Further improvement of wet process treatments in GEOS-Chem v 12 . 6 . 0 : Impact on global 1 distributions of aerosol precursors and aerosols 2

Abstract. Wet processes, including aqueous phase chemistry, wet scavenging, and wet surface uptakes during dry deposition, are important for global modeling of aerosol precursors and aerosols. In this study, we improved the treatments of these wet processes in the GEOS-Chem v12.6.0, including pH calculation for cloud, rain, and wet surface, fraction of cloud available for aqueous phase chemistry, rainout efficiencies for various types of cloud, empirical washout by rain and snow, and wet surface uptakes during dry deposition. We compared simulated surface mass concentrations of aerosol precursors and aerosols with surface monitoring networks over the United States, Europe, Asia, and Arctic regions, and showed that the model results with the updated wet processes agree better with measurements for most species. With the implementation of these updates, normalized mean biases (NMB) of surface nitric acid, nitrate, and ammonium are reduced from 78 %, 126 %, and 45 % to 13 %, 24 %, and 6.2 % over US sites, from 56 %, 105 %, and 91 % to −20 %, −5.1 %, and 22 % over Europe sites, and from 121 %, 269 %, and 167 % to −18 %, 40 %, and 86 % over Asia sites. Comparison with surface measured SO2, sulfate and black carbon at four Arctic sites indicated that these species simulated with the updated wet processes match well with observations except large underestimation of black carbon at one of the sites. Furthermore, we compared model simulation with aircraft measurement of nitric acid and aerosols during ATom-1 and ATom-2 periods and found seasonal variation and vertical profile of these species have been successfully improved by considering the updated wet processes. The investigation of impacts of updated wet process treatments on surface mass concentrations indicated that the updated wet processes have strong impacts on the global means of nitric acid, sulfate, nitrate, and ammonium and relative small impacts on the global means of sulfur dioxide, dust, sea salt, black carbon, and organic carbon.


empirical rates for nitric acid and water soluble aerosols in washout. These changes together 23 reduced the normalized mean biases (NMB) of simulated nitric acid, nitrate, and ammonium 24 mass concentrations at the United States' surface monitoring networks from 145 %, 168 %, and 25 81 % to 24 %, 25 %, and 13 %, respectively. However, the impacts of updated wet scavenging 26 scheme on the simulations over other regions (Europe, Asia, and remote areas) and free 27 atmosphere were not investigated. Moreover, L2019 only investigated the changes of nitric acid, 28 nitrate, and ammonium. The impact of updated wet scavenging scheme on other aerosols such as 29 sulfate, sea salt, dust, and carbonaceous aerosols were not investigated in that work. Due to the 30 large impact of updated wet scavenging on model simulations, a comprehensive validation of 31 https://doi.org /10.5194/gmd-2020-11 Preprint. Discussion started: 21 January 2020 c Author(s) 2020. CC BY 4.0 License. gases. Scavenging of aerosol by snow and cold-mixed precipitation was updated by Wang et al. 23 (2011Wang et al. 23 ( , 2014 Chem_species#Definition_of_Henry.27s_law_constants).

27
L2019 showed that the assumption of in-cloud condensation water with a fixed value (1  3 where F is the fraction of a water-soluble tracer in the grid-box scavenged by rainout, Δt (s) is 4 the model integration time step. k is the first-order rainout loss rate which represents the 5 conversion of cloud water to precipitation water. f c , P r (g⋅m -3 ⋅s -1 ), and LCW (g⋅m -3 ) are the grid-6 box mean cloud fraction, the rate of new precipitation formation, and liquid phase cloud water 7 content, respectively. 8 L2019 also showed that the difference between observations and simulations can be 9 further reduced, through (1) the update of empirical washout coefficients by rain for water- 10 soluble aerosol with the value which was calculated by the parameterization of Laakso et al.

11
(2003) for a 500 nm particle diameter, and (2) the new estimated washout coefficients for nitric 12 acid by referring to field measurements for particles with a 10 nm diameter (Laakso et al., 2003) 13 and the theoretical dependence of scavenging coefficients on particle sizes for particles < 10 nm 14 (Henzing et al., 2006). L2019 only focused on warm cloud wet scavenging, and did not 15 systematically consider the impact of wet process treatments on the simulated aerosol precursors 16 and aerosols. Here we show that a number of treatments in GC12 and L2019 can be further 17 updated (as detailed below) to improve the performance of GEOS-Chem in simulating spatial 18 and temporal variations of major aerosol precursors and aerosols in a global scale. 19 20 2.1 pH for cloud, rain, and wet surface 21 Water pH is important for dissolution and subsequent aqueous phase reactions of water- 22 soluble gases (Turnock et al., 2019;Ervens, 2015;Pandis and Seinfeld, 1989). Based on Henry's 23 law, dissolution of water-soluble gases can be calculated as:

25
where f w is the dissolution fraction for water-soluble gases, H * (mol⋅L -1 ⋅atm -1 ) is effective 26 Henry's law constant, R (0.08205 L⋅atm⋅K -1 ⋅mol -1 ) is the gas constant, T (K) is the temperature, 27 and LW (m 3 ⋅m -3 ) is the liquid water content. H * represents the impact of temperature, water acidy, and aqueous phase equilibrium on 1 solubility of water-soluble species (Seinfeld and Pandis, 2016 (Alexander et al., 2012). This iterative calculation is updated to use Newton's method in order to 1 arrive at a consistent result (Moch et al., 2019). To represent the removing of aerosols due to 2 rainout, GC12 assumes 30 % of sulfate, nitrate, and ammonium are removed away from cloud 3 water before cloud water pH calculation. To take into account the variations in the amount of 4 these species rained out, we propose to directly use the real-time rainout fractions for 5 corresponding species which are calculated during the treatment of wet scavenging to replace 6 this constant value (i.e., 30%). Additionally, in GC12, sulfate is assumed to be the only soluble 7 nonvolatile ion (SNVI) in cloud water, while ammonium and nitrate are treated as volatile 8 species similar to ammonia and nitric acid: 10 Previous studies found that observed ammonium-sulfate aerosol molar ratio is lower than 11 2 over the US (Silvern et al., 2017;Hidy et al., 2014

21
Rainwater pH, which is used for the calculation of water-soluble gases' effective Henry's 22 law constants for rain droplets (Eqs. [3][4][5], is assumed to be a constant value of 4.5 in GC12. 23 Rainwater pH is determined by the cloud water pH where the rain is produced, uptakes of water 24 and ions during rainfall processes, and evaporation of rain droplets. In addition, rainwater pH 25 also depends on temperature (Smith and Martell, 1976). Although it is difficult to fully trace 26 rainwater pH in the model based on current available information in GC12, we use cloud pH at 27 where rainout occurs to represent rainwater pH for rainout process and rainwater-mass-weighted 28 cloud pH above where washout occurs to represent rainfall water pH for washout process in this 29 work. and leaf water pH of 7. Surface water of land is dominated by leaf water whose pH is ~7. pH of 4 ocean surface water varies from 8 to 8.5 (Antonov, 2010;Jacobson, 2005 10 In GC12, the fraction of cloud available for aqueous phase chemistry is assumed to be 11 100 % of grid box cloud fraction when temperatures are above 258 K and 0 % of grid box cloud 12 fraction when temperatures are below 258 K. It means aqueous phase chemistry in mixed cloud 13 where temperatures are often below 258 K is not considered in GC12. However, many studies 14 indicated that supercooled cloud water can exist when temperatures are above 237 K (Rosenfeld 15 and Woodley, 2000;Sassen, 1985). Therefore, we propose to calculate aqueous phase cloud 16 fraction based on MERRA-2 cloud liquid content and cloud ice content when temperatures are 17 higher than 237 K and lower than 263 K:

Fraction of cloud available for aqueous phase chemistry
where f aq is aqueous phase cloud fraction, LWC (g m -3 ) is grid box mean liquid phase cloud 20 water content, and ICW (g m -3 ) is grid box mean ice phase cloud water content.

21
The surface pre-melt layer or quasi-liquid layer occurs when temperatures are above 263

22
K (Nenow, 1984;Ocampo and Klinger, 1983). Conklin et al. (1993) suggested that the ice 23 surface can be modeled as an aqueous phase when temperatures are higher than 265 K. So for 24 temperature higher than 263 K, we assume aqueous phase cloud fraction equals grid mean cloud 25 fraction:

27
The 263 K cut-off used here is to reflect the cover of meltwater on ice when temperature 28 is not too low.  7 where E r is the rainout efficiency for corresponding species. Eq. (11) is the same as Eq. (1) 8 except Eq. (11) contains E r in the rainout calculation. 9 In GC12, rainout efficiencies for water-soluble aerosols are assumed to be 100 % while 10 those for water-soluble gases, except nitric acid and SO 2 , are calculated via Henry's law 11 constants (Jacob et al., 2000). E r of nitric acid is assumed to be the same as water-soluble 12 aerosols due to its high solubility. Due to the low solubility of SO 2 in water, rainout of SO 2 is 13 limited by the aqueous phase oxidation of SO 2 by H 2 O 2 rather than the absorption by cloud water 14 (Chin et al., 1996). E r of SO 2 in GC12 is assumed to be the same as water-soluble aerosols but 15 limited by the availability of H 2 O 2 in the precipitating grid box. However, GEOS-Chem already 16 accounted for in-cloud oxidation of SO 2 as part of the aqueous phase chemical calculation, so 17 doing the same in the scavenging calculation would be double-counting the removal of SO 2 .

18
Considering the low solubility of SO 2 in water, it is more appropriate to calculate rainout 19 efficiency for SO 2 based on Henry's law. In the present work, we assume E r of SO 2 equals its 20 dissolution fraction:

22
with f w_SO2 calculated with Eq. (2). 23 In the present work, we also modified rainout efficiencies for hydrophilic black carbon 24 (BC) and primary organic carbon (POC), from 100% in GC12 to 50%. The rationale for the 25 modification is that, although the aging of BC and POC in the atmosphere converts these 26 aerosols from hydrophobic to hydrophilic, they are not as easy to be activated into cloud droplet particles (Yu et al., 2012;Yu and Luo, 2009) but this will be the subject of future work.

26
In addition to T, ice nucleation efficiency of particles also depend on their sizes and 27 smaller particles (diameter ＜ 500 nm) are less likely to act as IN (Niedermeier et al., 2015). 28 While most of mass of dust particles are dominated by those larger than 500 nm, a significant 29 fraction of BC particles are smaller than 500 nm. Based on sectional aerosol microphysics calculation in GEOS-Chem-APM (Yu and Luo, 2009), the mass fraction of BC particles with 1 diameter ＞ 500 nm is ~50 %. In this study, we assume E r for hydrophobic BC in both mixed 2 cloud (237 K ≤ T ＜ 258 K) and cold cloud (T < 237 K) are 50 % of those values for dust. 3 Water-soluble aerosols are removed via homogeneous freezing nucleation in cold cloud 4 (Wang et al., 2014;Liu et al., 2001). Strom et al. (1997) observed that ~ 40 % of preexisting 5 aerosol mass is incorporated in ice crystal. In this work, we assume cold cloud rainout 6 efficiencies are 40 % for water-soluble aerosol, 50 % for hydrophobic black carbon, and 100 % 7 for dust, respectively.

8
In GC12, cold cloud wet scavenging of nitric acid is treated the same as water-soluble 9 aerosol. However, in cold cloud (T < 237 K), nitric acid is removed by the partition on ice crystal 10 (Kärcher and Voigt, 2006;Voigt et al., 2006) is volume mixing ratio of in-cloud water and in-cloud ice, and 23 [HNO 3 ] vmr is volume mixing ration for nitric acid gas.  26 In GC12, ICCW for cold cloud (T < 237 K) is assumed to have a fixed value of 1 g⋅m -3

27
which is the same as that of warm cloud. This assumption significantly underestimates wet scavenging due to rainout in cold cloud (T < 237 K). L2019 replaced the fixed ICCW with cloud 1 water and rain water as shown in Equ. 1. However, water-soluble aerosols in cold cloud (T < 237 2 K) can also exist in ice due to freezing of supercooled water, therefore, we calculate ICCW for 3 cold cloud as:  Accumulation-mode washout coefficients were used for all aerosols except dust and sea salt, for 10 which the coarse mode coefficients were used. Previous studies noticed that washout rates by 11 rain derived from field measurements are 1 to 2 orders of magnitude larger than the values from 12 theoretical calculation (Wang et al., 2010;Luo et al., 2019). Therefore, L2019 recommended 13 using empirical washout coefficients for the simulation of washout by rain.
14 Wang et al. (2014) found that the large differences in washout rate between field 15 measurements and theoretical calculation not only appear in washout by rain but also appear in 16 washout by snow. In this work, we use the semi-empirical parameterization developed by Wang 17 et al. (2014) for the calculation of nitric acid and aerosol washout by both rain and snow. 18 Washout rate is calculated by an exponential equation: 20 where k wash (s -1 ) is the washout rate, P d (mm h -1 ) is rain or snow falling from upper layers, f r is 21 rainfall area fraction, Λ is washout scavenging coefficient, and b is an exponential coefficient. 22 The values of Λ and b for nitric acid and aerosol washout by rain (T > 268 K) and snow 23 (248 K < T < 268 K) are shown in Table 1. We assume precipitation at temperatures lower than 24 248 K is dominated by ice. GC12 assumed washout of aerosol by ice is the same as that by snow. replaced with the value of 2×10 -4 which is 20 times higher than the value by rain. Washout by ice 7 is assumed to be 1/5 of that by snow. 2.6 Wet surface uptakes during dry deposition 10 Uptakes at wet surface are strongly influenced by dissolution processes. The solubility of 11 SO 2 , H 2 O 2 , and NH 3 at wet surface needs to be calculated via effective Henry's law coefficient 12 because it is associated with a series of aqueous phase reactions (Seinfeld and Pandis, 2016  considering the updated wet processes described in section 2, and this case is called WETrev. All 1 simulations are run with 2º×2.5º horizontal resolution and 47 layers from surface to 0.01 hPa.  aerosol precursors and secondary inorganic aerosols over Europe and Asia were observed by 12 EMEP and EANET, respectively. As shown in Fig. 1 (a-c), simulated SO 2 for the 3 cases is 13 lower than observed values over the US but higher than the observations over Europe and Asia.
14 Over the US, simulated SO 2 is ~ 20 % lower than observations. One possible reason is that a  small. The aqueous concentration of ammonia is much lower than nitric acid, and therefore wet 1 processes show relatively small impact on the simulation of ammonia. WETrev case, the temperature limitation of aqueous phase chemistry is extended from 258 K to 10 237 K. This change allows aqueous phase chemistry to be simulated when temperatures are low.

11
After employing the new approaches of cloud water pH and aqueous phase cloud fraction indicates wet processes have a small impact on the simulation of BC and OC in these regions.

2
The small impact of wet processes on BC in the US and Europe is because 80 % of emitted BC is 3 assumed to be hydrophobic aerosol which needs 1.15 days to be converted to hydrophilic BC. 4 Updated wet processes has little impact on hydrophobic aerosol in low troposphere where wet  11 We also studied the impact of updated wet processes on SO 2 , sulfate and BC surface 12 mass concentrations at several Arctic sites where measurements are available. Figure 3 shows the 13 comparison of SO 2 at Nord (81.6ºN, 16.7ºW) and Zeppelin (78.9ºN, 11.9ºE). GC12 case matches 14 well with the observed SO 2 at Nord but 3 times overestimated SO 2 at Zeppelin in January and 15 December. The updated wet scavenging (yellow line) shows small impact on SO 2 simulation at 16 Arctic. Simulated SO 2 is slightly reduced during winter and spring. In WETrev case, we assumed 17 SO 2 dry deposition velocity is 0.01 cm s -1 when temperatures are lower than 253 K. It slightly 18 enhances SO 2 at the higher latitude site Nord during winter. Figure 4  and during the long-range transport hydrophobic BC is aged and covered to hydrophilic BC. The 1 assumption of reduced hydrophilic BC rainout efficiency in the WETrev case increases 2 simulated BC mass concentration and enhances agreement with observations at these Arctic sites. at the upper troposphere where pressure is lower than 300 hPa. As we mentioned earlier, L2019 18 may overestimate cold cloud wet scavenging of nitric acid due to the old treatments in GC12. 19 With updated cold cloud scavenging in WETrev, bias of nitric acid simulated by L2019 at the 20 upper troposphere is reduced. Figure 5  ATom-2 is more obvious than that during ATom-1. It is because black carbon emitted from open 26 fire in January is much less than that in July. Black carbon observed during ATom-2 is 27 dominated by hydrophilic black carbon which is more affected by wet scavenging processes, 28 while black carbon observed during ATom-1 is dominated by hydrophobic black carbon.

29
Updated wet scavenging shows small impact on organic carbon vertical profiles during both 30 ATom-1 and ATom-2.

17
In this study, we updated aqueous phase chemistry and wet scavenging for SO 2 and 18 sulfate, rainout efficiencies for warm, mixed, and cold cloud, empirical washout by rain and ammonium. The updated wet process treatments exhibit relatively small impacts on the 10 simulated global means of SO 2 , dust, sea salt, black carbon, and organic carbon. Although we 11 tried to make the updated wet process treatments to be the state-of-the-art, there still exit 12 limitations of the work presented in this study. For example, washout efficiencies of water 13 soluble species such as SO 2 and ammonia are sensitive to rain water pH values. In this study, we 14 simply assumed rainwater pHs for rainout and washout are cloud pH at where rainout occurs and 15 rainwater-mass-weighted cloud pH above where washout occurs, respectively. However, rain 16 water pH needs to be calculated by tracing the cloud process and precipitation process of rain 17 water lifecycle. The impact of traced rain water pH on wet scavenging needs to be further 18 investigated. 19 20 Code and data availability. The code of GEOS-Chem 12.6.0 is available through the GEOS-