Biodegradation by bacteria in clouds: An underestimated sink for some organics in the atmospheric multiphase system

Water-soluble organic compounds represent a significant fraction of total atmospheric carbon. The main oxidants towards them in the gas and aqueous phases are OH and NO3 radicals. In addition to chemical solutes, a great variety of microorganisms (e.g. bacteria, viruses, fungi) has been identified in cloud water. Previous lab studies suggested that for some organics, biodegradation by bacteria in water is comparable to their loss by chemical processes. We perform model sensitivity studies over large ranges of 15 biological and chemical process parameters using a box model with a detailed atmospheric multiphase chemical mechanism and biodegradation processes to explore the importance of biodegradation of organics in the aqueous phase. Accounting for the fact that only a small number fraction of cloud droplets (~0.0001 – 0.001) contains active bacteria cells, we consider only a few bacteria-containing droplets in the model cloud. We demonstrate that biodegradation might be most efficient for volatile organic compounds (VOC) 20 with intermediate solubility (~10 ≤ KH(eff) [M atm] ≤ 10, e.g., formic and acetic acids). This can be explained by the transport limitation due evaporation of organics from bacteria-free droplets to the gas phase, followed by the dissolution into bacteria-containing droplets. For non-volatile organics (NVOC), such as dicarboxylic acids, the upper limit of organic loss by biodegradation can be approximated by the amount of organics dissolved in the bacteria-containing droplets (< 0.01%). We compare results from this 25 detailed drop-resolved model to simplified model approaches, in which either (i) all cloud droplets are assumed to contain the same cell concentration (0.0001 – 0.001 cell droplet) or (ii) only droplets with intact bacteria cells are considered in the cloud (liquid water content ~10 vol/vol). Conclusions based on these approaches generally overestimate of the role of biodegradation, in particular, for highly soluble VOC. Our model sensitivity studies suggest that current atmospheric multiphase chemistry models are incomplete for 30 organics with intermediate solubility and high bacterial activity. https://doi.org/10.5194/acp-2020-778 Preprint. Discussion started: 13 August 2020 c © Author(s) 2020. CC BY 4.0 License.


Introduction
Clouds provide a medium for multiphase chemical reactions, in which chemical species from the aqueous, solid and gas phases are transformed and can affect significantly the transport and distribution of chemical species in the atmosphere (Lelieveld and Crutzen, 1991). The chemical composition of cloud water is a 35 complex mixture containing a multitude of organic and inorganic species with a range of chemical and physical properties (e.g., reactivity, solubility, volatility). The organic fraction includes volatile (VOC) and non-volatile organic compounds (NVOC), such as aldehydes, mono-and dicarboxylic acids, organonitrogen and organosulfur compounds. VOC are dissolved from the gas phase to the aqueous phase; NVOC enter the aqueous phase via nucleation scavenging of condensation nuclei (CCN) (Löflund et al., 2002) (Ervens,40 2015). Water-soluble organic carbon (WSOC) constitutes a significant portion of the total atmospheric organic carbon mass, ranging from ~14% to ~64% depending on the sampling location (Decesari et al., 2000) (Gao et al., 2016) (Varga et al., 2001).
In addition to chemical solutes, cloud water contains microorganisms such as bacteria, yeast and fungi (Delort et al., 2010) (Hu et al., 2018). Typical concentrations of bacteria cells are on the order of 10 6 to 10 8 45 cells L -1 ; fungi and yeast cell concentrations are usually lower (~10 5 to ~10 7 cells L -1 ) (Amato et al., 2007b) (Sattier et al., 2001). The atmosphere is a stressful environment for microorganisms (low temperature, UV exposure, acidic pH, quick hydration/drying cycles) (Sattier et al., 2001) which might limit the survival time of cells in the atmosphere. Several studies have shown that bacteria can grow and be metabolically active in cloud droplets. Marker compounds such as adenosine 5'-triphosphate (ATP) (Amato et al., 2007c) , rRNA 50 (Krumins et al., 2014) or mRNA (Amato et al., 2019) have been used to demonstrate metabolic activity in the atmosphere. Metabolic activity and cell generation of bacteria is likely restricted to the time cells spent in clouds due to the abundance of liquid water (Haddrell and Thomas, 2017)  ; bacteria have been found to be dormant at lower relative humidity than in clouds (Kaprelyants and Kell, 1993). 55 The metabolic activity of bacterial strains identified in cloud water (e.g. Pseudomonas, Sphingomonas) has been investigated in lab studies, and it was shown that they can biodegrade organics (e.g., malonate, succinate, adipate, pimelate, formaldehyde, methanol, acetate, formate, phenol and catechol (Delort et al., 2010) (Vaïtilingom et al., 2010(Vaïtilingom et al., , 2011(Vaïtilingom et al., , 2013 (Amato et al., 2007a) (Ariya et al., 2002) (Husárová et al., 2011) (Fankhauser et al., 2019) (Jaber et al., 2020). Based on comparisons of experimentally derived 60 biodegradation rates to chemical rates of oxidation reactions by radicals (e.g., OH, NO3) in the aqueous phase, it was concluded that they might be similar under some conditions, and that, depending on the abundance and metabolic activity of bacteria strains, oxidation and biodegradation processes of organics may compete in clouds.
There are several estimates of WSOC loss by bacteria on a global scale: Sattier et al. (2001) estimated a sink of 1-10 Tg yr -1 , smaller than the estimate by Vaïtilingom et al (2013) (10-50 Tg yr -1 ). However, the latter is likely an overestimate as complete respiration was implied, i.e. total conversion of organics into CO2.
More conservatively Ervens and Amato (2020) suggested a global WSOC loss of 8-11 Tg yr -1 , being comparable to that by chemical processes (8-20 Tg yr −1 ). Similarly, Fankhauser et al. (2019) postulated that the role of biodegradation is likely small but they did not quantify the loss of different organics by bacteria. 70 Current atmospheric multiphase chemistry models include chemical mechanisms of different complexity with up to thousands of chemical reactions describing the transformation of inorganic and organic compounds, e.g (Ervens et al., 2003a) (Mouchel-Vallon et al., 2017) (Tilgner et al., 2013) (Woo and McNeill, 2015). However, they do not include the biodegradation of organics by bacteria despite the available data sets discussed above. 75 Cloud chemistry models often assume initially identical composition of all cloud droplets. While this might be a reasonable assumption for the chemical droplet composition due to internally mixed CCN and the phase transfer from the gas phase into all droplets, it is not appropriate for the distribution of bacteria. Due to their small number fraction of the total CCN concentration (0.001 -0.1 %, (e.g., (Zhang et al., 2020) the fraction of cloud droplets that contain bacteria cells is small (< 0.001). Thus, to explore biodegradation of organics 80 in the atmospheric multiphase system, a realistic distribution within cloud droplet population needs to be assumed.
The aim of our study is to identify conditions, under which biodegradation in clouds is significant in the atmosphere. Using a cloud multiphase box model, we explore the biological and chemical degradation of VOC and NVOC over large parameter ranges of bacterial cell concentrations, biodegradation activities, 85 chemical rate constants and Henry's law constants of the organic substrates. We compare (1) the biodegradation rates in the aqueous phase to the chemical rates in both phases, and (2) the fraction of organics consumed by biodegradation to that by chemical processes. The results of our sensitivity studies elucidate, for which organics biodegradation competes with chemical processes. Our study will give guidance for future experimental and modeling studies to further complete atmospheric models in order to 90 more comprehensively describe organic degradation in the atmosphere.

Description of the multiphase box model
We use a multiphase box model with detailed gas and aqueous phase chemistry (75 species, 44 gas phase reactions, 31 aqueous reactions (Ervens et al., 2008). The two phases are coupled by 26 phase transfer 95 processes. Phase transfer is described kinetically based on the resistance model by Schwartz (1986).
In addition to the base chemical mechanism, we define one organic species 'Org' that undergoes chemical radical reactions in the gas and aqueous phases and biodegradation by bacteria only in the aqueous phase in a small subset of the droplets as shown in Figure 1.
We consider a polydisperse droplet population of 263 droplets cm -3 in 11 size classes with drop diameters 100 of 5 µm < Ddroplet < 20 µm and a total liquid water content of LWC = 6.8×10 -7 vol/vol. Only one droplet size class includes bacteria cells (Ddroplet = 20 µm; Ndroplet = cell concentration = 0.01 cm(gas) -3 ). Thus, the cell concentration in the cloud water (Ccell= 1.5×10 7 cell L(aq) -1 ) is similar to that found in ambient clouds and used in some lab experiments (Vaïtilingom et al., 2013). The model simulations are performed for 600s which corresponds approximately to the droplet lifetime during one cloud cycle (Ervens et al., 2004). We 105 assume an initial mixing ratio of the organic compound of 1 ppb. Table 1 includes chemical rate constants for radical (OH, NO3) reactions in the aqueous and gas phases for organic compounds, for which lab data on their biodegradation rates are available (Section 2.2.2). This data 110 covers ranges of 10 3 ≤ kchemaq [M -1 s -1 ] ≤ 10 10 and 10 -17 ≤ kchemgas [cm 3 s -1 ] ≤ 10 -10 , respectively, over which the model sensitivity studies in the following are performed.  acids at pH = 3 and pH = 6 respectively.
In order to generalize our results for different radical concentrations, we present in terms of chemical rates Typical radical concentrations are on the order of 10 -15 mol L -1 for OH and NO3 radicals in the aqueous phase of clouds (Arakaki et al., 2013) (Herrmann, 2003) and 10 6 cm -3 and 10 7 -10 8 cm -3 in the gas phase, respectively (Khan et al., 2008) (Cantrell et al., 1997).
The Henry's law constants for the same organic compounds are also listed Table 1. They cover a range of 130 10 2 ≤ KH [ M atm -1 ] ≤ 10 9 . For carboxylic acids, we also report effective Henry's law constants at pH = 3 and pH = 6 as being typical for cloud water.

Biodegradation rates
In the literature, experimental rates for metabolic processes are usually reported in units of [mol cell -1 s -1 ] ( cell concentration (Ccell,aq = 1.5×10 7 cell L -1 ) resulting in a range of 10 -18 ≤ kbact [L cell -1 s -1 ] ≤ 10 -11 . This cell concentration is on the same order of magnitude as found in many clouds (Amato et al., 2007c).
Experiments with 17 different cloud bacteria in artificial cloud water with pH = 5.0 and pH = 6.5 showed 140 also nearly identical results so it can be concluded that biodegradation rates are largely independent of pH for values typical in cloud water (Vaïtilingom et al., 2011). Similar results were shown by Razika et al. (2010) who demonstrated that biodegradation rates of phenol by Pseudomonas aeruginosa were very similar when incubated at pH = 5.8, 7.0 and 8.0, respectively. This independence of the biodegradation rate on the pH of the medium, within certain limits, results from the regulation of the intracellular composition and pH 145 in bacteria cells (~6.5 -7).

Relative contributions to biodegradation and chemical loss rates
Several previous studies compared biodegradation and chemical rates in order to conclude on the potential 160 importance of biological processes in clouds (Ariya et al., 2002) (Vaïtilingom et al., 2011(Vaïtilingom et al., , 2013) (Husárová et al., 2011)(Jaber et al., 2020. Similarly, we define the relative contributions of the bacterial and the chemical processes in the two phases to the total loss rate of an organic compound where Corg,aq and Corg,gas are the concentration of VOC in the aqueous [mol L -1 ] and gas phases [cm -3 ], respectively. Their relative contributions are then expressed as whereas these fractions always add up to 100%. As NVOC are only in the aqueous phase, frchemgas,NVOC = 0.

175
While frbact, frchemaq and frchemgas define the relative importance of the bacterial and chemical loss rates, they do not give quantitative information on the total loss of the organics. the sum of frbact, frchemaq and frchemgas that always yields 100% (Eqs 7-9), by definition, Lt does not have to reach 100%. As we only consider one cloud cycle in our simulations (t = 600 s ), the values of Lt, Lbact, Lchemgas and Lchemaq are rather small (a few percent at most); however, it should be kept in mind that particles likely undergo multiple cloud cycles during their residence time in the atmosphere. Thus, the contribution 190 of chemical and biological processes to the total loss for a specific organic can be extrapolated for longer time scales based on our results. However, as one of the main goals of our study is to compare microbial activity to the better constrained chemical losses, our conclusions will be independent of the time scales.

Results and discussion
Relative loss rate of organics by biological processes (frbact) 195

frbact of VOC
We first compare the contributions of biodegradation and chemical losses to the total loss rate for VOC. In order to cover a representative parameter range for the physicochemical properties of the organic compound,  Figure 2a, b, and c, we applied Rbact = 10 -8 s -1 , 10 -6 s -1 and 10 -4 s -1 and Rchemgas = 10 -6 s -1 , respectively. For the three sets of simulations shown in Figure 2b, d, and e, Rbact = 10 -6 s -1 and Rchemgas equal to 10 -6 s -1 , 10 -5 s -1 and 10 -4 s -1 , were assumed respectively. Thus, in total three values each for Rchemgas and Rbact are discussed in the following.
In general, for all combinations of Rchemgas and Rbact, the highest frbact is predicted for organics with the highest 205 solubility (KH [M atm -1 ] ≥ 10 8 ) and lowest chemical reaction rate in the aqueous phase (Rchemaq [s -1 ] ≤ 10 -11 ).
For the lowest biological activity (Rbact = 10 -8 s -1 , Figure 2a), frbact reaches a maximum value of ~100%. For higher biological activity (Rbact = 10 -6 s -1 and 10 -4 s -1 ), frbact is always smaller and only reaches at most ~80% (Figure 2b and 2c). This trend seems counterintuitive as for the highest Rbact, the highest importance of biodegradation may be expected. We will explore the reasons for this further in Section 3.1.2 where we 210 compare loss rates in individual droplets as a function of time.
soluble organics to the aqueous phase (> 90% for KH ≥ 10 6 M atm -1 ). If Rchemaq is low, the organics do not undergo efficient chemical processes in the aqueous phase. Therefore, frbact is highest for this parameter combination. The comparison of the results for Rbact = 10 -6 s -1 and three values of Rchemgas (Figure 2b, d, e) shows a decrease of the maximum value of frbact from 80% (Figure 2b) to 4% (Figure 2e) for similar ranges 225 of KH and Rchemaq because of the dependence of frbact on frchemgas: for compounds with highest Rchemgas, the dominant loss is the gas phase reaction, leading to a high frchemgas and consequently to a lower frbact ( Figure   2e). Therefore, the parameter ranges of KH and Rchemaq, where frbact is maximum, do not change for different Rchemgas, but idecrease when the gas phase chemistry dominates the loss of the organic.
Overall, the variation in frbact as a response to changes in Rchemaq, Rchemgas and KH shows different sensitivities:

230
For example, for organics with KH = 10 6 M atm -1 , Rchemaq = 10 -9 s -1 and Rchemgas = 10 -6 s -1 , frbact is ~8% when Rbact = 10 -8 s -1 (red range in Figure 2a). This fraction increases to frbact ~ 73%, i.e. by a factor of ~ 9, when Rbact = 10 -6 s -1 (blue part in Figure 2b) and approaches ~80% when Rbact = 10 -4 s -1 (light blue part in Figure   2c). Similarly, for organics with KH ~ 10 6 M atm -1 , Rchemaq = 10 -9 s -1 and Rbact = 10 -6 s -1 , an increase in Rchemgas from 10 -6 s -1 (Figure 2b) to 10 -5 s -1 (Figure 2d) and 10 -4 s -1 (Figure 2e) decreases frbact from 73% to 23% 235 and 3%. Based on these non-linear trends, one can hypothesize that (i) a change Rbact and/or Rchemgas translates into a less than proportional change in frbact, and (ii) an increase of Rbact might translate into a larger change in frbact than an increase of Rchemgas by the same factor. Therefore, frbact is more sensitive to a change in Rbact than in Rchemgas. Given that Rchemgas only differs by about two orders of magnitude for most organics relevant in the atmospheric multiphase system (Table 1), we conclude that frbact may be largely independent of the 240 gas phase chemical reactivity. Additional sensitivity studies (not shown) reveal that using combinations of Rbact and Rchemgas other than those in Figure 2, result in slightly different locations of the maximum of frbact but in similar shapes and widths of parameter spaces, for which frbact is maximum. Therefore, our conclusions on the sensitivities seem robust for wide parameters ranges and combinations.

245
For NVOC, the analysis of frbact is limited to exploring the ranges of Rbact and Rchemaq (Figure 1) frbact is nearly zero (Figure 3). In order to understand the reasons for these trends, we explore in the following the variables included determining frbact (Equation 7). However, unlike for the NVOC, frbact does not drop to ~0% but levels off at ~70% for KH = 10 9 M atm -1 (Figure 4d). For these two Rbact and KH = 10 5 M atm -1 , [Org]bact stays also nearly constant over the whole simulation time (600s) and is higher for lowest Rbact=10 -6 s -1 (Figure 4b). However, frbact is higher for the 265 highest Rbact=10 -4 s -1 (Figure 4d). Figure S3a and S3b show the ratio of the organic concentrations in bacteria-containing and bacteria-free droplets of the same diameter. For these conditions of low Rchemgas and Rchemaq, the concentration ratio is near unity for both NVOC and VOC (KH= 10 5 and 10 9 M atm -1 ) when Rbact = 10 -8 s -1 . For higher Rbact= 10 -6 s -1 and KH = 10 5 and 10 9 M atm -1 , the concentration ratio for the VOC is near unity and 10 -3 , respectively and for the NVOC it is much lower (~10 -6 ). For Rbact = 10 -4 s -1 , this ratio is 270 also higher for KH = 10 5 M atm -1 than for KH = 10 9 M atm -1 (~10 -1 and 10 -5 respectively) and << 10 -11 for NVOC. It can be summarized that (1) for VOC, the concentration in bacteria-containing droplets is higher for KH = 10 5 M atm -1 than for KH = 10 9 M atm -1 and (2) the concentration in bacteria-containing droplets is predicted to be always smaller for the NVOC than for the VOC with at least intermediate solubility.
This difference between VOC and NVOC can be explained by the schematic in Figure 5: The insets in the 275 droplets schematically depict the temporal evolution of the organic concentrations as shown in Figure 4.
The efficient consumption of organics by bacteria in the bacteria-containing droplets (I) leads to a significant decrease of the organic concentration in these droplets. For VOC, this results in a strong deviation from thermodynamic equilibrium of the gas and aqueous phase concentrations, as defined by Henry's law (II).
As a consequence, organics diffuse from the gas phase into the bacteria-containing droplets (III). As this 280 diffusion leads to a decrease of the gas phase concentration, thermodynamic equilibrium between the gas phase and the bacteria-free droplets is not fulfilled anymore (IV) resulting in a concentration gradient between these droplets and the gas phase. Finally, organics from the bacteria-free droplets evaporate to replenish the gas phase concentration (V) and eventually the organic concentration in the bacteria-containing

Lt for VOC
While the analysis of frbact in Section 3.1 quantifies the relative importance of the biological and chemical 305 processes for the organic loss, we explore in the following the absolute loss of these processes (Lt, Eq-10).
Unlike the sum of frbact, frchemaq, and frchemgas that always yields 100%, the ranges of Lt that will be discussed in the following are rather small, i.e. 10 -4 ≤ Lt [%] ≤ 12, given that we only simulate approximately one cloud cycle (600s). Figure 6 shows Lt for the same parameters (Rchemgas, Rbact) as in Figure 2. For  increasing Rchemgas (10 -6 s -1 , 10 -5 s -1 , 10 -4 s -1 , Figure 6 c-e), with the overall highest Lt (~12%) for the highest Rchemgas and lowest KH. Comparing Figure 6a, b, and  Section 3.2.2 where the contribution of the bacterial process to Lt (i.e. Lbact) will be explored. Comparing the shapes of the three panels, shows that Lt is additive as with increasing Rbact, Lt reaches a maximum (1.6%) at KH ~ 10 4 M atm -1 and Rchemaq~ 10 -5 s -1 (Figure 6c). Similar to our findings for frbact (Section 3.1.1.), we also see in the trends of Lt that a change of several orders of magnitude in Rbact or Rchemgas translates into a smaller change in Lt. Thus, one can conclude that Lt has a similarly low sensitivity to the various 320 parameters as frbact.

340
To understand the contribution of bacteria (Lbact) in the total consumption of the organics (Lt), we explore Lbact for the same values of Rbact and Rchemgas as in Figures 2 and 6 (Figure 7). As suggested in the comparison of Figures 6a-c, the contribution of the organic loss by bacteria increases with increasing Rbact (Figure 7ac), i.e. Lbact increases from 0.025% to 0.45%.
The different shape of Figure 7 compared to other panels is somewhat misleading as the scales of the z-345 axes differ. For all conditions, there is a contribution of Lbact ~ 0.025% for the lowest Rchemaq and highest KH, i.e. when chemical activity in the aqueous phase is lowest and solubility is highest. However, when Rbact ≥ 10 -6 s -1 , the maximum value of Lbact is predicted for a narrow range of ~10 4 ≤ KH [M atm -1 ] ≤ 10 6 , nearly independent of Rchemaq. The highest value of Lbact (0.4%) is observed for highest Rbact (Rbact = 10 -4 s -1; Figure   7c). This corresponds to nearly a quarter of the total loss (Lt ≤ 1.4%, Figure 6c). To further explore why For each set of conditions, i.e. in the various panels of Figure 7, the maximum value Lbact is independent of Rchemaq and does not greatly vary for the same Rbact (Lbact,max ~ 0.05%) (Figure 7b, d, e). This is different than the trends of frbact (Figure 2) that show a decrease with increasing Rchemaq. By definition (Eq-7), frbact and 360 frchemaq are coupled and thus an increase in one value causes a decrease in the other. Contrary, Lbact is independent of the chemical contributions as it describes the absolute consumption rate related to the initial organic concentration. Comparing the trends in Figures 2 and 7 biodegradation that were solely based on comparing rates (Jaber et al., 2020) are misleading. In these studies, it was concluded that biodegradation for highly soluble compounds is likely most important.

Lt and Lbact for NVOC
For NVOC, the analysis of trends in Lt is simpler as it is the sum of Lbact and Lchemaq only. Lt increases with increasing Rchemaq, nearly independently of Rbact up to a maximum value of ~1.1% (Figure 8a). Comparison 375 of Lt to Lbact (Figure 8b) shows that the consumption by bacteria is smaller by several orders of magnitude (Lbact,max = 0.005%) and thus it is not a major contribution to the total loss. Figures 3 and 8b  rates were compared in a bulk aqueous phase (Vaïtilingom et al., 2013) or in a population of droplets with identical composition (Jaber et al., 2020). However, since the bacteria cell concentration (~0.01 cells cm -3 ) in clouds in much smaller than the droplet (= 263 droplets cm -3 , in our model), this results in a ratio of ~4.10 -395 5 bacteria cells per droplet. (ii) A multiphase system is considered with only droplets that contain an intact bacteria cell (Fankhauser et al., 2019) resulting in a liquid water content of ~ 10 -11 vol/vol. In Figure 9, we schematically contrast these approaches. While it is clear, that none of the two approaches reflect the conditions as encountered in real clouds, we will analyze in the following the extent to which these simplified model approaches lead to similar results as predicted in our detailed model discussed so far 400 (Figure 1).

3.2.2)
, we compare frbact and Lbact for KH = 10 5 M atm -1 , Rchemgas= 10 -6 s -1 , Rchemaq = 10 -7 s -1 as a function of Rbact (Figure 9 c, d). Similar to the results from the detailed model (Section 3.1.2), the low LWC approach leads to frbact < 100% at t = 600 s because organics are efficiently consumed in the bacteria-containing droplets. However, for the bulk model, frbact ~ 100% for the highest Rbact, even after 600 s, because bacterial 420 processes take place in all droplets with a reduced efficiency as compared to the processes in the single droplet in the detailed model. Consequently, the concentration of organics in all droplets remains relatively high for extended time scales, even for the highest Rbact. Figure 9d shows the dependence of Lbact between the three models on Rbact: for Rbact 10 -6 s -1 , Lbact is similar for the detailed and the bulk models (~5 ×10 -4 %) whereas it is twice as high for the low LWC model (~1×10 -3 %). However, for the highest Rbact (10 -4 s -1 ), the 425 https://doi.org/10.5194/acp-2020-778 Preprint. Discussion started: 13 August 2020 c Author(s) 2020. CC BY 4.0 License.
highest Lbact is predicted by the bulk approach (~5%) which is more than two orders of magnitude higher than predicted from the detailed model (~0.2 %). The similarity of Lbact between the detailed and the bulk models for Rbact=10 -6 s -1 is due to the efficient replenishment of organics with intermediate KH (Figure 5), i.e. the amount of organics available to bacteria are similar. However, for the highest Rbact, Lbact is much higher for the bulk model because the consumption of the organics occurs without any kinetic limitations of 430 the various transport processes that are only presented by the detailed model.
case for highest Rchemaq (10 -7 s -1 ) ( Figure S4a). However, for Rbact = 10 -4 s -1 , frbact is ~0% (at t = 600 s) in the 435 detailed model for both Rchemaq for the reasons explained in Section 3.1.2 whereas frbact ~ 99% for the bulk model because of continuous bacteria activity in all droplets. Moreover, frbact and Lbact are always 100% for the low LWC model. The role of aqueous phase chemistry in the multiphase system is predicted to be negligible by the low LWC approach as it is equally reduced as the LWC, leading to the dominance of the bacterial processes. If we compare Lbact for the bulk (blue line in Figure S4b) and the detailed models (red 440 line), one finds an increasing discrepancy between the model results: For the lowest Rbact considered here (10 -8 s -1 ) and the two Rchemaq: 10 -11 s -1 and 10 -7 s -1 respectively, Lbact for the bulk model is predicted to be ~ 9.3 and 11.6 higher than for the detailed model. For Rbact = 10 -6 s -1 and the two Rchemaq, this factor increases to 2.5×10 4 and 1.3×10 6 . For the highest Rbact = 10 -4 s -1 , Lbact between the bulk and our detailed model after 600s of simulation cannot be reasonably compared as in the detailed model after < 100 seconds, the NVOC 445 in the bacteria-containing droplets are completely depleted for high Rbact and, thus, no further consumption occurs (Lbact ~ 0%) (Section 3.2.3). In the bulk model, the organic consumption by bacteria is predicted to occur continuously as the bacteria could, in theory, consume all organics on 'infinite' time scales.
Our model comparison shows that for VOC and NVOC, both the bulk and low-LWC models overestimate the importance of biodegradation, i.e. both frbact and Lbact. The biases are highest in the bulk approach for 450 high Rbact and high KH. Also, the comparison shows that the bulk approach leads to wrong conclusions in terms of the importance of biodegradation as a function of the solubility of organics. Our model analysis emphasizes that a detailed model is needed in order to correctly represent the partitioning of VOC between the gas phase and the droplets with and without bacteria.

Atmospheric implications: How important is biodegradation for organic compounds
455 identified in cloud water?

Lbact of organic cloud water constituents
Based on our model sensitivity studies discussed in the previous sections and the literature data for cloud water organics summarized in Tables 1 and 2, we assess the importance of biodegradation of these compounds. The 3-d representations in Figure 10 are the same as in Figures 7 a-c (VOC) and Figure 8b 460 (NVOC), respectively, with slightly shifted viewing angles for better clarity. The added symbols correspond to Lbact of the compounds as listed in Table 1 for a single Rbact in each figure panel. Given that Rchemgas only differs by about two orders of magnitude for the organics in Table 1, we present the results with Rchemgas = 10 -6 s -1 . The reasoning of the assumed single Rbact is further discussed in Section 4.2.
For the carboxylic acids, two values are shown, i.e. KH(eff) and Rchemaq at pH = 3 and 6 ( Table 3) Similarly, rate constants for other (e.g. NO3) reactions could be derived using data summarized in Table 1.

Dependence of Lbact on bacteria cell concentration in cloud water
According to Eq-3, Rbact is the product of the biodegradation rate constant kbact [L Cell -1 s -1 ] ( Eq-14 The resulting Ccell,theoretical, [Cell L -1 ] are listed in Table 3, together with kbact [L Cell -1 s -1 ]. Among the VOC, Ccell,theoretical is predicted to be ~10 2 ≤ Ccell,theoretical [Cell L -1 ] ≤ 10 7 for acetic acid, formic acid and phenol 495 while it is much higher (~10 7 ≤ Ccell,theoretical [Cell L -1 ] ≤ 10 15 ) for catechol, formaldehyde and methanol that have much lower kbact than the former compounds. Typical bacteria cell concentration in cloud water are in the range of 10 6 -10 8 Cell L -1 (Amato et al., 2007b). Comparing this range to Ccell,theoretical reveals that for several compounds Ccell,theoretical falls into a realistic range (grey shaded cells in Table 3): for ambient cell concentrations in cloud water, malonic acid/malonate, succinic acid/succinate, acetic acid/acetate, formic 500 acid/formate, cathecol and phenol might fall into a range where the maximum consumption of organics by biodegradation can be expected, while for formaldehyde and methanol Ccell,theoretical is unrealistic ( ≥ 10 9 cells L -1 ) based on the available studies of bacteria cell concentrations in cloud water.

For which organics is biodegradation an efficient sink in the atmosphere?
The maximum value of Lbact is ~ 0.7% but we should keep in mind that also the predicted Lt is not higher than a few percent (Figure 6 a-c) as we restrict our simulations to the time scale of approximate one cloud 510 cycle. However, in a relative sense, our results allow us to compare the importance of biodegradation to chemical loss for the compounds shown in Figure 10. To quantify the contribution of bacteria in this total consumption (Lt) we list the ratio Lbact/Lt in Table 3. Our results clearly show that biodegradation might add significantly to the loss of formic acid/formate (Lbact/Lt= 0.65) and acetic acid/acetate (Lbact/Lt =0.70) at cell concentrations of ~7·10 7 and 2·10 6 Cells L -1 , respectively. These acids contribute on average ~68% to the 515 total dissolved carbon in cloud and fog water (Herckes et al., 2013). Their removal by dry and wet deposition are considered the major loss processes in the atmosphere (Khare et al., 1999). However, several studies also suggested that the oxidation for formic acid/formate in cloud water may be a net sink (Jacob, 1986).
Measurements during different seasons in the Amazon showed that indeed formic and acetic acids have stronger sinks during the wet season (Herlihy et al., 1987). While wet deposition was described to contribute 520 to a large extent the observed removal, chemical and possibly bacterial processes were suggested to act as additional sinks. While the latter was regarded as being inefficient due to long incubation time as observed in lab experiments, more recent experiments suggest that such incubation times are likely not occurring in the atmosphere where bacteria cells are continuously exposed to water and substrates.
We conclude that biodegradation of these major cloud water organics may be a significant sink under 525 ambient conditions, possibly even comparable to the loss by chemical reactions. While phenol is not a major contributor to the WSOC content in cloud water (5-95 nM ) (Jaber et al., 2020) as compared to 10µm (Ervens et al., 2003b) (Löflund et al., 2002) (Sun et al., 2016) for formic and acetic acids respectively, its degradation processes in the atmosphere might be of interest due to its toxic properties. Overall, our results are in agreement with previous findings that neither chemical processes nor biodegradation are major WSOC 530 losses as compared to deposition (Ervens and Amato, 2020) (Fankhauser et al., 2019). However, in order to comprehensively describe the loss processes and time scales of organic degradation and residence time scales in the atmosphere, both chemical and biological processes should be considered. Hence, we suggest that biological processes of organics with properties similar to those of formic and acetic acids and phenol (!10 3 < KH,eff [M atm -1 ] < ~10 6 , kbact =10 -11 ) should be included in atmospheric multiphase models.

535
The importance of biodegradation of NVOC is limited by the number fraction of cloud droplets that contain bacteria. Malonate/malonic acid and succinate/succinic acid contribute on average to < 5% to the total organic aerosol mass in ambient particles, e.g. (Kawamura and Ikushima, 1993) (Fu et al., 2013),. Their loss by chemical and biological processes will neither affect the total carbon budget to a large extent nor the budget of the individual compounds. These conclusions can be generalized for other NVOC aerosol 540 constituents, for which biodegradation has been suggested to occur in the atmosphere. Our assumption of static cloud droplets in the box model is certainly a simplified representation of cloud microphysics. Droplets might experience collision/coalescence in clouds leading to mixing of the cloud water constituents in the resulting larger droplets. However, such processes are unlikely to add significantly to the loss of NVOC by bacteria due to (i) the number small fraction of bacteria-containing droplets (0.001 -0.0001) and (ii) the 545 limited atmospheric residence time of large droplets which are efficiently removed by precipitation as a function of drop size (Beard and Ochs, 1984).

Summary and Conclusions
Our model sensitivity study is the first comprehensive analysis of the importance of biodegradation of organics by bacteria in the atmospheric multiphase system in comparison to chemical loss for wide ranges 550 of chemical and biodegradation kinetic data. We use a box model with drop-size-resolved aqueous phase chemistry and additional biological processes that only occur in a small number of cloud droplets, in agreement with ambient ratios of cell and droplet concentrations. We neglect the fact that bacteria cells may form agglomerates in the atmosphere and consequently, there might more than one cell per droplet. This effects could be included in our model by multiplying the biological activity in the respective droplets with 555 the number of cells.
We compare the predicted loss rates of chemical processes in both phases to those of biological processes in the aqueous phase only. In addition to presenting the relative loss rates (frbact, frchem) as in previous studies (Jaber et al., 2020) (Vaïtilingom et al., 2010), we discuss the relative amounts of organics (Lt) consumed by chemical (Lchem) and biological processes (Lbact). We find that the relative loss rate of organics by biological 560 processes (frbact) is generally higher for VOC than for NVOC. However, the total loss of the organics (Lt) is predicted not to reach any value higher than ~12% because our simulations were restricted to a period of 600 s (~ drop lifetime within one cloud cycle); it would be higher if the total particle processing time during multiple cloud cycles in the atmosphere were considered. The contribution of bacteria (Lbact) to the total loss is predicted to be highest for VOC with intermediate solubility (~10 4 ≤ KH [M atm -1 ] ≤ ~10 6 ). This can be 565 explained by the replenishment of VOC in the bacteria-containing droplets upon uptake from the gas phase and evaporation from the bacteria-free droplets, in which less efficient consumption of the organics occurs.
Less soluble organics (KH < 10 4 M atm -1 ) that partition to a smaller extent (< 1%) to the aqueous phase are mostly consumed by chemical processes in the gas phase; more soluble compounds (KH > 10 6 M atm -1 ) are predominately partitioned to the aqueous phase and, thus, the evaporation to the gas phase and consequently 570 the redistribution from the bacteria-free to the bacteria-containing droplets is kinetically more limited and thus less efficient. The ratio of the consumption of VOC by bacteria to the total loss (Lbact/Lt) might be as high as 0.7 for high biological activity and cell concentrations (~ 10 8 cells L -1 ). These values suggest that biological processes might add significantly (>70%) to the loss processes in the atmospheric multiphase system for organics with intermediate solubility such as formic acid/formate, acetic acid/acetate or phenol.

575
For NVOC, the amount of organics consumed by bacteria is restricted to the fraction of the organic dissolved in bacteria-containing droplets (~ 0.001%) as no efficient replenishment from the gas phase or from the other droplets occurs. Thus, biodegradation of NVOC does not significantly affect their atmospheric budget.
In addition to our detailed model with a realistic bacteria cell distribution within a cloud droplet population, we apply simpler model approaches: (i) Similar to many chemical model studies, bacteria cells are 580 distributed equally in all droplets ('bulk approach') resulting in cell concentrations of 10 -5 cells per droplet, clearly an unphysical assumption for intact bacteria cells. (ii) A multiphase system with only cloud droplets which contain bacteria resulting in a liquid water content of ~10 -11 vol/vol as compared to ~10 -7 vol/vol in clouds ('low LWC approach'). Comparing Lbact predicted from these approaches to results of our detailed model shows that all approaches agree in predicting Lbact for organics of low solubility (KH(eff) < 10 4 M atm -585 1 ). However, for such species the importance of biodegradation is low due to their inefficient partitioning to the aqueous phase. The bulk approach increasingly overestimates Lbact of organics with higher KH; the greatest discrepancy is predicted for highly soluble compounds (KH,eff > 10 6 M atm -1 ) as the bulk approach does not take into account the kinetic limitation due to the organic redistribution between the bacteria-free and bacteria-containing droplets. As the bulk approach implies organics in all droplets, it does not allow 590 limiting on the organic consumption by biodegradation. Predictions of the relative role of biodegradation as compared to chemical processes by the 'low LWC approach' are biased high because the loss due to aqueous phase processes is only considered in an unrealistically small fraction of droplets.
The current data sets for microbial rates of organic compounds are limited to very few compounds. Our model sensitivity study shows that biodegradation by bacteria in clouds is most efficient for compounds 595 with intermediate (effective) Henry's law constants (~10 4 M atm -1 < KH(eff) < 10 6 M atm -1 ) as found for common cloud water constituents such as formic acid/formate and acetic acid/acetate but also for less abundant species such as phenol, largely dependent of their chemical reactivity. Our framework allows to estimate the potential importance of biodegradation of organics, in comparison to chemical processes. It also gives guidance to future lab and model studies to further explore the role of biodegradation of specific 600 organics in the multiphase system. https://doi.org/10.5194/acp-2020-778 Preprint. Discussion started: 13 August 2020 c Author(s) 2020. CC BY 4.0 License.