioxygen. J Chem Soc Perkin 2 5:787–792. 10(15):7425–7438. 32. Boucher O, et al. (2013) Clouds and aerosols. Climate Change 2013: The Physical Science Basis. 14. Barnes I, Hjorth J, Mihalopoulos N (2006) Dimethyl sulfide and dimethyl sulfoxide and Contribution of Working Group I to the Fifth Assessment Report of the Intergovernmental their oxidation in the atmosphere. Chem Rev 106(3):940–975. Panel on Climate Change, eds Stocker TF, et al. (Cambridge Univ Press, Cambridge, United 15. Faloona I (2009) Sulfur processing in the marine atmospheric boundary layer: A review Kingdom). and critical assessment of modeling uncertainties. Atmos Environ 43(18):2841–2854. 33. Simpson WR, Brown SS, Saiz-Lopez A, Thornton JA, Glasow Rv (2015) Tropospheric

16. Lucas DD, Prinn RG (2002) Mechanistic studies of dimethylsulfide oxidation products using halogen chemistry: Sources, cycling, and impacts. Chem Rev 115(10):4035–4062. EARTH, ATMOSPHERIC, AND PLANETARY SCIENCES an observationally constrained model. J Geophys Res Atmos 107(D14):ACH 12-1–ACH 12-26. 34. Grandey BS, Wang C (2015) Enhanced marine sulphur emissions offset global 17. Bork N, Elm J, Olenius T, Vehkamaki H (2014) Methane sulfonic acid-enhanced for- warming and impact rainfall. Sci Rep 5:13055. mation of molecular clusters of sulfuric acid and dimethyl amine. Atmos Chem Phys 35. Pruppacher HR, Jaenicke R (1995) The processing of water-vapor and aerosols by 14(22):12023–12030. atmospheric clouds, a global estimate. Atmos Res 38(1-4):283–295. 18. Chen H, et al. (2015) New particle formation and growth from methanesulfonic acid, 36. Berndt T, Richters S (2012) Products of the reaction of OH radicals with dimethyl sulphide

trimethylamine and water. Phys Chem Chem Phys 17(20):13699–13709. in the absence of NOx: Experiment and simulation. Atmos Environ 47:316–322. 19. Lelieveld J, Crutzen PJ, Rodhe H (1989) Zonal Average Cloud Characteristics for Global 37. Schwartz SE (1986) Mass transport considerations pertinent to aqueous phase reac- Atmospheric Chemistry Modelling (Department of Meteorology, University of Stock- tions of gases in liquid water clouds. Chemistry of Multiphase Atmospheric Systems, holm, Stockholm). ed Jaeschke W (Springer, Berlin), pp 415–472. 20. Campolongo F, Saltelli A, Jensen NR, Wilson J, Hjorth J (1999) The role of multiphase 38. Ervens B, et al. (2003) CAPRAM 2.4 (MODAC mechanism): An extended and con- chemistry in the oxidation of dimethylsulphide (DMS). A latitude dependent analysis. densed tropospheric aqueous phase mechanism and its application. J Geophys Res J Atmos Chem 32(3):327–356. Atmos 108(D14):4426.

Hoffmann et al. PNAS | October 18, 2016 | vol. 113 | no. 42 | 11781 Downloaded by guest on September 30, 2021