Factors controlling tropospheric 03, OH, NOx, and SO2 over the tropical Pacific during PEM-Tropics B

. Observations over the tropical Pacific during the Pacific Exploratory Mission (PEM)-Tropics B experiment (March-April 1999) are analyzed. Concentrations of CO and long-lived nonmethane hydrocarbons in the region are significantly enhanced due to transport of pollutants from northern industrial continents. This pollutant import also enhances moderately 03 concentrations but not NOx concentrations. It therefore tends to depress OH concentrations over the tropical Pacific. These effects contrast to the large enhancements of 03 and NOx concentrations and the moderate increase of OH concentrations due to biomass burning outflow during the PEM-Tropics A experiment (September-October 1996). Observed CH3I concentrations, as in PEM-Tropics A, indicate that convective mass outflux in the mid- dle and upper troposphere is largely independent of altitude over the tropical Pacific. Constraining a one-dimensional model with CH31 observations yields a 1 O-day timescale for convective turnover of the free troposphere, a factor of 2 faster than during PEM-Tropics A. Model simulated HO2, CH20, H202, and CH3OOH concentrations are generally in agree- ment with observations. However, simulated OH concentrations are lower (~25%) than observations above 6 km. Whereas models tend to overestimate previous field measurements, simulated HNO3 concentrations during PEM-Tropics B are too low (a factor of 2-4 below 6 km) compared to observations. Budget analyses indicate that chemical production of 03 accounts for only 50% of chemical loss; significant transport of 03 into the region appears to take place within the tropics. Convective transport of CH3OOH enhances the production of HOx and 03 in the upper troposphere, but this effect is offset by HOx loss due to the scavenging of H202. Convective transport and scavenging of reactive nitrogen species imply a necessary source of 0.4-1 Tg yr -1 of NO., in the free troposphere (above 4 km) over the tropics. A large fraction of the source could be from marine lightning. Oxidation of DMS transported by convection from the boundary layer could explain the observed free tropo- spheric SO2 concentrations over the tropical Pacific. This source of DMS due to convection, however, would imply in the model free tropospheric concentrations much higher than observed. The model overestimate cannot be reconciled using recent kinetics measurements of the DMS-OH adduct reaction at low pressures and temperatures and may reflect enhanced OH oxidation of DMS during convection. Model results show a similar trend but simulated column concentrations are 15% lower than observations. The underestimate is due to that in the upper troposphere (section

In addition to the O3-HOx-NOx chemistry over the tropical Pacific, examined by Wang et al. [2000] for PEM-Tropics A, we will investigate sources of SO2 in the free troposphere. Thornton et al. [ 1999] reviewed 1991-1996 aircraft SO2 measurements over the Pacific. They suggested that anthropogenic activities along the North Pacific rim and volcanic sources along the western Pacific rim contribute significantly to freetroposphere SO2 concentrations. The contribution from DMS, emitted by oceans, appears to dominate in the boundary layer. Quantifying the link between oceanic DMS and free troposphere SO2 is, however, important because of the implications on climate feedback [Charlson et al., 1987]. We will investigate how much free-troposphere SO2 during PEM-Tropics B may be attributed to DMS oxidation. The widely adopted rate constants for the DMS-OH adduct reaction by Hynes et al. [1986] were derived from experiments conducted for atmospheric conditions near surface. We will derive rate constants more appropriate for free tropospheric conditions on the basis of recent low-temperature and low-pressure measurements by Hynes et al. [ 1995] and Barone et al. [ 1996].
We will analyze observations of 15øS-15øN from the DC-8 flights to understand chemistry over the tropical Pacific; a large number of industrial plumes observed north of 15øN are therefore excluded. During the return flights from Easter Island to NASA Dryden Flight Research Center, highly polluted continental outflow was sampled. We excluded these last two flights from our analysis. Instruments used on board DC-8 are described by perspective investigators in this issue. We will take into account instrument sensitivities in our analysis. We describe in section 2 the 1-D model applied in the analysis. In section 3, we compare the chemical characteristics between PEM-Tropics A and B and examine the effects of long-range transport of pollutants over the tropical Pacific. In section 4, we analyze using the 1-D model the convective turnover timescale constrained by CH3I observations and the impact of convection on 03, HOx, NOx, DMS, and SO2 concentrations.
Conclusions are given in section 5.

One-Dimensional Model
Concentrations of chemical species often change rapidly with altitude. This rapid change reflects in part the changing photochemical environment with altitude and in part the relatively slow transport in the vertical compared with that in the horizontal. The vertical distribution of chemical species often provides a sensitive test for our understanding of chemistry and transport. A 1-D model is suited for such analysis. The 1-D model used in this work is described by Wang et al. [2000]. It is based on the model by Trainer et al. [1987Trainer et al. [ , 1991 and McKeen et al. [ 1997]. The vertical transport of the model was originally computed using a diffusion scheme, in which the rate of tracer transport depends on the diffusion coefficient Kz and the vertical gradient of the tracer [e.g., Liu et al., 1984]. The diffusion scheme alone, however, cannot represent vertical mixing of the tropospheric column in the tropics [Wang et al., 2000], where convection dominates. During convection, air in the lower troposphere is lifted by updrafts into the free troposphere and air in the free troposphere is brought by subsidence to the lower troposphere. This direction-oriented transport differs from the underlying assumption of random motions in all directions in diffusion transport. Wang et al. [2000] implemented an explicit scheme of convective transport using specified convective mass fluxes. A diffusion coefficient is specified in the model to reproduce the observed vertical gradient of CH3I concentrations below 2 km (section 4). Soluble species H202 and HNO3 are scavenged during convective transport [Wang et al., 2000].
The model extends to 16 km with decreasing vertical resolutions from 10 m near the surface to 1 km at the ceiling of the DC-8 aircraft (12-km altitude). The time step for transport and chemistry is 30 s. The model is run for 60 days to obtain steady state results. The kinetics data are taken from DeMore et al. [1997] and Atkinson et al. [1997], with updates from Sander et al. [2000]. The rate constants for the DMS-OH adduct reaction are uncertain, particularly for free tropospheric conditions [DeMore et al., 1997]. We derive new rate constants from more recent kinetics measurements by Hynes et al. [1995] and Barone et al. [1996] and compare those to the widely adopted rate constants from Hynes et al. [1986] in section 4.4. We did not include hydrolysis of N205 on aerosols due in part to the uncertainties in the calculation of aerosol surface areas [Schultz et al., 2000]; its implications on the budget of NOx in the model are discussed in section 4.3. The total ozone column is specified to 262 Dobson Unit (DU), the average observed by the Total Ozone Mapping Spectrometer during PEM-Tropics B. The oceanic surface albedo is specified to i 32,735 from fresh emissions. During PEM-Tropics A in austral spring, chemistry over the southern tropical Pacific was strongly perturbed by pollutants from biomass burning Schultz et al., 1999;Talbot et al., 1999;Wang et al., 2000]. Using C2H2 as a tracer, Wang et al. [2000] estimated that biomass burning outflow enhances 03 concentrations, 03 production, and concentrations of NOx and OH by 60, 45, 75, and 7%, respectively.
Air masses with enhanced concentrations of C2H2 were also observed during PEM-Tropics B in austral fall [Blake et al., this issue]. However, this enhancement shows distinctively different chemical characteristics from those of PEM-Tropics A indicating a different origin from biomass burning. Figure 1 compares the correlation of C2H 2 and C2C14 at 15øS -15øN between PEM-Tropics A and B. Tetrachloroethene is emitted from dry cleaning and industrial usage [McCulloch and Midgley, 1996] and is therefore a good industrial tracer. Its chemical lifetime is -3 months in the tropics. Acetylene is emitted from either fossil fuel combustion and industry or biomass burning and has a lifetime of a few weeks in the tropics. Figure 1 shows a positive C2H2-C2C14 correlation during PEM-Tropics B, implying that C2H2 originated from northern industrialized continents. The scattering in the data reflects the difference in the production processes of C2H 2 and C2C14. Their industrial origin is further supported by decreasing concentrations of C2H 2 and C2C14 as well as CO, C2H6, and C3H 8 with latitude from north to south [Blake et al., this issue]. In contrast, C2H 2 concentrations in the northern tropics were close to background levels and much lower than in the southern tropics during PEM-Tropics A [Wang et al., 2000]. Furthermore, concentrations of C2H2 were mostly independent of comparatively low C2C14 concentrations during PEM-Tropics A, consistent with a biomass burning origin of C2H 2 in austral spring.
We use C2H2 as a tracer for industrial pollutant outflow and group PEM-Tropics B observations into three equally divided tertiles to analyze further air chemical characteristics in the region. Figure 2 shows the vertical profiles of median concentrations of C2H2, CO, acetone, 03, OH, and NO for the lower, middle, and upper tertiles of C2H 2. The median concentrations of C2H2 in the lower tertile are only ~ 10 pptv, about factors of 2 and 4 lower than medians in the middle and upper tertiles, respectively. Concentrations of CO show a similar trend. The median concentrations of CO in the lower tertile of C2H 2 is -45 ppbv, only a few ppbv above the background CO concentration of 40 ppbv from CH 4 oxidation. Median CO concentrations in the middle and upper tertiles of C2H2 are significantly higher than the 40-ppbv level expected from CH 4 oxidation. The grouping of observations by C2H2 concentrations therefore separates air masses strongly influenced by pollutant outflow from those relatively clean and can be used to estimate the impact of pollutant outflow. The concentrations of CO and C2H2 exhibit generally decreasing concentrations with altitude, which was not observed during PEM-Tropics A [Wang et al., 2000], implying that the lower troposphere is an important conduit for the long-range transport of northern industrial plumes but not for southern biomass burning plumes.
The median concentrations of acetone are moderately higher in higher tertiles of C2H 2 (Figure 2) reflecting in part production of acetone from oxidation of anthropogenic hydrocarbons such as propane, the concentrations of which correlate well with C2H2 [Blake et al., this issue]. However, the observed enhancements of 0-200 pptv due to pollutant import over a background of 400 pptv are marginal considering that standard deviations are -100 pptv (not shown). These measurements suggest that the sources of acetone are located largely in the tropical region, most likely of biogenic origin [Singh et al., 1994[Singh et al., , 2001Wang et al., 1998b].

Timescale of Transport and Its Effects on Tropical OH Concentrations
During PEM-Tropics A, biomass burning outflow was observed to enhance significantly not only reduced compounds like C2H 2 and CO but also 03 and NO concentrations; Short-lived NOx is depleted in the long-range transport during PEM-Tropics B but longer-lived NOt, CO, and C2H2 are not. During PEM-Tropics A, on the other hand, the timescale of transport of biomass burning outflow is short enough that NOx concentrations were also enriched. Quantitative assessments of transport timescales from available observations require a more rigorous treatment that we will undertake in a subsequent study. We will suggest in section 4.3 that NOx observed during PEM-Tropics B is mainly from lightning.
Concentrations of OH in general depend strongly on concentrations of NOx, CO, and hydrocarbons. Higher NOx concentrations shift HO2 towards OH and tend to increase OH concentrations. Carbon monoxide and hydrocarbons have the opposite effects. Davis   Model results show a similar trend but simulated column concentrations are 15% lower than observations. The underestimate is due to that in the upper troposphere (section 4.2).

Concentrations of OH were not measured during PEM-Tropics
A; we use model results here. Table 1 shows a small increase of 7% from the lower to upper tertile of C2H 2. The shorter transport timescale during PEM-Tropics A allowed concurrent increases of short-lived NOx with longer-lived CO and hydrocarbons in air masses influenced by biomass burning outflow. These two factors have offsetting effects resulting in a small enhancement of OH concentrations.

Convective Turnover Timescale
A good chemical tracer for convection over the tropical ocean is CH3I Cohan e! al., 1999;Wang et al., 2000]. Emitted from the ocean, CH3I has a short lifetime of-3 days against photolysis in the tropics. The observed vertical profile of CH3I in the free troposphere reflects a balance between the supply by convective transport and the loss by photolysis and offers good constraints on the rate of convec-1-D simulation of CH3I using the GCM convective statistics shows a C-shaped profile with higher concentrations in boundary layer (due to emission) and in the upper troposphere (due to convective transport). Observations of CH3I during PEM-Tropics A, in comparison, show little altitude dependence in the free troposphere, suggesting that oceanic convective outflux distributes evenly with altitude. Figure 3 shows the median profile of CH31 observed during PEM-Tropics B. Observed CH31 concentrations again show little altitude dependence in the free troposphere corroborating with our finding from PEM-Tropics A.
We estimated a 20-day turnover timescale during PEM-Tropics A on the basis of observed CH3I observations. Figure   3 shows that the diffusion transport (K__ = 10 m s -l, the same as for PEM-Tropics A) alone captures the vertical gradient in the boundary layer but grossly underestimates CH3I concentrations in the free troposphere. The mass fluxes adjusted to simulate PEM-Tropics A observations lead to large underestimates of CH3I concentrations in the free troposphere. A doubling of the mass fluxes is necessary to explain PEM-Tropics B CH3I observations. It implies a 10-day turnover timescale of the free troposphere due to convection during PEM-Tropics B. with observations for the three tertiles of C2H2 separately because observed HNO3 and PAN concentrations vary with C2H2 tertiles.

Hydrogen Oxides and 03
Hydroxyl radicals are the key species for the oxidation of reduced compounds such as CO and hydrocarbons in the troposphere [IPCC, 1996]. During oxidation, OH is convened to HO2, which is recycled back to OH by reacting with NO. A key intermediate product of hydrocarbon oxidation is CH20. The lifetime of CH20 is relatively short (5-10 hours) in the tropics. Its concentration therefore serves as a good gauge for photochemical activities. However, previously large discrepancies have been found between model simulated and observed CH20 concentrations [e.g., Fried et al., 1997]. Most of CH20 is produced from CH4 oxidation during PEM-Tropics B. A substantial fraction of data points are below the limit of detection (LOD) of 50 pptv. The LOD fraction increases from 30% at 7 km to 50% at 12 km ( Figure 5). When the LOD fraction is >30%, we show only the median values. When the LOD fraction is >50% at 9-10 km, we use the LOD limit of 50 pptv for the median of the data bin, which is an upper limit. Model simulated CH20 concentrations generally agree with observations but with a consistent high bias tions due to the offsetting effects of convective transport of CH3OOH and scavenging of H20 2. One convective effect not taking into account here is the convective transport of H20, which is reflected in the observed H2 ¸ concentrations. Its effect on the HOx primary sources is unlikely to be significant above 12 km due to the large contribution by convectively transported CH3OOH. However, convective transport of H20 and subsequent moistening of the atmosphere could substantially boost the primary HOx production in the middle and lower troposphere because of the dominance of the O(ID)+H2 ¸ reaction (Figure 7). Our model, however, is inadequate to assess this effect. Figure 9 shows model diagnosed median 03 production and loss rates and a deficit in the 03 budget. Column 03 production at 0-16 km is 1.3x10 • molecules cm -2 s -• only ~50% of the column loss (

Free Tropospheric DMS and SO2
Dimethylsulfide in the atmosphere is oxidized mostly by OH and to a lesser extent by NO3 [Berresheim et al., 1995]. Whereas the rate constants for H abstraction reaction by the radicals are reasonably known, kinetics data for the adduct reaction of DMS and OH are uncertain [DeMote et al., 1997].
While the former reaction is bimolecular, the latter reaction involves the DMS-OH adduct that either decomposes or reacts with 02 and hence is more complex. The bulk rate constants for the adduct reaction were derived by Hynes et al. [1986] for near surface atmospheric conditions. Their empirical fitting function has since been widely adopted [e.g., ,4tkinson et al.
where T is in Kelvin. Barone et al. [1996] also showed pressure dependence of ki and k2. We include this uncertainty by assuming a linear pressure dependence to compute another set of rate constants k', i.e., k'l=kl*(P/100) and k'2=k'2*(P/100), where P is in Torr.

For the value of k3, Hynes et al. [1995] and Barone et al.
[1996] found it to be independent of T and P at 8x10 -•3 and 10x10 -13 cm 3 molecule -I s -l, respectively. We adopt k3=9x10 -13 cm 3 molecule -I s -I We therefore derive a rate expression for the adduct channel,

Figure 13 compares these rate constants to those for the abstraction channel and the addition reaction rate constants by
Hynes et al. [1986]. The H-abstraction rate constant decreases with altitude while the addition rate constant generally increases; the latter channel is more important or even dominant in the free troposphere. The DMS-OH adduct reaction accounts for 20, 30, and 40% in boundary layer and 75, 60, and 75% at 8 kin, respectively, for the rate constants by Hynes et al. [1986] and of (5) and (6). Equation (6) with the pressure dependence is expected to give faster rate constants than (5). The relatively good agreement between that from Hynes et al.
[1986] and (6) in the free troposphere is fortuitous since the rate expression given by Hynes et al. [1986], strictly speaking, applies only for 1 atmosphere pressure whereas (6) is pressure dependent. The new measurements at low pressures and temperatures imply higher branching ratios for the addition channel in the boundary layer than predicted by Hynes et al. [1986]. Our primary interest is in DMS and SO2 concentrations in the free troposphere. We will adopt in our 1-D simulations equations (5) and (6) to bracket the range of addition rate constants.
In the simulation of DMS we specify its concentration at 500 m as observed. The resulting DMS concentrations in the free troposphere reflect the source from convective transport compensated for by oxidation by OH and NO 3. Figure 14 shows that 50-90% of the observations are below the LOD at altitudes above 3 km, where the model predicts concentrations of 5-10 pptv. The observed medians in the free troposphere are lower than plotted values since we used the LOD value as an upper limit for data bins with LOD fraction > 50%. To narrow The virtually unobservable level of DMS in the free troposphere, at first sight, implies that DMS cannot be a major source of SO2 outside the boundary layer. This issue is further masked by other sources apparently making contributions to free tropospheric SO2 including anthropogenic and volcanic sources [Thornton et al., 1999]. However, convection can transport into the free troposphere large amounts of insoluble DMS. It is reasonable to assume that SO2 is produced from transported DMS in the free troposphere. The yields of SO2 from DMS oxidation are yet another uncertainty [e.g., Davis et al., 1999]. We adopt oxidation yields of 1 for OH and NO 3 abstraction reactions and 0.6 for the DMS-OH adduct reaction Koch et al., 1999]. Figure 15 shows the observed median SO2 profile and model results using the rate expressions of (5) and (6), respectively. To simplify the interpretation of the model results, we assume a complete scavenging of SO2 during convective transport and specify the SO2 concentration at 500 m as observed. The latter led to the kink at 500 m in the model results. These measures ensure that free tropospheric SO2 simulated in the model are produced from DMS oxidation. The difference in the yield of SO2 between the abstraction and addition channels is reflected in the general decrease of model simulated SO2 concentrations with altitude as the addition channel becomes more important (Figure 13). It is also reflected in the lower SO2 concentrations obtained using the faster rate expression of (6) than using that of (5) for the addition channel. Model results suggest that convective transport of DMS from the boundary layer could explain the observed level of SO2 concentrations in the free troposphere over the tropical Pacific.

Conclusions
We analyzed observations from PEM-  Convection during PEM-Tropics B was more vigorous than during PEM-Tropics A. Using CH31 as a tracer for convective transport in a 1-D model, we found a turnover timescale of 10 days by convection during PEM-Tropics B, twice as fast as during PEM-Tropics A. Observations of CH31 from PEM-Tropics A and B suggest that convective outflow in the tropics is evenly distributed with altitude up to the DC-8 flight ceiling of 12 km, implying stronger outflow in the middle troposphere than predicted in a general circulation model [Wang et al., 2000]. Observations of CH3I above 12 km are necessary to examine if such a distribution profile of convective outflow extends into the top layer of the tropical troposphere.
Model simulated HO2, CH20, H202, CH3OOH, and PAN concentrations showed reasonable agreement with the observations. Although simulated CH20 concentrations are consis-,,•,,,1,, higher h,, ~9q nnt,7 than c•h•or-wocl mocli•nq thiq agreement is still noteworthy in light of large discrepancies found in previous field campaigns. Proper care needs to be taken to account for CH20 data below the LOD. Despite of the good agreement for HO2, model simulated OH concentrations were consistently lower (~25%) than observations above 6 km. The agreement between simulated and observed peroxides is not as good as that for PEM-Tropics A; model simulated CH3OOH are consistently higher by ~ 100 pptv. The discrepan-cies between simulated and observed OH and peroxides are, however, within the measurement uncertainties. Unlike the model overestimates of HNO3 in previous studies, simulated HNO3 concentrations during PEM-Tropics B were much lower than observations. Invoking N205 hydrolysis in aerosols could mitigate the underestimates. It is, however, difficult to reconcile the difference between PEM-Tropics A and B. During PEM-Tropics A the model predicted concentrations of HNO3 close to or moderately higher than the observations without invoking N205 hydrolysis in aerosols.
The comparatively faster convective transport and less influence from long-range transport of pollutants during PEM-Tropics B than A makes PEM-Tropics B a good case for studying the effects of convection on chemistry over the tropics. Convective transport of CH3OOH was the dominant primary HO x source above 10 km. The corresponding increase in HOx and 03 production in the upper troposphere was offset by the decrease in the lower and middle troposphere due to convective scavenging of another HO x reservoir, H202. Convective transport increased column 03 production by 4% but decreased column HO x production and OH concentrations by 7 and 4%, respectively. We could not analyze the effect of convection on enhancing H20 concentrations in our model. Convection could in this sense be a primary driver for HO x chemistry in the tropical troposphere since the reaction of O(]D) and H20 provides the dominant primary HO x source in the middle and lower troposphere.
The column budget of 03 had a large deficit; column production was only half of the loss. The deficit is likely compensated for by transport of 03 from other regions of the tropics since transport from northern industrial continents had little impact on 03 and NOx concentrations. Convective transport of reactive nitrogen also imposed a budget deficit of NOx. The loss of upper tropospheric HNO3 and PAN due to the subsiding branch of convection could not be compensated for by the ascending branch because of convective scavenging of HNO3 and very low concentrations of PAN in the lower troposphere. In the lower troposphere, thermal decomposition of PAN transported by subsidence provided the necessary NO.,. source to account for scavenging and deposition of HNO3. The implied NOx source in the free troposphere (above 4 km) is 0.4-1 Tg N yr -] over the tropics (20øS -20øN). The likely sources are in situ marine lightning or transport of NO x from continental lightning. The former source is supported by GEOS observations of marine cumulus clouds along the calculated back trajectories of high NOx encounters. The magnitude of this source highlights the need to understand NO x production by marine lightning.
Concentrations of DMS were barely detectable in the free troposphere with 60%-100% observational data below the LOD (1 pptv). The widely adopted rate constants by Hynes et al. [1986] were measured for atmospheric conditions near the surface. For free tropospheric studies, we derived new rate Our results suggest that caution must be exercised when using observed very low concentrations of DMS as evidence of little contribution by DMS to free tropospheric SO2. Although more work is needed to clarify potential DMS oxidation pathways, it is clear that the amount of DMS transported by convection was large enough to explain observed SO2 concentrations in the free troposphere over the tropical Pacific.