The atmospheric sulfur cycle over the Amazon Basin: 2. Wet season

,

turbulent boundary layer of the aircraft.The flow rates and sample volumes for all species investigated were measured with integrating mass flowmeters (Teledyne Hastings-Raydist).At ground level, air was pulled directly into the DMS and H2S preconcentration devices.
The techniques used for sampling and determination of atmospheric DMS have been described in detail previously [Andreae et al., 1985] 520 nm emission wavelengths.The analytical procedure was calibrated prior to each analysis period using freshly prepared standard solutions of sodium sulfide.All samples were analyzed within 8 hours after collection.The results were corrected for the interference caused by the hydrolysis of atmospheric COS on the impregnated filters [Cooper and Saltzman, 1987].This correction was made by placing two impregnated filters in series.Since the conversion efficiency of COS is only 1-2%, the amount of COS converted on the second filter is essentially identical to that on the first, and an accurate correction can be made by subtracting the value obtained on the second filter from the total sulfide concentration measured on the first filter.The average COS correction was 10 ppt, about 20% of the H2S signal in the boundary layer.
SO2 samples were collected from the same manifold, using a filter pack with a K2CO3-impregnated filter preceded by a Teflon filter to remove aerosols.To trap SO2, we used a fast-flow cellulose filter (Schleicher and Schuell, FF 2) impregnated with K2CO3/glycerol.The preparation, storage, and blank behavior of the impregnated filters have been described in detail by Daum and Leahy [1983].The collection efficiency and capacity of the impregnated filters for SO2 were studied in laboratory and field experiments [Andreae et al., 1988;Berresheim et al., 1990].Under the meteorological conditions prevailing during ABLE 2B and using sampling periods of 30-120 min, the collection efficiency of the impregnated filters was >90% in the boundary layer and >80% in the free troposphere.
The aerosol sampling system and procedures are described in detail elsewhere [Talbot et al., 1988, this

Reduced Sulfur Species: Aircraft Measurements
In Table 1 we present a summary of the sulfur species concentrations over the Amazon Basin during the wet and dry seasons.In the case of DMS, we were able to obtain a comprehensive data set for both seasons; technical problems made MeSH measurements during the wet season impossible, and for H2S we have ground-level data only for the dry season and aircraft data only for the wet season.
The concentrations of DMS in the wet and dry seasons were identical within the variability of the data.This is evident in Figure 1, where the data from both seasons are plotted versus altitude.During most flights, the concentration of DMS was almost constant from the lowest sampling level (about 150 m) to the top of the mixed layer (about 300 to 1500 m, depending on meteorological conditions and time of day).It decreased through the cloud convection layer (CCL) to values near 1 ppt in the free troposphere.DMS showed no statistically significant diel variation in the mixed layer, although there was a tendency for high concentrations to occur between 1000 and 1200 local time (LT), which was probably due to fast transport of DMS upward from the forest into the growing mixed layer.
The cross-basin survey between Manaus and Be16m during the wet season (April 23-24, 1989) confirmed the results of the dry season experiment (Figure 2).Over the Amazonas Estuary near Be16m, DMS levels exceeded 70 ppt (only a lower limit can be given, as the instrument went off-scale at 70 ppt); they decreased rapidly between the coast (near 48øW) and 52øW.West of this longitude, values typical of the interior Amazon Basin were measured.The steep longitudinal gradient observed here cannot be used as evidence for a short atmospheric lifetime of DMS, however.Streamline charts show that airmass transport was nearly parallel to the coast (Figure 3), so that marine air was sampled only very close to the coast.Consequently, the rapid transition from predominantly marine to predominantly continental air masses along the flight track, rather than fast photochemical oxidation of DMS, appears to be responsible for the observed sharp geographical gradient in DMS concentration.

Sources and Fluxes of Reduced Sulfur Species
In order to investigate the source mechanisms and fluxes of reduced sulfur species into the atmosphere from the Amazon forest ecosystem, we conducted measurements using soil chambers and a micrometeorological mast at the Ducke Forest Reserve.Both soils and plants have been shown to emit sulfur gases (for a recent review, see Andreae [1990]).We measured the emissions of DMS and H2S from the yellow clay soil dominant in the region at two sites in the forest near the micrometeorological tower.The measurement sites are described in more detail by Bakwin et 1 for the wet season boundary layer represents a mean concentration of aerosol sulfate throughout the study.This mean value is, however, the result of compounding data from two rather distinct distributions, as the composite vertical plot in Figure 7  The chemical evolution of the boundary layer is computed over a three-day simulation period, starting from initial concentrations representative of the free troposphere (Table 1) which are also taken as fixed upper boundary conditions at   layer.The simulated concentration of SO2 is 13 ppt, about half the average observed SO2 concentration of 24 ppt (Table 1).However, the main sink for SO2 in the model is aqueous This perturbation of the sulfur cycle in the tropics is even more evident in tropical Africa, where the rainwater concentrations of sulfate and other species derived from biomass burning are substantially higher than in Amazonia.In the rain forest of the northern Congo, for example, J.P. Lacaux (personal communication, 1988) found a mean rainwater sulfate concentration 5 times higher than in Amazonia.Together with increased amounts of nitric and organic acids, this may result in significant acidification of tropical ecosystems [McDowell, 1988].

No major differences are evident between the sulfate concentrations we measured using airborne sampling and ion-chromatographic analysis and those obtained by Artaxo et al. [1988, this issue] using ground-level sampling and analysis by photon-induced X ray emission (PIXE). This suggests that airborne sampling did not introduce any substantial bias. The sulfate value given in Table
. DMS was preconcentrated by adsorption to gold wool contained in quartz glass tubes.Duplicate samples were collected simultaneously; the flow rate through each sampling tube was regulated to about 2 L min -•.The sampling time varied from 10 min in the mixed layer and at grotlnd level to about 1 hour at the highest flight levels.Scrubber tubes filled with 5% Na2CO3 on Anakrom C22 (40/50 mesh) were placed in the sample airstream before the gold adsorption tubes.Freshly filled tubes were used for each flight.Laboratory experiments and the results of a recent intercomparison experiment have shown that these scrubber tubes are equally effective in preventing negative artifacts in the determination of DMS as other ozoneremoval devices (M.O. Andreae, unpublished data, 1989).All samples were analyzed within 6 hours after each flight.Previous studies had shown that the DMS on the gold surface is stable for at least 10 days if the samples are stored in the dark.The determination of DMS consisted of thermal desorption followed by cryogenic trapping, chromatographic separation, and flame photometric detection [Andreae et al., were placed in polyethylene vials and extracted immediately following the flights with 10 mL of deionized water.(The Zefluor filters were first wetted with 1 mL of methanol.)The impregnated filters were treated with 10 mL of 0.06% H202, whereby all sulfur (IV) species adsorbed on the filter were dissolved and converted to sulfate.Rainwater samples were collected on an event basis by setting out polyethylene collectors during active precipitation only.The samples were placed in polyethylene bottles which had been rinsed with deionized water.Small amounts (about 1 mL) of chloroform were added to prevent microbial growth.Further details on the sampling sites and protocol can be found in the work by Andreae et al. [this issue].Sulfate in both aerosol and impregnated filter extracts was determined by ion chromatography using the Dionex HPICexperiment took place in April/May 1987, during the late part of the wet season.The large-scale meteorological conditions prevailing during the experiment have been described by Harriss et al. [this issue] and M. Garstang et al.(The Amazon Boundary Layer Experiment (ABLE 2B): A meteorological perspective, submitted to Bulletin of the American Meteorological Society, 1989).The circulation was dominated by flow from the east to southeast, with the Intertropical Convergence Zone (ITCZ) either passing over Manaus or lying slightly to the north.An anticylcone, present for most of the time over central Brazil, with its center near Brasilia, caused inflow of air from the dry tropical regions of central Brazil into the Amazon Basin.Occasional shifts in the position of the ITCZ and the strength of the northern hemisphere subtropical anticyclone resulted in the injection of pulses of air from the northern hemisphere, which may have introduced dust from the Saharan region [Talbot et al., this issue (a)].This dust is characteristically associated with aerosol sulfate and nitrate originating from pollution and/or biomass burning [Savoie et al., 1989].16,815

Fig. 1 .Fig. 2 .
Fig. 1.Vertical distribution of dimethylsulfide (DMS) in the atmosphere over the central Amazon Basin during the dry and wet seasons.For the wet season, the results of the model simulation are superimposed (see text).

Fig. 4 .
Fig. 3. Streamline charts for the 700-hPa level and weather synopsis over western South America for April 23 and 24, 1987.

Fig. 5 .Fig. 6 .Fig. 7 .Fig. 8 .
Fig. 5. Flux of DMS between the Amazon forest canopy and the overlying atmosphere during the wet season.The squares indicate the flux estimates obtained from the observed gradients and exchange coefficients.Error bars show one standard deviation for time intervals where multiple measurements were available.The solid and dashed lines represent the source characteristics used in the model (see text).
Fig. 9a Fig. 10.Simulated concentrations of DMS (ppt) between 0 and 50 m, as a function of time of day.
phase oxidation, which is poorly constrained.The production rate of SO2 in the model (10 ppt d -j) suggests that biogenic emissions from the forest must account for a significant fraction of the SO2 levels observed in ABLE 2B.Other sources (e.g., transport from outside the Basin) could also be important.The average sulfate concentration predicted by the model in the boundary layer (9 ppt) is at the low end of observed values (Figure7) and is in reasonable agreement with the background concentration of 12 ppt derived from the observations after removing the marine source characterized by Na +, and the anthropogenic source characterized' by NO•-(equation (1)).It appears therefore that the background sulfate concentrations observed during ABLE 2B can be explained by biogenic emissions from the forest.Under most conditions, however, this biogenic background is small compared to the marine and anthropogenic contributions (Figures7 and 8).A significant local source of SO2 from the photochemical oxidation of reduced biogenic sulfur gases and, in contrast, a predominantly advective origin of sulfate from long-range transport are consistent with the ab.•ence of any significant correlation between SO2 and sulfate in our data set.CONCLUSIONSThe results of our studies on the sulfur cycle over Amazonia have led us to reevaluate roles played by bi0genic emissions, marine aerosols, and long-range transport of combustion-derived st•lfate aerosols in controlling the concentrations of sulfur species over the remote tropical continents.The reduced sulfur gas emissions from soils were low during the wet season, the DMS flux was about 0.04 nmol m -2 min -•, the H2S flux less than 0.03 nmol m -2 min -•.These fluxes are at the low end of the range observed from soils in the United States [Goldan et al., 1987; Lamb et aT., 1987].The DMS flux from the forest canopy, obtained by a gradient-flux technique, was 0.46 _+ 0.20 nmol m -2 min -•, similar to values measured in North America.The H2S flux from the canopy was not measured directly.Model calculations based on the concentrations of DMS, MESH, and H2S in the planetary boundary layer (PBL) suggest that the canopy emission fluxes for these species were 0.2, 0.1, and 1.1 nmol m -2 min -• respectively for a total reactive reduced sulfur flux of about 1.4 nmol m -2 min -• .This flux is somewhat lower than our dry season estimate of 2.9 nmol m -2 min -• and in reasonable agreement with the measurements made in North America by Lamb et al. [ 1987].When the ventilation of the reduced sulfur gases and SO 2 from the boundary layer into the free troposphere and the dry deposition of SO2 are taken into account, the-oxidation of biogenic sulfur species can account for a , nitrate, and sodium in aerosol and rain, as described by multivariate regression analysis, suggest that marine aerosols, including sulfate from the oxidation of marine biogenic dimethylsulfide, and combustion-derived sulfate aerosols brought into the Amazon Basin by long-range transport account for the remaining •90% of sulfate standing stock and deposition.Transport from outside the Amazon Basin has been implicated in the ABLE 2B observations for other species besides SOl, e.g., NOy [Bakwin et aT., this issue (b); Jacob and Wofsy, this issue], aerosol [Talbot et aT., this issue (a)], and organic acids [Talbot et aT., this •ssue (b)].Based on our measurements in the wet and dry season, we estimate a mean annual reduced sulfur gas emission of about 2 _+ 1 nmol m -2 min -• from the rain forest ecosystems of central Amazonia.The weak geographical gradients in sulfur species concentrations Over Amazonia observed during both expeditions suggest that this estimate may be representative for all of the wet tropical regions of South America.In fact, it may be valid for the wet tropics worldwide, as suggested by our sulfur gas measurements in the PBL over the rain forest of equatorial Africa.Over the Congo forest we observed slightly lower concentrations of DMS and H2S (5.3 -4.5 and 25 _+ 16 ppt, respectively) than over Amazonia (H.G. Bingemer et al., Measurements of sulfur gases and aerosols in and above the Equatorial African rain forest, submitted to Journal of Geophysical Research, 1990).But in view of the somewhat shorter lifetime of the sulfur gases over Africa caused by higher OH levels, these concentra-conclude that the wet tropical continents make only a small contribution to the global atmospheric sulfur cycle, since they represent only about 3% of the Earth's surface and have a relatively low sulfur emission rate.The emissions from tropical wet forests account for only about 1% of global biogenic sulfur emissions.This is especially modest in comparison with anthropogenic emissions from fossil fuel burning (about