In this report we analyze airborne measurements to suggest that methanol in biomass burning smoke is lost heterogeneously in clouds. When a smoke plume intersected a cumulus cloud during the SAFARI 2000 field project, the observed methanol gas phase concentration rapidly declined. Current understanding of gas and aqueous phase chemistry cannot explain the loss of methanol documented by these measurements. Two plausible heterogeneous reactions are proposed to explain the observed simultaneous loss and production of methanol and formaldehyde, respectively. If the rapid heterogeneous processing of methanol, seen in a cloud impacted by smoke, occurs in more pristine clouds, it could affect the oxidizing capacity of the troposphere on a global scale.
 After methane, methanol (CH3OH) is the second most abundant organic trace gas in the Earth's atmosphere [Singh et al., 2000, 2001]. In the troposphere CH3OH reacts directly with OH in the gas and aqueous phases, contributing to sources of reactive species such as formaldehyde (HCHO), ozone and hydrogen radicals. Currently, the sources and sinks of methanol are not well understood [Singh et al., 2000; Heikes et al., 2002; Galbally and Kirstine, 2002]. The mass of methanol lost in the atmosphere due to gas and aqueous phase OH oxidation is estimated to be ∼100 and ∼10 Tg/year [Heikes et al., 2002], respectively, which is smaller than the total estimated source. According to these estimates, the lifetime of methanol in the atmosphere is primarily controlled by gas phase oxidation. Below the possibility of a heterogeneous sink for methanol is explored.
2. Measurement Summary
 The SAFARI 2000 (S2K) field project employed a combination of ground, airborne, and satellite instruments to study the atmospheric effects of biomass burning [Hobbs et al., 2003]. The mixing of clouds and smoke was common during the S2K experiment (see auxiliary photos). The bottom figure in the auxiliary file shows a smoke plume mixing with a capping cumulus cloud on August 18, 2000. The latter situation provides a unique opportunity to explore how various trace gases, particularly oxygenated organics, interact with clouds.
 A number of trace gases were measured with an airborne FTIR (AFTIR) instrument [Yokelson et al., 1999, 2003a] during the S2K field experiment. Laboratory studies indicate that the inlet has a good passing efficiency for a large number of molecules, including CH3OH, HCHO, and C2H4 [Yokelson et al., 2003b]. Most reactive and stable trace gases present above 5–20 ppbv (e.g., the CH3OH detection limit is ∼5 ppb) were measured by the AFTIR.
 On August 18, 2000 the University of Washington's Convair-580 aircraft penetrated a fresh biomass-burning plume at various altitudes below and within a capping cumulus cloud. The AFTIR measured rapid loss and production of CH3OH and HCHO, respectively, which correlated with increased exposure to the cloud [Yokelson et al., 2003a]. Isolated, and cloud-free, plumes from three other fires were sampled at different stages in their evolutions, ranging in age from 5–45 minutes, but no significant changes of CH3OH and HCHO were detected in these cases [Yokelson et al., 2003a; Hobbs et al., 2003]. This is consistent with the stability of these two compounds for the first 2–3 hours of “cloud-free” smoke aging for a biomass fire observed in Alaska [Goode et al., 2000].
 The smoke plume was penetrated and sampled at three altitudes: once below the cloud, and twice within the cumulus cloud at different altitudes above the cloud base (Table 1). We assume that the degree of cloud processing of the smoke increases with altitude above the cloud base. From the rate of the observed ozone recovery in the smoke [Yokelson et al., 2003a], the horizontal dimensions of the cloud (∼3 km), the measured horizontal windspeed (5 ± 1 m s−1), and typical updraft velocities (∼1–2 m s−1), we can estimate how long the smoke sampled was cloud-processed. The results are summarized in Table 1. We use CO as a conserved tracer to account for the dilution of the smoke [Hobbs et al., 2003]. The CH3OH gas-phase concentration measured by the AFTIR in the lowest in-cloud sample (sample 2) was not close to 36 ppb, which would occur from dilution alone, but was 15 ppb after only ∼3 minutes of exposure to cloudy conditions (Figure 1). After ∼10 minutes of exposure (Table 1, sample 3), the methanol concentration was below the AFTIR detection [Yokelson et al., 1999, 2003a].
Table 1. Observations of Chemical Changes Due to Cloud Processing of Smokea
Excess Mixing Ratio (ppb)
Excess mixing ratio (EMR) is defined as ppbv in the smoke minus ppbv in the background. Numbers in parentheses are dimensionless quantities, which are obtained by dividing the species EMR by the CO EMR. Sample 1 gives the EMR measured below the cloud base in a diluting column. Trace gas-to-CO ratios in sample 1 (cloud free) provide the benchmark for detecting chemical changes in samples 2 and 3, which were taken after the smoke intersected a cumulus cloud. 2a and 3a give the calculated EMR at different heights, due to dilution only, which were obtained by multiplying trace gas to CO ratios in sample 1 by the EMR of CO in samples 2 and 3, respectively. 2b and 3b give the measured EMR after ∼3 and ∼10 mins of smoke exposure to the cloud, respectively. 2c and 3c give the inferred chemical changes (sample (b)-sample (a)) due to cloud processing of the smoke. AFTIR precision and calibration accuracy are better than 1% and 5%, respectively (1σ). The detection limits, which dominate the uncertainty (2σ) in this data are: CH3OH and CO (∼5 ppbv); NO, NO2, and HCHO (15–20 ppbv).
−25 to −30
3. Gas- and Aqueous-Phase Oxidation
 Biomass burning emits large quantities of chlorine species into the atmosphere [Li et al., 2003]. We performed some simple calculations to explore the possibility of oxidation chemistry involving chlorine radicals. A drop in CH3OH concentration by more than one-half in ∼3 mins (Figure 1) requires a steady state concentration of Cl in the gas phase of ∼7 × 107 cm−3 (rate constant for Cl reaction with CH3OH is 5.5 × 10−11 cm3 s−1) [Atkinson et al., 1997]. At such high Cl levels, C2H4 oxidation would have been complete, since it reacts with Cl about 5 times faster than methanol (rate constant for C2H4 reaction with Cl is 3 × 10−10 cm3 s−1) [Atkinson et al., 1997]. Clearly, this did not happen since no significant loss in C2H4/CO ratio was measured with the AFTIR during even the first 10 mins of exposure to the cloud [see Yokelson et al., 2003a, Table 1]. In addition, in cloud-free smoke, produced by another fire in the same region, the rate of oxidation in the gas phase for a large number of species was consistent with OH chemistry [Hobbs et al., 2003].
 Having ruled out chlorine chemistry, the rapid decrease in the CH3OH concentration cannot be explained by “standard” gas and aqueous phase chemistry (see Appendix A in the auxiliary file, reactions 1–4). In 3 mins (Figure 1), less than 0.1 ppb of CH3OH can be lost through gas-phase reaction with OH. The maximum aqueous phase loss, where we assumed water droplets to be in thermodynamic equilibrium with the gas phase, can account for only ∼7 ppb loss of CH3OH in 3 mins. In fact, more than 3 times as much CH3OH was lost during this time period (Table 1). Overall, measurable loss through bulk aqueous phase chemistry is highly unlikely due to mass transport limitations, since significant gas transfer into the bulk of water droplets cannot occur in 3 to 10 mins of exposure [Jacob, 1986] (see auxiliary file (Appendix A), for more details). Below we propose a few plausible heterogeneous chemical processes involving OH and NO2 that may help to explain the rapid loss of gaseous CH3OH in the cloud.
4. Possible Heterogeneous Reactions
 Here we use the measured rate of change in the CH3OH gas phase concentration, as well as typical microphysical properties of cloud drops, to determine a pseudo first-order rate constant for methanol reaction on cloud droplet surfaces. In deriving the pseudo first-order rate constants below, we assumed only the gas phase supplied the oxidant to the surface of cloud drops. However, in reality, the steady state oxidant concentration at the surface is also affected by supply from the bulk phase as well as possible in situ conversion mechanisms at the surface (e.g., the reaction of O(1D) with surface water to produce in situ OH). Some recent molecular dynamics simulations suggest that the OH concentration is enriched at the interface of aqueous salt solutions [Roeselova et al., 2003]. Whether this conclusion applies to water droplets in clouds requires further investigation. However, it is beyond the scope of this paper to distinguish what processes may have contributed to the oxidant concentration at or near the surface.
 The most likely heterogeneous process to consider here is the reaction of adsorbed CH3OH with OH:
This heterogeneous process may also help to explain why a rapid production of HCHO was simultaneously observed in the cloud (Table 1). Laboratory measurements indicate that CH3OH on dry ammonium sulfate surfaces (relative humidity (RH) ∼35%) can react with OH to yield formaldehyde [Oh and Andino, 2000, 2001]. According to these measurements, the rate of gas phase methanol reaction with OH is enhanced by about 30% in the presence of dry ammonium sulfate surfaces at ∼35% RH. However, no yield for CH3OH reaction with OH on dry ammonium sulfate was measured at ∼3% RH [Sorensen et al., 2002]. Since the RH conditions were different in the above experiments, we suspect that the rate of this reaction may have a dependence on the RH and/or the physical phase of the particles. In the Sorensen et al. study, it was assumed that OH was only supplied to the surface from the gas phase. While this assumption may hold at 3% RH, at higher RHs other mechanisms (noted above) may also supply OH to the surface of particles, thereby enhancing the overall rate of this surface oxidation process. More experimental work is needed to investigate whether the above reaction can occur at a faster rate on cloud drop surfaces.
 To obtain a pseudo first-order rate constant for the heterogeneous CH3OH reaction with OH, the measurements shown in Figure 1 were used. Assuming a constant concentration for gas phase OH in the smoke ([OH] = 2 × 107 cm−3) [Hobbs et al., 2003], the change in the CH3OH gas-phase concentration can be calculated from,
where, [CH3OH]i is the expected concentration of CH3OH in sample 2 (∼36 ppbv), and [CH3OH]f is the measured concentration of CH3OH in sample 2. We set [CH3OH]f = 15 ppbv and Δt = 180 s (see Table 1) to derive a value for the rate of this heterogeneous process (khetOH) at −5°C. The rest of the parameters in the above equation are constant and are given in the auxiliary file (Appendix A).
 The normalized pseudo first-order heterogeneous rate constant obtained here (∼3.7 × 10−8 cm3 s−1 cm, see Appendix A in the auxiliary file) is about 300 times larger than the gas-phase rate constant for the conditions sampled in this cloud (S = 0.0063 cm2 cm−3, where S is the surface area of cloud droplets per unit volume of air). Although this relative increase appears to be too large, it is perhaps more useful to compare the heterogeneous and aqueous phase loss rates to assess whether the inferred rates derived here are plausible. To do so we analyze the loss rates calculated and measured for sample 2 (Table 1). The maximum aqueous phase methanol loss for sample 2 is 7 ppb (Figure 1). On the other hand, the observed loss for sample 2 is 21 ppb (Table 1), indicating that the heterogeneous loss is ∼3 times faster (21/7) than the maximum aqueous phase loss. In a recent study, the reaction of Cl2•− radical anion with ethanol was found to be at least two times faster at the air-water interface than in the bulk [Strekowski et al., 2003]. Thus the inferred loss rate calculated here for the methanol reaction with OH on the cloud surface is consistent with current information (admittedly limited) on reactions of radicals with small alcohols on aqueous surfaces.
 Another possible heterogeneous process is the reaction of adsorbed methanol with gas phase NO2:
where methyl nitrite (CH3ONO) photolyzes in the gas phase, as shown above, to yield NO and formaldehyde. Taylor et al.  showed that the photochemical lifetime of CH3ONO is nearly independent of altitude and is ∼2 mins. Laboratory studies indicate that the reaction of CH3OH with NO2 is enhanced catalytically on various surfaces, including stainless steel, Pyrex and smog chamber walls [Takagi et al., 1986]. We used pseudo first-order kinetics to derive a normalized rate constant (∼4.3 × 10−13 cm3 s−1 cm, see Appendix A in the auxiliary file) for the CH3OH reaction with NO2. The procedure used is identical to that described above for the reaction with OH.
 Of the two reactions proposed above, the OH reaction is more plausible since methanol loss (∼25–30 ppb) and HCHO production in sample 3 (31 ppb) are nearly identical (see Table 1). However, the measurements seem to indicate that NO2 and methanol were consumed as NO and HCHO were produced in the cloud (Table 1), which provides some support for the NO2 surface reaction.
5. Atmospheric Implications
 In Figure 2 the ratio of heterogeneous-to-gas phase loss of CH3OH is shown using the rate constants given in Appendix A for typical ranges of oxidant concentrations. The heterogeneous loss is calculated assuming that an air mass spends on average ∼1 h in cloudy air each day [Jacob, 1999]. The gas phase loss is calculated for 24 h. If the heterogeneous rate constants derived in this work from field data are applicable to the global troposphere, then clouds can provide a major sink for gas-phase CH3OH through reaction with OH. For the NO2 reaction, the effect could be significant only for polluted conditions.
 In clouds affected by smoke there might be significant organic coverage on the surface. This might, in turn, promote the methanol uptake and oxidation on the surface. In clouds unaffected by smoke, the surface tension of water droplets is lower than pure water due to organic contamination [Facchini et al., 2000]. Partial or complete organic coatings may provide a more suitable medium for CH3OH uptake and for reactions to occur in the global atmosphere [Djikaev and Tabazadeh, 2003]. However, the rate of heterogeneous CH3OH oxidation in the global atmosphere may be lower than that determined here for a “dirty” cloud, since organic layers on aqueous surfaces are probably thicker and more prominent in smoky environments.
 Clearly, future targeted laboratory and fieldwork are needed to determine whether the heterogeneous CH3OH loss, reported here to occur rapidly in a cloud affected by smoke, also occurs in more pristine clouds. The mixing of clouds and smoke is quite common throughout the tropics [e.g., Reid et al., 1998]. If the heterogeneous reactions proposed here occur in cleaner environments, then oxidation of methanol on cloud drops would be a more important process, since methanol and clouds are globally distributed in the Earth's atmosphere. Thus, heterogeneous oxidation in clouds has the potential to provide a major global source for reactive organic species (e.g., HCHO, CH3ONO), as well as a sink for CH3OH.
 We thank Prof. Roger Atkinson for helpful comments and discussions. Data collection was supported by grants NAG5-9022 and NAG5-7675 from NASA's Radiation Science Program and grants ATM-9901G24, ATM-9900494, and ATM-0228003 from NSF's Division of Atmospheric Sciences.